rabbitmq-server/deps/rabbit/test/backing_queue_SUITE.erl

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

1888 lines
73 KiB
Erlang
Raw Permalink Normal View History

%% This Source Code Form is subject to the terms of the Mozilla Public
%% License, v. 2.0. If a copy of the MPL was not distributed with this
%% file, You can obtain one at https://mozilla.org/MPL/2.0/.
%%
%% Copyright (c) 2007-2024 Broadcom. All Rights Reserved. The term “Broadcom” refers to Broadcom Inc. and/or its subsidiaries. All rights reserved.
%%
-module(backing_queue_SUITE).
-include_lib("common_test/include/ct.hrl").
-include_lib("amqp_client/include/amqp_client.hrl").
Make #amqqueue{} a private record + add a feature flag The #amqqueue{} record is now isolated in the `amqqueue` module: all accesses and modifications to the record are made through this helper. The type `rabbit_types:amqqueue()` is now `amqqueue:amqqueue()`. `amqqueue` knows about the new #amqqueue{} record introduced with quorum queues. If the actual record is the old one (as defined in RabbitMQ 3.7.x), it calls `amqqueue_v1` which knows about the old definition. When it needs to produce an #amqqueue{} record, it verifies if it is allowed to, using the new Feature flags subsystem and the `quorum_queue` feature flag specifically, and proceeds. If an old format #amqqueue{} is required, the creation is deferred to `amqqueue_v1`. The new modules come with a couple headers: they provide macros to replace pattern matching with `when` conditions. The `amqqueue_v*.hrl` headers are generated using the `macros()` functions in `amqqueue` and `amqqueue_v1` modules. Some operations now depend on the state of the `quorum_queue` feature flag. In particular, creating a quorum queue is denied if the feature flag is disabled. In the process, the following modules were moved from rabbitmq-common to rabbitmq-server: * rabbit_backing_queue * rabbit_queue_master_locator Likewise, the few functions were copied from `rabbit_misc` to `rabbit_amqqueue` because they depend on `amqqueue`. They don't really need to live in rabbitmq-common and they prevent further progress with the feature flags: feature flags need informations about the cluster. The `quorum_queue` feature flag comes with a migration function which is responsible of converting the #amqqueue{} records in Mnesia tables `rabbit_queue` and `rabbit_durable_queue`. Therefore, the upgrade function which did this is removed in this commit. [#159298729]
2018-10-11 18:12:39 +08:00
-include("amqqueue.hrl").
Message Containers (#5077) This PR implements an approach for a "protocol (data format) agnostic core" where the format of the message isn't converted at point of reception. Currently all non AMQP 0.9.1 originating messages are converted into a AMQP 0.9.1 flavoured basic_message record before sent to a queue. If the messages are then consumed by the originating protocol they are converted back from AMQP 0.9.1. For some protocols such as MQTT 3.1 this isn't too expensive as MQTT is mostly a fairly easily mapped subset of AMQP 0.9.1 but for others such as AMQP 1.0 the conversions are awkward and in some cases lossy even if consuming from the originating protocol. This PR instead wraps all incoming messages in their originating form into a generic, extensible message container type (mc). The container module exposes an API to get common message details such as size and various properties (ttl, priority etc) directly from the source data type. Each protocol needs to implement the mc behaviour such that when a message originating form one protocol is consumed by another protocol we convert it to the target protocol at that point. The message container also contains annotations, dead letter records and other meta data we need to record during the lifetime of a message. The original protocol message is never modified unless it is consumed. This includes conversion modules to and from amqp, amqpl (AMQP 0.9.1) and mqtt. COMMIT HISTORY: * Refactor away from using the delivery{} record In many places including exchange types. This should make it easier to move towards using a message container type instead of basic_message. Add mc module and move direct replies outside of exchange Lots of changes incl classic queues Implement stream support incl amqp conversions simplify mc state record move mc.erl mc dlx stuff recent history exchange Make tracking work But doesn't take a protocol agnostic approach as we just convert everything into AMQP legacy and back. Might be good enough for now. Tracing as a whole may want a bit of a re-vamp at some point. tidy make quorum queue peek work by legacy conversion dead lettering fixes dead lettering fixes CMQ fixes rabbit_trace type fixes fixes fix Fix classic queue props test assertion fix feature flag and backwards compat Enable message_container feature flag in some SUITEs Dialyzer fixes fixes fix test fixes Various Manually update a gazelle generated file until a gazelle enhancement can be made https://github.com/rabbitmq/rules_erlang/issues/185 Add message_containers_SUITE to bazel and regen bazel files with gazelle from rules_erlang@main Simplify essential proprty access Such as durable, ttl and priority by extracting them into annotations at message container init time. Move type to remove dependenc on amqp10 stuff in mc.erl mostly because I don't know how to make bazel do the right thing add more stuff Refine routing header stuff wip Cosmetics Do not use "maybe" as type name as "maybe" is a keyword since OTP 25 which makes Erlang LS complain. * Dedup death queue names * Fix function clause crashes Fix failing tests in the MQTT shared_SUITE: A classic queue message ID can be undefined as set in https://github.com/rabbitmq/rabbitmq-server/blob/fbe79ff47b4edbc0fd95457e623d6593161ad198/deps/rabbit/src/rabbit_classic_queue_index_v2.erl#L1048 Fix failing tests in the MQTT shared_SUITE-mixed: When feature flag message_containers is disabled, the message is not an #mc{} record, but a #basic_message{} record. * Fix is_utf8_no_null crash Prior to this commit, the function crashed if invalid UTF-8 was provided, e.g.: ``` 1> rabbit_misc:is_valid_shortstr(<<"😇"/utf16>>). ** exception error: no function clause matching rabbit_misc:is_utf8_no_null(<<216,61,222,7>>) (rabbit_misc.erl, line 1481) ``` * Implement mqtt mc behaviour For now via amqp translation. This is still work in progress, but the following SUITEs pass: ``` make -C deps/rabbitmq_mqtt ct-shared t=[mqtt,v5,cluster_size_1] FULL=1 make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_1] FULL=1 ``` * Shorten mc file names Module name length matters because for each persistent message the #mc{} record is persisted to disk. ``` 1> iolist_size(term_to_iovec({mc, rabbit_mc_amqp_legacy})). 30 2> iolist_size(term_to_iovec({mc, mc_amqpl})). 17 ``` This commit renames the mc modules: ``` ag -l rabbit_mc_amqp_legacy | xargs sed -i 's/rabbit_mc_amqp_legacy/mc_amqpl/g' ag -l rabbit_mc_amqp | xargs sed -i 's/rabbit_mc_amqp/mc_amqp/g' ag -l rabbit_mqtt_mc | xargs sed -i 's/rabbit_mqtt_mc/mc_mqtt/g' ``` * mc: make deaths an annotation + fixes * Fix mc_mqtt protocol_state callback * Fix test will_delay_node_restart ``` make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_3]:will_delay_node_restart FULL=1 ``` * Bazel run gazelle * mix format rabbitmqctl.ex * Ensure ttl annotation is refelected in amqp legacy protocol state * Fix id access in message store * Fix rabbit_message_interceptor_SUITE * dializer fixes * Fix rabbit:rabbit_message_interceptor_SUITE-mixed set_annotation/3 should not result in duplicate keys * Fix MQTT shared_SUITE-mixed Up to 3.12 non-MQTT publishes were always QoS 1 regardless of delivery_mode. https://github.com/rabbitmq/rabbitmq-server/blob/75a953ce286a10aca910c098805a4f545989af38/deps/rabbitmq_mqtt/src/rabbit_mqtt_processor.erl#L2075-L2076 From now on, non-MQTT publishes are QoS 1 if durable. This makes more sense. The MQTT plugin must send a #basic_message{} to an old node that does not understand message containers. * Field content of 'v1_0.data' can be binary Fix ``` bazel test //deps/rabbitmq_mqtt:shared_SUITE-mixed \ --test_env FOCUS="-group [mqtt,v4,cluster_size_1] -case trace" \ -t- --test_sharding_strategy=disabled ``` * Remove route/2 and implement route/3 for all exchange types. This removes the route/2 callback from rabbit_exchange_type and makes route/3 mandatory instead. This is a breaking change and will require all implementations of exchange types to update their code, however this is necessary anyway for them to correctly handle the mc type. stream filtering fixes * Translate directly from MQTT to AMQP 0.9.1 * handle undecoded properties in mc_compat amqpl: put clause in right order recover death deatails from amqp data * Replace callback init_amqp with convert_from * Fix return value of lists:keyfind/3 * Translate directly from AMQP 0.9.1 to MQTT * Fix MQTT payload size MQTT payload can be a list when converted from AMQP 0.9.1 for example First conversions tests Plus some other conversion related fixes. bazel bazel translate amqp 1.0 null to undefined mc: property/2 and correlation_id/message_id return type tagged values. To ensure we can support a variety of types better. The type type tags are AMQP 1.0 flavoured. fix death recovery mc_mqtt: impl new api Add callbacks to allow protocols to compact data before storage And make readable if needing to query things repeatedly. bazel fix * more decoding * tracking mixed versions compat * mc: flip default of `durable` annotation to save some data. Assuming most messages are durable and that in memory messages suffer less from persistence overhead it makes sense for a non existent `durable` annotation to mean durable=true. * mc conversion tests and tidy up * mc make x_header unstrict again * amqpl: death record fixes * bazel * amqp -> amqpl conversion test * Fix crash in mc_amqp:size/1 Body can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0/ ct-system t=java ``` on branch native-amqp. * Fix crash in lists:flatten/1 Data can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp. * Fix crash in rabbit_writer Running test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp resulted in the following crash: ``` crasher: initial call: rabbit_writer:enter_mainloop/2 pid: <0.711.0> registered_name: [] exception error: bad argument in function size/1 called as size([<<0>>,<<"Sw">>,[<<160,2>>,<<"hi">>]]) *** argument 1: not tuple or binary in call from rabbit_binary_generator:build_content_frames/7 (rabbit_binary_generator.erl, line 89) in call from rabbit_binary_generator:build_simple_content_frames/4 (rabbit_binary_generator.erl, line 61) in call from rabbit_writer:assemble_frames/5 (rabbit_writer.erl, line 334) in call from rabbit_writer:internal_send_command_async/3 (rabbit_writer.erl, line 365) in call from rabbit_writer:handle_message/2 (rabbit_writer.erl, line 265) in call from rabbit_writer:handle_message/3 (rabbit_writer.erl, line 232) in call from rabbit_writer:mainloop1/2 (rabbit_writer.erl, line 223) ``` because #content.payload_fragments_rev is currently supposed to be a flat list of binaries instead of being an iolist. This commit fixes this crash inefficiently by calling iolist_to_binary/1. A better solution would be to allow AMQP legacy's #content.payload_fragments_rev to be an iolist. * Add accidentally deleted line back * mc: optimise mc_amqp internal format By removint the outer records for message and delivery annotations as well as application properties and footers. * mc: optimis mc_amqp map_add by using upsert * mc: refactoring and bug fixes * mc_SUITE routingheader assertions * mc remove serialize/1 callback as only used by amqp * mc_amqp: avoid returning a nested list from protocol_state * test and bug fix * move infer_type to mc_util * mc fixes and additiona assertions * Support headers exchange routing for MQTT messages When a headers exchange is bound to the MQTT topic exchange, routing will be performend based on both MQTT topic (by the topic exchange) and MQTT User Property (by the headers exchange). This combines the best worlds of both MQTT 5.0 and AMQP 0.9.1 and enables powerful routing topologies. When the User Property contains the same name multiple times, only the last name (and value) will be considered by the headers exchange. * Fix crash when sending from stream to amqpl When publishing a message via the stream protocol and consuming it via AMQP 0.9.1, the following crash occurred prior to this commit: ``` crasher: initial call: rabbit_channel:init/1 pid: <0.818.0> registered_name: [] exception exit: {{badmatch,undefined}, [{rabbit_channel,handle_deliver0,4, [{file,"rabbit_channel.erl"}, {line,2728}]}, {lists,foldl,3,[{file,"lists.erl"},{line,1594}]}, {rabbit_channel,handle_cast,2, [{file,"rabbit_channel.erl"}, {line,728}]}, {gen_server2,handle_msg,2, [{file,"gen_server2.erl"},{line,1056}]}, {proc_lib,wake_up,3, [{file,"proc_lib.erl"},{line,251}]}]} ``` This commit first gives `mc:init/3` the chance to set exchange and routing_keys annotations. If not set, `rabbit_stream_queue` will set these annotations assuming the message was originally published via the stream protocol. * Support consistent hash exchange routing for MQTT 5.0 When a consistent hash exchange is bound to the MQTT topic exchange, MQTT 5.0 messages can be routed to queues consistently based on the Correlation-Data in the PUBLISH packet. * Convert MQTT 5.0 User Property * to AMQP 0.9.1 headers * from AMQP 0.9.1 headers * to AMQP 1.0 application properties and message annotations * from AMQP 1.0 application properties and message annotations * Make use of Annotations in mc_mqtt:protocol_state/2 mc_mqtt:protocol_state/2 includes Annotations as parameter. It's cleaner to make use of these Annotations when computing the protocol state instead of relying on the caller (rabbitmq_mqtt_processor) to compute the protocol state. * Enforce AMQP 0.9.1 field name length limit The AMQP 0.9.1 spec prohibits field names longer than 128 characters. Therefore, when converting AMQP 1.0 message annotations, application properties or MQTT 5.0 User Property to AMQP 0.9.1 headers, drop any names longer than 128 characters. * Fix type specs Apply feedback from Michael Davis Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * Add mc_mqtt unit test suite Implement mc_mqtt:x_header/2 * Translate indicator that payload is UTF-8 encoded when converting between MQTT 5.0 and AMQP 1.0 * Translate single amqp-value section from AMQP 1.0 to MQTT Convert to a text representation, if possible, and indicate to MQTT client that the payload is UTF-8 encoded. This way, the MQTT client will be able to parse the payload. If conversion to text representation is not possible, encode the payload using the AMQP 1.0 type system and indiate the encoding via Content-Type message/vnd.rabbitmq.amqp. This Content-Type is not registered. Type "message" makes sense since it's a message. Vendor tree "vnd.rabbitmq.amqp" makes sense since merely subtype "amqp" is not registered. * Fix payload conversion * Translate Response Topic between MQTT and AMQP Translate MQTT 5.0 Response Topic to AMQP 1.0 reply-to address and vice versa. The Response Topic must be a UTF-8 encoded string. This commit re-uses the already defined RabbitMQ target addresses: ``` "/topic/" RK Publish to amq.topic with routing key RK "/exchange/" X "/" RK Publish to exchange X with routing key RK ``` By default, the MQTT topic exchange is configure dto be amq.topic using the 1st target address. When an operator modifies the mqtt.exchange, the 2nd target address is used. * Apply PR feedback and fix formatting Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * tidy up * Add MQTT message_containers test * consistent hash exchange: avoid amqp legacy conversion When hashing on a header value. * Avoid converting to amqp legacy when using exchange federation * Fix test flake * test and dialyzer fixes * dialyzer fix * Add MQTT protocol interoperability tests Test receiving from and sending to MQTT 5.0 and * AMQP 0.9.1 * AMQP 1.0 * STOMP * Streams * Regenerate portions of deps/rabbit/app.bzl with gazelle I'm not exactly sure how this happened, but gazell seems to have been run with an older version of the rules_erlang gazelle extension at some point. This caused generation of a structure that is no longer used. This commit updates the structure to the current pattern. * mc: refactoring * mc_amqpl: handle delivery annotations Just in case they are included. Also use iolist_to_iovec to create flat list of binaries when converting from amqp with amqp encoded payload. --------- Co-authored-by: David Ansari <david.ansari@gmx.de> Co-authored-by: Michael Davis <mcarsondavis@gmail.com> Co-authored-by: Rin Kuryloski <kuryloskip@vmware.com>
2023-08-31 18:27:13 +08:00
-compile(nowarn_export_all).
-compile(export_all).
-define(PERSISTENT_MSG_STORE, msg_store_persistent).
-define(TRANSIENT_MSG_STORE, msg_store_transient).
-define(TIMEOUT, 30000).
-define(VHOST, <<"/">>).
-define(VARIABLE_QUEUE_TESTCASES, [
variable_queue_partial_segments_delta_thing,
variable_queue_all_the_bits_not_covered_elsewhere_A,
variable_queue_all_the_bits_not_covered_elsewhere_B,
variable_queue_drop,
variable_queue_fold_msg_on_disk,
variable_queue_dropfetchwhile,
variable_queue_dropwhile_restart,
variable_queue_dropwhile_sync_restart,
variable_queue_ack_limiting,
variable_queue_purge,
variable_queue_requeue,
variable_queue_requeue_ram_beta,
2023-10-04 18:11:54 +08:00
variable_queue_fold
]).
-define(BACKING_QUEUE_TESTCASES, [
bq_queue_index,
bq_queue_index_props,
{variable_queue_default, [parallel], ?VARIABLE_QUEUE_TESTCASES},
bq_variable_queue_delete_msg_store_files_callback,
bq_queue_recover
]).
all() ->
[
{group, backing_queue_tests}
].
groups() ->
2021-10-08 21:26:15 +08:00
Common = [
{backing_queue_embed_limit_0, [], ?BACKING_QUEUE_TESTCASES},
{backing_queue_embed_limit_1024, [], ?BACKING_QUEUE_TESTCASES}
],
V2Only = [
v2_delete_segment_file_completely_acked,
v2_delete_segment_file_partially_acked,
v2_delete_segment_file_partially_acked_with_holes
],
[
{backing_queue_tests, [], [
msg_store,
msg_store_file_scan,
{backing_queue_v2, [], Common ++ V2Only}
]}
].
2021-10-11 18:30:39 +08:00
group(backing_queue_tests) ->
[
%% Several tests based on lazy queues may take more than 30 minutes.
{timetrap, {hours, 1}}
];
group(_) ->
[].
%% -------------------------------------------------------------------
%% Testsuite setup/teardown.
%% -------------------------------------------------------------------
init_per_suite(Config) ->
rabbit_ct_helpers:log_environment(),
rabbit_ct_helpers:run_setup_steps(Config).
end_per_suite(Config) ->
rabbit_ct_helpers:run_teardown_steps(Config).
init_per_group(Group, Config) ->
case lists:member({group, Group}, all()) of
true ->
ClusterSize = 1,
Config1 = rabbit_ct_helpers:set_config(Config, [
{rmq_nodename_suffix, Group},
{rmq_nodes_count, ClusterSize}
]),
rabbit_ct_helpers:run_steps(Config1,
rabbit_ct_broker_helpers:setup_steps() ++
rabbit_ct_client_helpers:setup_steps() ++ [
fun(C) -> init_per_group1(Group, C) end
]);
false ->
rabbit_ct_helpers:run_steps(Config, [
fun(C) -> init_per_group1(Group, C) end
])
end.
init_per_group1(backing_queue_tests, Config) ->
%% @todo Is that test still relevant?
Module = rabbit_ct_broker_helpers:rpc(Config, 0,
application, get_env, [rabbit, backing_queue_module]),
case Module of
{ok, rabbit_priority_queue} ->
rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, setup_backing_queue_test_group, [Config]);
_ ->
{skip, rabbit_misc:format(
"Backing queue module not supported by this test group: ~tp~n",
[Module])}
end;
init_per_group1(backing_queue_embed_limit_0, Config) ->
ok = rabbit_ct_broker_helpers:rpc(Config, 0,
application, set_env, [rabbit, queue_index_embed_msgs_below, 0]),
Config;
init_per_group1(backing_queue_embed_limit_1024, Config) ->
ok = rabbit_ct_broker_helpers:rpc(Config, 0,
application, set_env, [rabbit, queue_index_embed_msgs_below, 1024]),
Config;
init_per_group1(variable_queue_default, Config) ->
rabbit_ct_helpers:set_config(Config, {variable_queue_type, default});
%% @todo These groups are no longer used?
init_per_group1(from_cluster_node1, Config) ->
rabbit_ct_helpers:set_config(Config, {test_direction, {0, 1}});
init_per_group1(from_cluster_node2, Config) ->
rabbit_ct_helpers:set_config(Config, {test_direction, {1, 0}});
init_per_group1(_, Config) ->
Config.
end_per_group(Group, Config) ->
case lists:member({group, Group}, all()) of
true ->
rabbit_ct_helpers:run_steps(Config,
[fun(C) -> end_per_group1(Group, C) end] ++
rabbit_ct_client_helpers:teardown_steps() ++
rabbit_ct_broker_helpers:teardown_steps());
false ->
Config
end.
end_per_group1(backing_queue_tests, Config) ->
rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, teardown_backing_queue_test_group, [Config]);
end_per_group1(Group, Config)
when Group =:= backing_queue_embed_limit_0
orelse Group =:= backing_queue_embed_limit_1024 ->
ok = rabbit_ct_broker_helpers:rpc(Config, 0,
application, set_env, [rabbit, queue_index_embed_msgs_below,
?config(rmq_queue_index_embed_msgs_below, Config)]),
Config;
end_per_group1(_, Config) ->
Config.
init_per_testcase(Testcase, Config) when Testcase == variable_queue_requeue;
Testcase == variable_queue_fold ->
rabbit_ct_helpers:testcase_started(Config, Testcase);
init_per_testcase(Testcase, Config) ->
rabbit_ct_helpers:testcase_started(Config, Testcase).
end_per_testcase(Testcase, Config) when Testcase == variable_queue_requeue;
Testcase == variable_queue_fold ->
rabbit_ct_helpers:testcase_finished(Config, Testcase);
end_per_testcase(Testcase, Config) ->
rabbit_ct_helpers:testcase_finished(Config, Testcase).
%% -------------------------------------------------------------------
%% Message store.
%% -------------------------------------------------------------------
msg_store(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, msg_store1, [Config]).
msg_store1(_Config) ->
%% We simulate the SeqId (used as a message ref for the flying optimisation)
%% using the process dictionary.
GenRefFun = fun(Key) -> V = case get(Key) of undefined -> 0; V0 -> V0 end, put(Key, V + 1), V end,
GenRef = fun() -> GenRefFun(msc) end,
restart_msg_store_empty(),
MsgIds = [{GenRef(), msg_id_bin(M)} || M <- lists:seq(1,100)],
{MsgIds1stHalf, MsgIds2ndHalf} = lists:split(length(MsgIds) div 2, MsgIds),
Ref = rabbit_guid:gen(),
{Cap, MSCState} = msg_store_client_init_capture(
?PERSISTENT_MSG_STORE, Ref),
Ref2 = rabbit_guid:gen(),
{Cap2, MSC2State} = msg_store_client_init_capture(
?PERSISTENT_MSG_STORE, Ref2),
%% check we don't contain any of the msgs we're about to publish
false = msg_store_contains(false, MsgIds, MSCState),
%% test confirm logic
passed = test_msg_store_confirms([hd(MsgIds)], Cap, GenRef, MSCState),
%% check we don't contain any of the msgs we're about to publish
false = msg_store_contains(false, MsgIds, MSCState),
%% publish the first half
ok = msg_store_write(MsgIds1stHalf, MSCState),
%% sync on the first half
ok = on_disk_await(Cap, MsgIds1stHalf),
%% publish the second half
ok = msg_store_write(MsgIds2ndHalf, MSCState),
%% check they're all in there
true = msg_store_contains(true, MsgIds, MSCState),
%% publish the latter half twice so we hit the caching and ref
%% count code. We need to do this through a 2nd client since a
%% single client is not supposed to write the same message more
%% than once without first removing it.
ok = msg_store_write([{GenRefFun(msc2), MsgId} || {_, MsgId} <- MsgIds2ndHalf], MSC2State),
%% check they're still all in there
true = msg_store_contains(true, MsgIds, MSCState),
%% sync on the 2nd half
ok = on_disk_await(Cap2, MsgIds2ndHalf),
%% cleanup
ok = on_disk_stop(Cap2),
ok = rabbit_msg_store:client_delete_and_terminate(MSC2State),
ok = on_disk_stop(Cap),
%% read them all
MSCState1 = msg_store_read(MsgIds, MSCState),
%% read them all again - this will hit the cache, not disk
MSCState2 = msg_store_read(MsgIds, MSCState1),
%% remove them all
CQ shared store write optimisations (#8507) * CQ: Don't use FHC for writes in shared store * CQ: Send confirms when flushing to disk in shared store Before they would only be sent periodically or when rolling over to a new file. * CQ: Fast-confirm when flushing data to disk We know the messages are on disk or were acked so there is no need to do sets intersections/subtracts in this scenario. * Fix a Dialyzer warning * Faster confirms for unwritten messages Instead of having the message store send a message to the queue with the confirms for messages ignored due to the flying optimisation, we have the queue handle the confirms directly when removing the messages. This avoids sending potentially 1 Erlang message per 1 AMQP message to the queue. * Refactor rabbit_msg_file:pread into rabbit_msg_store Also make use of the opened file for multi-reads instead of opening/reading/closing each time. * CQ: Make sure we keep the updated CState when using read_many * CQ shared store: Run compaction on older file candidates The way I initially did this the maybe_gc would be triggered based on candidates from 15s ago, but run against candidates from just now. This is sub-optimal because when messages are consumed rapidly, just-now candidates are likely to be in a file about to be deleted, and we don't want to run compaction on those. Instead, when sending the maybe_gc we also send the candidates we had at the time. Then 15s later we check if the file still exists. If it's gone, great! No compaction to do. * CQ: Add a few todos for later
2023-06-21 02:04:17 +08:00
{ok, _} = msg_store_remove(MsgIds, MSCState2),
%% check first half doesn't exist
false = msg_store_contains(false, MsgIds1stHalf, MSCState2),
%% check second half does exist
true = msg_store_contains(true, MsgIds2ndHalf, MSCState2),
%% read the second half again
MSCState3 = msg_store_read(MsgIds2ndHalf, MSCState2),
%% read the second half again, just for fun (aka code coverage)
MSCState4 = msg_store_read(MsgIds2ndHalf, MSCState3),
ok = rabbit_msg_store:client_terminate(MSCState4),
%% stop and restart, preserving every other msg in 2nd half
ok = rabbit_variable_queue:stop_msg_store(?VHOST),
ok = rabbit_variable_queue:start_msg_store(?VHOST,
[], {fun ([]) -> finished;
([{_, MsgId}|MsgIdsTail])
when length(MsgIdsTail) rem 2 == 0 ->
{MsgId, 1, MsgIdsTail};
([{_, MsgId}|MsgIdsTail]) ->
{MsgId, 0, MsgIdsTail}
end, MsgIds2ndHalf}),
MSCState5 = msg_store_client_init(?PERSISTENT_MSG_STORE, Ref),
%% check we have the right msgs left
lists:foldl(
fun ({_, MsgId}, Bool) ->
not(Bool = rabbit_msg_store:contains(MsgId, MSCState5))
end, false, MsgIds2ndHalf),
ok = rabbit_msg_store:client_terminate(MSCState5),
%% restart empty
restart_msg_store_empty(),
MSCState6 = msg_store_client_init(?PERSISTENT_MSG_STORE, Ref),
%% check we don't contain any of the msgs
false = msg_store_contains(false, MsgIds, MSCState6),
%% publish the first half again
ok = msg_store_write(MsgIds1stHalf, MSCState6),
%% this should force some sort of sync internally otherwise misread
ok = rabbit_msg_store:client_terminate(
msg_store_read(MsgIds1stHalf, MSCState6)),
MSCState7 = msg_store_client_init(?PERSISTENT_MSG_STORE, Ref),
CQ shared store write optimisations (#8507) * CQ: Don't use FHC for writes in shared store * CQ: Send confirms when flushing to disk in shared store Before they would only be sent periodically or when rolling over to a new file. * CQ: Fast-confirm when flushing data to disk We know the messages are on disk or were acked so there is no need to do sets intersections/subtracts in this scenario. * Fix a Dialyzer warning * Faster confirms for unwritten messages Instead of having the message store send a message to the queue with the confirms for messages ignored due to the flying optimisation, we have the queue handle the confirms directly when removing the messages. This avoids sending potentially 1 Erlang message per 1 AMQP message to the queue. * Refactor rabbit_msg_file:pread into rabbit_msg_store Also make use of the opened file for multi-reads instead of opening/reading/closing each time. * CQ: Make sure we keep the updated CState when using read_many * CQ shared store: Run compaction on older file candidates The way I initially did this the maybe_gc would be triggered based on candidates from 15s ago, but run against candidates from just now. This is sub-optimal because when messages are consumed rapidly, just-now candidates are likely to be in a file about to be deleted, and we don't want to run compaction on those. Instead, when sending the maybe_gc we also send the candidates we had at the time. Then 15s later we check if the file still exists. If it's gone, great! No compaction to do. * CQ: Add a few todos for later
2023-06-21 02:04:17 +08:00
{ok, _} = msg_store_remove(MsgIds1stHalf, MSCState7),
ok = rabbit_msg_store:client_terminate(MSCState7),
%% restart empty
restart_msg_store_empty(), %% now safe to reuse msg_ids
%% push a lot of msgs in... at least 100 files worth
{ok, FileSize} = application:get_env(rabbit, msg_store_file_size_limit),
PayloadSizeBits = 65536,
BigCount = trunc(100 * FileSize / (PayloadSizeBits div 8)),
MsgIdsBig = [{GenRef(), msg_id_bin(X)} || X <- lists:seq(1, BigCount)],
Payload = << 0:PayloadSizeBits >>,
ok = with_msg_store_client(
?PERSISTENT_MSG_STORE, Ref,
fun (MSCStateM) ->
[ok = rabbit_msg_store:write(SeqId, MsgId, Payload, MSCStateM) ||
{SeqId, MsgId} <- MsgIdsBig],
MSCStateM
end),
%% now read them to ensure we hit the fast client-side reading
ok = foreach_with_msg_store_client(
?PERSISTENT_MSG_STORE, Ref,
fun ({_, MsgId}, MSCStateM) ->
{{ok, Payload}, MSCStateN} = rabbit_msg_store:read(
MsgId, MSCStateM),
MSCStateN
end, MsgIdsBig),
%% We remove every other other message first, then do it again a second
%% time with another set of messages and then a third time. We start
%% with younger messages on purpose. So we split the list in three
%% lists keeping the message reference.
Part = fun
PartFun([], _, Acc) ->
Acc;
PartFun([E|Tail], N, Acc) ->
Pos = 1 + (N rem 3),
AccL = element(Pos, Acc),
PartFun(Tail, N + 1, setelement(Pos, Acc, [E|AccL]))
end,
{One, Two, Three} = Part(MsgIdsBig, 0, {[], [], []}),
ok = msg_store_remove(?PERSISTENT_MSG_STORE, Ref, One),
%% This is likely to hit GC (under 50% good data left in files, but no empty files).
ok = msg_store_remove(?PERSISTENT_MSG_STORE, Ref, Two),
%% Files are empty now and will get removed.
ok = msg_store_remove(?PERSISTENT_MSG_STORE, Ref, Three),
%% ensure empty
ok = with_msg_store_client(
?PERSISTENT_MSG_STORE, Ref,
fun (MSCStateM) ->
false = msg_store_contains(false, MsgIdsBig, MSCStateM),
MSCStateM
end),
%%
passed = test_msg_store_client_delete_and_terminate(fun() -> GenRefFun(msc_cdat) end),
%% restart empty
restart_msg_store_empty(),
passed.
restart_msg_store_empty() ->
ok = rabbit_variable_queue:stop_msg_store(?VHOST),
ok = rabbit_variable_queue:start_msg_store(?VHOST,
undefined, {fun (ok) -> finished end, ok}).
msg_id_bin(X) ->
erlang:md5(term_to_binary(X)).
on_disk_capture() ->
receive
{await, MsgIds, Pid} -> on_disk_capture([], MsgIds, Pid);
stop -> done
end.
on_disk_capture([_|_], _Awaiting, Pid) ->
Pid ! {self(), surplus};
on_disk_capture(OnDisk, Awaiting, Pid) ->
receive
{on_disk, MsgIdsS} ->
MsgIds = sets:to_list(MsgIdsS),
on_disk_capture(OnDisk ++ (MsgIds -- Awaiting), Awaiting -- MsgIds,
Pid);
stop ->
done
after (case Awaiting of [] -> 200; _ -> ?TIMEOUT end) ->
case Awaiting of
[] -> Pid ! {self(), arrived}, on_disk_capture();
_ -> Pid ! {self(), timeout}
end
end.
on_disk_await(Pid, MsgIds0) when is_list(MsgIds0) ->
{_, MsgIds} = lists:unzip(MsgIds0),
Pid ! {await, MsgIds, self()},
receive
{Pid, arrived} -> ok;
{Pid, Error} -> Error
end.
on_disk_stop(Pid) ->
MRef = erlang:monitor(process, Pid),
Pid ! stop,
receive {'DOWN', MRef, process, Pid, _Reason} ->
ok
end.
msg_store_client_init_capture(MsgStore, Ref) ->
Pid = spawn(fun on_disk_capture/0),
{Pid, rabbit_vhost_msg_store:client_init(?VHOST, MsgStore, Ref,
fun (MsgIds, _ActionTaken) ->
Pid ! {on_disk, MsgIds}
end)}.
msg_store_contains(Atom, MsgIds, MSCState) ->
Atom = lists:foldl(
fun ({_, MsgId}, Atom1) when Atom1 =:= Atom ->
rabbit_msg_store:contains(MsgId, MSCState) end,
Atom, MsgIds).
msg_store_read(MsgIds, MSCState) ->
lists:foldl(fun ({_, MsgId}, MSCStateM) ->
{{ok, MsgId}, MSCStateN} = rabbit_msg_store:read(
MsgId, MSCStateM),
MSCStateN
end, MSCState, MsgIds).
msg_store_write(MsgIds, MSCState) ->
ok = lists:foldl(fun ({SeqId, MsgId}, ok) ->
rabbit_msg_store:write(SeqId, MsgId, MsgId, MSCState)
end, ok, MsgIds).
msg_store_write_flow(MsgIds, MSCState) ->
ok = lists:foldl(fun ({SeqId, MsgId}, ok) ->
rabbit_msg_store:write_flow(SeqId, MsgId, MsgId, MSCState)
end, ok, MsgIds).
msg_store_remove(MsgIds, MSCState) ->
rabbit_msg_store:remove(MsgIds, MSCState).
msg_store_remove(MsgStore, Ref, MsgIds) ->
with_msg_store_client(MsgStore, Ref,
fun (MSCStateM) ->
CQ shared store write optimisations (#8507) * CQ: Don't use FHC for writes in shared store * CQ: Send confirms when flushing to disk in shared store Before they would only be sent periodically or when rolling over to a new file. * CQ: Fast-confirm when flushing data to disk We know the messages are on disk or were acked so there is no need to do sets intersections/subtracts in this scenario. * Fix a Dialyzer warning * Faster confirms for unwritten messages Instead of having the message store send a message to the queue with the confirms for messages ignored due to the flying optimisation, we have the queue handle the confirms directly when removing the messages. This avoids sending potentially 1 Erlang message per 1 AMQP message to the queue. * Refactor rabbit_msg_file:pread into rabbit_msg_store Also make use of the opened file for multi-reads instead of opening/reading/closing each time. * CQ: Make sure we keep the updated CState when using read_many * CQ shared store: Run compaction on older file candidates The way I initially did this the maybe_gc would be triggered based on candidates from 15s ago, but run against candidates from just now. This is sub-optimal because when messages are consumed rapidly, just-now candidates are likely to be in a file about to be deleted, and we don't want to run compaction on those. Instead, when sending the maybe_gc we also send the candidates we had at the time. Then 15s later we check if the file still exists. If it's gone, great! No compaction to do. * CQ: Add a few todos for later
2023-06-21 02:04:17 +08:00
{ok, _} = msg_store_remove(MsgIds, MSCStateM),
MSCStateM
end).
with_msg_store_client(MsgStore, Ref, Fun) ->
rabbit_msg_store:client_terminate(
Fun(msg_store_client_init(MsgStore, Ref))).
foreach_with_msg_store_client(MsgStore, Ref, Fun, L) ->
rabbit_msg_store:client_terminate(
lists:foldl(fun (MsgId, MSCState) -> Fun(MsgId, MSCState) end,
msg_store_client_init(MsgStore, Ref), L)).
test_msg_store_confirms(MsgIds, Cap, GenRef, MSCState) ->
%% write -> confirmed
MsgIds1 = [{GenRef(), MsgId} || {_, MsgId} <- MsgIds],
ok = msg_store_write(MsgIds1, MSCState),
ok = on_disk_await(Cap, MsgIds1),
%% remove -> _
CQ shared store write optimisations (#8507) * CQ: Don't use FHC for writes in shared store * CQ: Send confirms when flushing to disk in shared store Before they would only be sent periodically or when rolling over to a new file. * CQ: Fast-confirm when flushing data to disk We know the messages are on disk or were acked so there is no need to do sets intersections/subtracts in this scenario. * Fix a Dialyzer warning * Faster confirms for unwritten messages Instead of having the message store send a message to the queue with the confirms for messages ignored due to the flying optimisation, we have the queue handle the confirms directly when removing the messages. This avoids sending potentially 1 Erlang message per 1 AMQP message to the queue. * Refactor rabbit_msg_file:pread into rabbit_msg_store Also make use of the opened file for multi-reads instead of opening/reading/closing each time. * CQ: Make sure we keep the updated CState when using read_many * CQ shared store: Run compaction on older file candidates The way I initially did this the maybe_gc would be triggered based on candidates from 15s ago, but run against candidates from just now. This is sub-optimal because when messages are consumed rapidly, just-now candidates are likely to be in a file about to be deleted, and we don't want to run compaction on those. Instead, when sending the maybe_gc we also send the candidates we had at the time. Then 15s later we check if the file still exists. If it's gone, great! No compaction to do. * CQ: Add a few todos for later
2023-06-21 02:04:17 +08:00
{ok, []} = msg_store_remove(MsgIds1, MSCState),
ok = on_disk_await(Cap, []),
%% write, remove -> confirmed
MsgIds2 = [{GenRef(), MsgId} || {_, MsgId} <- MsgIds],
ok = msg_store_write(MsgIds2, MSCState),
CQ shared store write optimisations (#8507) * CQ: Don't use FHC for writes in shared store * CQ: Send confirms when flushing to disk in shared store Before they would only be sent periodically or when rolling over to a new file. * CQ: Fast-confirm when flushing data to disk We know the messages are on disk or were acked so there is no need to do sets intersections/subtracts in this scenario. * Fix a Dialyzer warning * Faster confirms for unwritten messages Instead of having the message store send a message to the queue with the confirms for messages ignored due to the flying optimisation, we have the queue handle the confirms directly when removing the messages. This avoids sending potentially 1 Erlang message per 1 AMQP message to the queue. * Refactor rabbit_msg_file:pread into rabbit_msg_store Also make use of the opened file for multi-reads instead of opening/reading/closing each time. * CQ: Make sure we keep the updated CState when using read_many * CQ shared store: Run compaction on older file candidates The way I initially did this the maybe_gc would be triggered based on candidates from 15s ago, but run against candidates from just now. This is sub-optimal because when messages are consumed rapidly, just-now candidates are likely to be in a file about to be deleted, and we don't want to run compaction on those. Instead, when sending the maybe_gc we also send the candidates we had at the time. Then 15s later we check if the file still exists. If it's gone, great! No compaction to do. * CQ: Add a few todos for later
2023-06-21 02:04:17 +08:00
{ok, ConfirmedMsgIds2} = msg_store_remove(MsgIds2, MSCState),
ok = on_disk_await(Cap, lists:filter(fun({_, MsgId}) -> not lists:member(MsgId, ConfirmedMsgIds2) end, MsgIds2)),
%% write, remove, write -> confirmed, confirmed
MsgIds3 = [{GenRef(), MsgId} || {_, MsgId} <- MsgIds],
ok = msg_store_write(MsgIds3, MSCState),
CQ shared store write optimisations (#8507) * CQ: Don't use FHC for writes in shared store * CQ: Send confirms when flushing to disk in shared store Before they would only be sent periodically or when rolling over to a new file. * CQ: Fast-confirm when flushing data to disk We know the messages are on disk or were acked so there is no need to do sets intersections/subtracts in this scenario. * Fix a Dialyzer warning * Faster confirms for unwritten messages Instead of having the message store send a message to the queue with the confirms for messages ignored due to the flying optimisation, we have the queue handle the confirms directly when removing the messages. This avoids sending potentially 1 Erlang message per 1 AMQP message to the queue. * Refactor rabbit_msg_file:pread into rabbit_msg_store Also make use of the opened file for multi-reads instead of opening/reading/closing each time. * CQ: Make sure we keep the updated CState when using read_many * CQ shared store: Run compaction on older file candidates The way I initially did this the maybe_gc would be triggered based on candidates from 15s ago, but run against candidates from just now. This is sub-optimal because when messages are consumed rapidly, just-now candidates are likely to be in a file about to be deleted, and we don't want to run compaction on those. Instead, when sending the maybe_gc we also send the candidates we had at the time. Then 15s later we check if the file still exists. If it's gone, great! No compaction to do. * CQ: Add a few todos for later
2023-06-21 02:04:17 +08:00
{ok, ConfirmedMsgIds3} = msg_store_remove(MsgIds3, MSCState),
MsgIds4 = [{GenRef(), MsgId} || {_, MsgId} <- MsgIds],
ok = msg_store_write(MsgIds4, MSCState),
CQ shared store write optimisations (#8507) * CQ: Don't use FHC for writes in shared store * CQ: Send confirms when flushing to disk in shared store Before they would only be sent periodically or when rolling over to a new file. * CQ: Fast-confirm when flushing data to disk We know the messages are on disk or were acked so there is no need to do sets intersections/subtracts in this scenario. * Fix a Dialyzer warning * Faster confirms for unwritten messages Instead of having the message store send a message to the queue with the confirms for messages ignored due to the flying optimisation, we have the queue handle the confirms directly when removing the messages. This avoids sending potentially 1 Erlang message per 1 AMQP message to the queue. * Refactor rabbit_msg_file:pread into rabbit_msg_store Also make use of the opened file for multi-reads instead of opening/reading/closing each time. * CQ: Make sure we keep the updated CState when using read_many * CQ shared store: Run compaction on older file candidates The way I initially did this the maybe_gc would be triggered based on candidates from 15s ago, but run against candidates from just now. This is sub-optimal because when messages are consumed rapidly, just-now candidates are likely to be in a file about to be deleted, and we don't want to run compaction on those. Instead, when sending the maybe_gc we also send the candidates we had at the time. Then 15s later we check if the file still exists. If it's gone, great! No compaction to do. * CQ: Add a few todos for later
2023-06-21 02:04:17 +08:00
ok = on_disk_await(Cap, lists:filter(fun({_, MsgId}) -> not lists:member(MsgId, ConfirmedMsgIds3) end, MsgIds3) ++ MsgIds4),
%% remove, write -> confirmed
CQ shared store write optimisations (#8507) * CQ: Don't use FHC for writes in shared store * CQ: Send confirms when flushing to disk in shared store Before they would only be sent periodically or when rolling over to a new file. * CQ: Fast-confirm when flushing data to disk We know the messages are on disk or were acked so there is no need to do sets intersections/subtracts in this scenario. * Fix a Dialyzer warning * Faster confirms for unwritten messages Instead of having the message store send a message to the queue with the confirms for messages ignored due to the flying optimisation, we have the queue handle the confirms directly when removing the messages. This avoids sending potentially 1 Erlang message per 1 AMQP message to the queue. * Refactor rabbit_msg_file:pread into rabbit_msg_store Also make use of the opened file for multi-reads instead of opening/reading/closing each time. * CQ: Make sure we keep the updated CState when using read_many * CQ shared store: Run compaction on older file candidates The way I initially did this the maybe_gc would be triggered based on candidates from 15s ago, but run against candidates from just now. This is sub-optimal because when messages are consumed rapidly, just-now candidates are likely to be in a file about to be deleted, and we don't want to run compaction on those. Instead, when sending the maybe_gc we also send the candidates we had at the time. Then 15s later we check if the file still exists. If it's gone, great! No compaction to do. * CQ: Add a few todos for later
2023-06-21 02:04:17 +08:00
{ok, []} = msg_store_remove(MsgIds4, MSCState),
MsgIds5 = [{GenRef(), MsgId} || {_, MsgId} <- MsgIds],
ok = msg_store_write(MsgIds5, MSCState),
ok = on_disk_await(Cap, MsgIds5),
%% remove, write, remove -> confirmed
CQ shared store write optimisations (#8507) * CQ: Don't use FHC for writes in shared store * CQ: Send confirms when flushing to disk in shared store Before they would only be sent periodically or when rolling over to a new file. * CQ: Fast-confirm when flushing data to disk We know the messages are on disk or were acked so there is no need to do sets intersections/subtracts in this scenario. * Fix a Dialyzer warning * Faster confirms for unwritten messages Instead of having the message store send a message to the queue with the confirms for messages ignored due to the flying optimisation, we have the queue handle the confirms directly when removing the messages. This avoids sending potentially 1 Erlang message per 1 AMQP message to the queue. * Refactor rabbit_msg_file:pread into rabbit_msg_store Also make use of the opened file for multi-reads instead of opening/reading/closing each time. * CQ: Make sure we keep the updated CState when using read_many * CQ shared store: Run compaction on older file candidates The way I initially did this the maybe_gc would be triggered based on candidates from 15s ago, but run against candidates from just now. This is sub-optimal because when messages are consumed rapidly, just-now candidates are likely to be in a file about to be deleted, and we don't want to run compaction on those. Instead, when sending the maybe_gc we also send the candidates we had at the time. Then 15s later we check if the file still exists. If it's gone, great! No compaction to do. * CQ: Add a few todos for later
2023-06-21 02:04:17 +08:00
{ok, []} = msg_store_remove(MsgIds5, MSCState),
MsgIds6 = [{GenRef(), MsgId} || {_, MsgId} <- MsgIds],
ok = msg_store_write(MsgIds6, MSCState),
CQ shared store write optimisations (#8507) * CQ: Don't use FHC for writes in shared store * CQ: Send confirms when flushing to disk in shared store Before they would only be sent periodically or when rolling over to a new file. * CQ: Fast-confirm when flushing data to disk We know the messages are on disk or were acked so there is no need to do sets intersections/subtracts in this scenario. * Fix a Dialyzer warning * Faster confirms for unwritten messages Instead of having the message store send a message to the queue with the confirms for messages ignored due to the flying optimisation, we have the queue handle the confirms directly when removing the messages. This avoids sending potentially 1 Erlang message per 1 AMQP message to the queue. * Refactor rabbit_msg_file:pread into rabbit_msg_store Also make use of the opened file for multi-reads instead of opening/reading/closing each time. * CQ: Make sure we keep the updated CState when using read_many * CQ shared store: Run compaction on older file candidates The way I initially did this the maybe_gc would be triggered based on candidates from 15s ago, but run against candidates from just now. This is sub-optimal because when messages are consumed rapidly, just-now candidates are likely to be in a file about to be deleted, and we don't want to run compaction on those. Instead, when sending the maybe_gc we also send the candidates we had at the time. Then 15s later we check if the file still exists. If it's gone, great! No compaction to do. * CQ: Add a few todos for later
2023-06-21 02:04:17 +08:00
{ok, ConfirmedMsgIds6} = msg_store_remove(MsgIds6, MSCState),
ok = on_disk_await(Cap, lists:filter(fun({_, MsgId}) -> not lists:member(MsgId, ConfirmedMsgIds6) end, MsgIds6)),
%% confirmation on timer-based sync
passed = test_msg_store_confirm_timer(GenRef),
passed.
test_msg_store_confirm_timer(GenRef) ->
Ref = rabbit_guid:gen(),
MsgId = msg_id_bin(1),
Self = self(),
MSCState = rabbit_vhost_msg_store:client_init(
?VHOST,
?PERSISTENT_MSG_STORE,
Ref,
fun (MsgIds, _ActionTaken) ->
case sets:is_element(MsgId, MsgIds) of
true -> Self ! on_disk;
false -> ok
end
end),
MsgIdsChecked = [{GenRef(), MsgId}],
ok = msg_store_write(MsgIdsChecked, MSCState),
ok = msg_store_keep_busy_until_confirm([msg_id_bin(2)], GenRef, MSCState, false),
CQ shared store write optimisations (#8507) * CQ: Don't use FHC for writes in shared store * CQ: Send confirms when flushing to disk in shared store Before they would only be sent periodically or when rolling over to a new file. * CQ: Fast-confirm when flushing data to disk We know the messages are on disk or were acked so there is no need to do sets intersections/subtracts in this scenario. * Fix a Dialyzer warning * Faster confirms for unwritten messages Instead of having the message store send a message to the queue with the confirms for messages ignored due to the flying optimisation, we have the queue handle the confirms directly when removing the messages. This avoids sending potentially 1 Erlang message per 1 AMQP message to the queue. * Refactor rabbit_msg_file:pread into rabbit_msg_store Also make use of the opened file for multi-reads instead of opening/reading/closing each time. * CQ: Make sure we keep the updated CState when using read_many * CQ shared store: Run compaction on older file candidates The way I initially did this the maybe_gc would be triggered based on candidates from 15s ago, but run against candidates from just now. This is sub-optimal because when messages are consumed rapidly, just-now candidates are likely to be in a file about to be deleted, and we don't want to run compaction on those. Instead, when sending the maybe_gc we also send the candidates we had at the time. Then 15s later we check if the file still exists. If it's gone, great! No compaction to do. * CQ: Add a few todos for later
2023-06-21 02:04:17 +08:00
{ok, _} = msg_store_remove(MsgIdsChecked, MSCState),
ok = rabbit_msg_store:client_delete_and_terminate(MSCState),
passed.
msg_store_keep_busy_until_confirm(MsgIds, GenRef, MSCState, Blocked) ->
After = case Blocked of
false -> 0;
true -> ?MAX_WAIT
end,
Recurse = fun () -> msg_store_keep_busy_until_confirm(
MsgIds, GenRef, MSCState, credit_flow:blocked()) end,
receive
on_disk -> ok;
{bump_credit, Msg} -> credit_flow:handle_bump_msg(Msg),
Recurse()
after After ->
MsgIds1 = [{GenRef(), MsgId} || MsgId <- MsgIds],
ok = msg_store_write_flow(MsgIds1, MSCState),
CQ shared store write optimisations (#8507) * CQ: Don't use FHC for writes in shared store * CQ: Send confirms when flushing to disk in shared store Before they would only be sent periodically or when rolling over to a new file. * CQ: Fast-confirm when flushing data to disk We know the messages are on disk or were acked so there is no need to do sets intersections/subtracts in this scenario. * Fix a Dialyzer warning * Faster confirms for unwritten messages Instead of having the message store send a message to the queue with the confirms for messages ignored due to the flying optimisation, we have the queue handle the confirms directly when removing the messages. This avoids sending potentially 1 Erlang message per 1 AMQP message to the queue. * Refactor rabbit_msg_file:pread into rabbit_msg_store Also make use of the opened file for multi-reads instead of opening/reading/closing each time. * CQ: Make sure we keep the updated CState when using read_many * CQ shared store: Run compaction on older file candidates The way I initially did this the maybe_gc would be triggered based on candidates from 15s ago, but run against candidates from just now. This is sub-optimal because when messages are consumed rapidly, just-now candidates are likely to be in a file about to be deleted, and we don't want to run compaction on those. Instead, when sending the maybe_gc we also send the candidates we had at the time. Then 15s later we check if the file still exists. If it's gone, great! No compaction to do. * CQ: Add a few todos for later
2023-06-21 02:04:17 +08:00
{ok, _} = msg_store_remove(MsgIds1, MSCState),
Recurse()
end.
test_msg_store_client_delete_and_terminate(GenRef) ->
restart_msg_store_empty(),
MsgIds = [{GenRef(), msg_id_bin(M)} || M <- lists:seq(1, 10)],
Ref = rabbit_guid:gen(),
MSCState = msg_store_client_init(?PERSISTENT_MSG_STORE, Ref),
ok = msg_store_write(MsgIds, MSCState),
%% test the 'dying client' fast path for writes
ok = rabbit_msg_store:client_delete_and_terminate(MSCState),
passed.
%% -------------------------------------------------------------------
%% Message store file scanning.
%% -------------------------------------------------------------------
%% While it is possible although very unlikely that this test case
%% produces false positives, all failures of this test case should
%% be investigated thoroughly as they test an algorithm that is
%% central to the reliability of the data in the shared message store.
%% Failing files can be found in the CT private data.
msg_store_file_scan(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, msg_store_file_scan1, [Config]).
msg_store_file_scan1(Config) ->
Scan = fun (Blocks) ->
Expected = gen_result(Blocks),
Path = gen_msg_file(Config, Blocks),
Result = rabbit_msg_store:scan_file_for_valid_messages(Path),
ok = file:delete(Path),
case Result of
Expected -> ok;
_ -> {expected, Expected, got, Result}
end
end,
%% Empty files.
ok = Scan([]),
ok = Scan([{pad, 1024}]),
ok = Scan([{pad, 1024 * 1024}]),
%% One-message files.
ok = Scan([{msg, gen_id(), <<0>>}]),
ok = Scan([{msg, gen_id(), <<255>>}]),
ok = Scan([{msg, gen_id(), gen_msg()}]),
ok = Scan([{pad, 1024}, {msg, gen_id(), gen_msg()}]),
ok = Scan([{pad, 1024 * 1024}, {msg, gen_id(), gen_msg()}]),
ok = Scan([{msg, gen_id(), gen_msg()}, {pad, 1024}]),
ok = Scan([{msg, gen_id(), gen_msg()}, {pad, 1024 * 1024}]),
%% Multiple messages.
ok = Scan([{msg, gen_id(), gen_msg()} || _ <- lists:seq(1, 2)]),
ok = Scan([{msg, gen_id(), gen_msg()} || _ <- lists:seq(1, 5)]),
ok = Scan([{msg, gen_id(), gen_msg()} || _ <- lists:seq(1, 20)]),
ok = Scan([{msg, gen_id(), gen_msg()} || _ <- lists:seq(1, 100)]),
%% Multiple messages with padding.
ok = Scan([
{pad, 1024},
{msg, gen_id(), gen_msg()},
{msg, gen_id(), gen_msg()}
]),
ok = Scan([
{msg, gen_id(), gen_msg()},
{pad, 1024},
{msg, gen_id(), gen_msg()}
]),
ok = Scan([
{msg, gen_id(), gen_msg()},
{msg, gen_id(), gen_msg()},
{pad, 1024}
]),
ok = Scan([
{pad, 1024},
{msg, gen_id(), gen_msg()},
{pad, 1024},
{msg, gen_id(), gen_msg()}
]),
ok = Scan([
{msg, gen_id(), gen_msg()},
{pad, 1024},
{msg, gen_id(), gen_msg()},
{pad, 1024}
]),
ok = Scan([
{pad, 1024},
{msg, gen_id(), gen_msg()},
{msg, gen_id(), gen_msg()},
{pad, 1024}
]),
ok = Scan([
{pad, 1024},
{msg, gen_id(), gen_msg()},
{pad, 1024},
{msg, gen_id(), gen_msg()},
{pad, 1024}
]),
OneOf = fun(A, B) ->
case rand:uniform() of
F when F < +0.5 -> A;
_ -> B
end
end,
ok = Scan([OneOf({msg, gen_id(), gen_msg()}, {pad, 1024}) || _ <- lists:seq(1, 2)]),
ok = Scan([OneOf({msg, gen_id(), gen_msg()}, {pad, 1024}) || _ <- lists:seq(1, 5)]),
ok = Scan([OneOf({msg, gen_id(), gen_msg()}, {pad, 1024}) || _ <- lists:seq(1, 20)]),
ok = Scan([OneOf({msg, gen_id(), gen_msg()}, {pad, 1024}) || _ <- lists:seq(1, 100)]),
%% Duplicate messages.
Msg = {msg, gen_id(), gen_msg()},
ok = Scan([Msg, Msg]),
ok = Scan([Msg, Msg, Msg, Msg, Msg]),
ok = Scan([Msg, {pad, 1024}, Msg]),
ok = Scan([Msg]
++ [OneOf({msg, gen_id(), gen_msg()}, {pad, 1024}) || _ <- lists:seq(1, 100)]
++ [Msg]),
%% Truncated start of message.
ok = Scan([{bin, <<21:56, "deadbeefdeadbeef", "hello", 255>>}]),
ok = Scan([{bin, <<21:48, "deadbeefdeadbeef", "hello", 255>>}]),
ok = Scan([{bin, <<21:40, "deadbeefdeadbeef", "hello", 255>>}]),
ok = Scan([{bin, <<21:32, "deadbeefdeadbeef", "hello", 255>>}]),
ok = Scan([{bin, <<21:24, "deadbeefdeadbeef", "hello", 255>>}]),
ok = Scan([{bin, <<21:16, "deadbeefdeadbeef", "hello", 255>>}]),
ok = Scan([{bin, <<21:8, "deadbeefdeadbeef", "hello", 255>>}]),
ok = Scan([{bin, <<"deadbeefdeadbeef", "hello", 255>>}]),
ok = Scan([{bin, <<"beefdeadbeef", "hello", 255>>}]),
ok = Scan([{bin, <<"deadbeef", "hello", 255>>}]),
ok = Scan([{bin, <<"beef", "hello", 255>>}]),
ok = Scan([{bin, <<"hello", 255>>}]),
ok = Scan([{bin, <<255>>}]),
%% Truncated end of message (unlikely).
ok = Scan([{bin, <<255>>}]),
ok = Scan([{bin, <<255, 255>>}]),
ok = Scan([{bin, <<255, 255, 255>>}]),
ok = Scan([{bin, <<255, 255, 255, 255>>}]),
ok = Scan([{bin, <<255, 255, 255, 255, 255>>}]),
ok = Scan([{bin, <<255, 255, 255, 255, 255, 255>>}]),
ok = Scan([{bin, <<255, 255, 255, 255, 255, 255, 255>>}]),
ok = Scan([{bin, <<255, 255, 255, 255, 255, 255, 255, 255>>}]),
ok = Scan([{bin, <<15:64, "deadbeefdeadbee">>}]),
ok = Scan([{bin, <<16:64, "deadbeefdeadbeef">>}]),
ok = Scan([{bin, <<17:64, "deadbeefdeadbeef", 0>>}]),
ok = Scan([{bin, <<17:64, "deadbeefdeadbeef", 255>>}]),
ok = Scan([{bin, <<17:64, "deadbeefdeadbeef", 255, 254>>}]),
%% Messages with no content.
ok = Scan([{bin, <<0:64, "deadbeefdeadbeef", 255>>}]),
ok = Scan([{msg, gen_id(), <<>>}]),
CQ: Fix shared store scanner missing messages It was still possible, although rare, to have message store files lose message data, when the following conditions were met: * the message data contains byte values 255 (255 is used as an OK marker after a message) * the message is located after a 0-filled hole in the file * the length of the data is at least 4096 bytes and if we misread it (as detailed below) we encounter a 255 byte where we expect the OK marker The trick for the code to previously misread the length can be explained as follow: A message is stored in the following format: <<Len:64, MsgIdAndMsg:Len/unit:8, 255>> With MsgId always being 16 bytes in length. So Len is always at least 16, if the message data Msg is empty. But technically it never is. Now if we have a zero filled hole just before this message, we may end up with this: <<0, Len:64, MsgIdAndMsg:Len/unit:8, 255>> When we are scanning we are testing bytes to see if there is a message there or not. We look for a Len that gives us byte 255 after MsgIdAndMsg. Len of value 4096 looks like this in binary: <<0:48, 16, 0>> Problem is if we have leading zeroes, Len may look like this: <<0, 0:48, 16, 0>> If we take the first 64 bits we get a potential length of 16. We look at the byte after the next 16 bytes. If it is 255, we think this is a message and skip by this amount of bytes, and mistakenly miss the real message. Solving this by changing the file format would be simple enough, but we don't have the luxury to afford that. A different solution was found, which is to combine file scanning with checking that the message exists in the message store index (populated from queues at startup, and kept up to date over the life time of the store). Then we know for sure that the message above doesn't exist, because the MsgId won't be found in the index. If it is, then the file number and offset will not match, and the check will fail. There remains a small chance that we get it wrong during dirty recovery. Only a better file format would improve that.
2024-09-26 22:43:55 +08:00
%% Tricky messages.
%%
%% These only get properly detected when the index is populated.
%% In this test case we simulate the index with a fun.
TrickyScan = fun (Blocks, Expected, Fun) ->
Path = gen_msg_file(Config, Blocks),
Result = rabbit_msg_store:scan_file_for_valid_messages(Path, Fun),
case Result of
Expected -> ok;
_ -> {expected, Expected, got, Result}
end
end,
ok = TrickyScan(
[{bin, <<0, 0:48, 17, 17, "idididididididid", 255, 0:4352/unit:8, 255>>}],
{ok, [{<<"idididididididid">>, 4378, 1}]},
fun(Obj = {<<"idididididididid">>, 4378, 1}) -> {valid, Obj}; (_) -> invalid end),
%% All good!!
passed.
gen_id() ->
rand:bytes(16).
gen_msg() ->
gen_msg(1024 * 1024).
gen_msg(MaxSize) ->
Bytes = rand:bytes(rand:uniform(MaxSize)),
%% We remove 255 to avoid false positives. In a running
%% rabbit node we will not get false positives because
%% we also check messages against the index.
<< <<case B of 255 -> 254; _ -> B end>> || <<B>> <= Bytes >>.
gen_msg_file(Config, Blocks) ->
PrivDir = ?config(priv_dir, Config),
TmpFile = integer_to_list(erlang:unique_integer([positive])),
Path = filename:join(PrivDir, TmpFile),
ok = file:write_file(Path, [case Block of
{bin, Bin} ->
Bin;
{pad, Size} ->
%% Empty space between messages is expected to be zeroes.
<<0:Size/unit:8>>;
{msg, MsgId, Msg} ->
Size = 16 + byte_size(Msg),
[<<Size:64>>, MsgId, Msg, <<255>>]
end || Block <- Blocks]),
Path.
gen_result(Blocks) ->
Messages = gen_result(Blocks, 0, []),
CQ: Fix shared store scanner missing messages It was still possible, although rare, to have message store files lose message data, when the following conditions were met: * the message data contains byte values 255 (255 is used as an OK marker after a message) * the message is located after a 0-filled hole in the file * the length of the data is at least 4096 bytes and if we misread it (as detailed below) we encounter a 255 byte where we expect the OK marker The trick for the code to previously misread the length can be explained as follow: A message is stored in the following format: <<Len:64, MsgIdAndMsg:Len/unit:8, 255>> With MsgId always being 16 bytes in length. So Len is always at least 16, if the message data Msg is empty. But technically it never is. Now if we have a zero filled hole just before this message, we may end up with this: <<0, Len:64, MsgIdAndMsg:Len/unit:8, 255>> When we are scanning we are testing bytes to see if there is a message there or not. We look for a Len that gives us byte 255 after MsgIdAndMsg. Len of value 4096 looks like this in binary: <<0:48, 16, 0>> Problem is if we have leading zeroes, Len may look like this: <<0, 0:48, 16, 0>> If we take the first 64 bits we get a potential length of 16. We look at the byte after the next 16 bytes. If it is 255, we think this is a message and skip by this amount of bytes, and mistakenly miss the real message. Solving this by changing the file format would be simple enough, but we don't have the luxury to afford that. A different solution was found, which is to combine file scanning with checking that the message exists in the message store index (populated from queues at startup, and kept up to date over the life time of the store). Then we know for sure that the message above doesn't exist, because the MsgId won't be found in the index. If it is, then the file number and offset will not match, and the check will fail. There remains a small chance that we get it wrong during dirty recovery. Only a better file format would improve that.
2024-09-26 22:43:55 +08:00
{ok, Messages}.
gen_result([], _, Acc) ->
Acc;
gen_result([{bin, Bin}|Tail], Offset, Acc) ->
gen_result(Tail, Offset + byte_size(Bin), Acc);
gen_result([{pad, Size}|Tail], Offset, Acc) ->
gen_result(Tail, Offset + Size, Acc);
gen_result([{msg, MsgId, Msg}|Tail], Offset, Acc) ->
Size = 9 + 16 + byte_size(Msg),
%% Only the first MsgId found is returned when duplicates exist.
case lists:keymember(MsgId, 1, Acc) of
false ->
gen_result(Tail, Offset + Size, [{MsgId, Size, Offset}|Acc]);
true ->
gen_result(Tail, Offset + Size, Acc)
end.
%% -------------------------------------------------------------------
%% Backing queue.
%% -------------------------------------------------------------------
setup_backing_queue_test_group(Config) ->
2021-10-11 18:30:39 +08:00
{ok, MaxJournal} =
application:get_env(rabbit, queue_index_max_journal_entries),
application:set_env(rabbit, queue_index_max_journal_entries, 128),
{ok, Bytes} =
application:get_env(rabbit, queue_index_embed_msgs_below),
rabbit_ct_helpers:set_config(Config, [
2021-10-11 18:30:39 +08:00
{rmq_queue_index_max_journal_entries, MaxJournal},
{rmq_queue_index_embed_msgs_below, Bytes}
]).
teardown_backing_queue_test_group(Config) ->
%% FIXME: Undo all the setup function did.
application:set_env(rabbit, queue_index_max_journal_entries,
?config(rmq_queue_index_max_journal_entries, Config)),
%% We will have restarted the message store, and thus changed
%% the order of the children of rabbit_sup. This will cause
%% problems if there are subsequent failures - see bug 24262.
ok = restart_app(),
Config.
bq_queue_index(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, bq_queue_index1, [Config]).
2021-10-11 17:49:11 +08:00
index_mod() ->
rabbit_classic_queue_index_v2.
2021-10-11 17:49:11 +08:00
bq_queue_index1(_Config) ->
init_queue_index(),
2021-10-11 17:49:11 +08:00
IndexMod = index_mod(),
SegmentSize = IndexMod:next_segment_boundary(0),
TwoSegs = SegmentSize + SegmentSize,
MostOfASegment = trunc(SegmentSize*0.75),
SeqIdsA = lists:seq(0, MostOfASegment-1),
SeqIdsB = lists:seq(MostOfASegment, 2*MostOfASegment),
SeqIdsC = lists:seq(0, trunc(SegmentSize/2)),
SeqIdsD = lists:seq(0, SegmentSize*4),
VerifyReadWithPublishedFun = fun verify_read_with_published_v2/3,
with_empty_test_queue(
fun (Qi0, QName) ->
2021-10-11 17:49:11 +08:00
{0, 0, Qi1} = IndexMod:bounds(Qi0),
{Qi2, SeqIdsMsgIdsA} = queue_index_publish(SeqIdsA, false, Qi1),
2021-10-11 17:49:11 +08:00
{0, SegmentSize, Qi3} = IndexMod:bounds(Qi2),
{ReadA, Qi4} = IndexMod:read(0, SegmentSize, Qi3),
ok = VerifyReadWithPublishedFun(false, ReadA,
lists:reverse(SeqIdsMsgIdsA)),
%% should get length back as 0, as all the msgs were transient
{0, 0, Qi6} = restart_test_queue(Qi4, QName),
2021-10-11 17:49:11 +08:00
{0, 0, Qi7} = IndexMod:bounds(Qi6),
{Qi8, SeqIdsMsgIdsB} = queue_index_publish(SeqIdsB, true, Qi7),
2021-10-11 17:49:11 +08:00
{0, TwoSegs, Qi9} = IndexMod:bounds(Qi8),
{ReadB, Qi10} = IndexMod:read(0, SegmentSize, Qi9),
ok = VerifyReadWithPublishedFun(true, ReadB,
lists:reverse(SeqIdsMsgIdsB)),
%% should get length back as MostOfASegment
LenB = length(SeqIdsB),
BytesB = LenB * 10,
{LenB, BytesB, Qi12} = restart_test_queue(Qi10, QName),
2021-10-11 17:49:11 +08:00
{0, TwoSegs, Qi13} = IndexMod:bounds(Qi12),
Qi15 = case IndexMod of
rabbit_queue_index ->
Qi14 = IndexMod:deliver(SeqIdsB, Qi13),
{ReadC, Qi14b} = IndexMod:read(0, SegmentSize, Qi14),
ok = VerifyReadWithPublishedFun(true, ReadC,
2021-10-11 17:49:11 +08:00
lists:reverse(SeqIdsMsgIdsB)),
Qi14b;
_ ->
Qi13
end,
{_DeletedSegments, Qi16} = IndexMod:ack(SeqIdsB, Qi15),
Qi17 = IndexMod:flush(Qi16),
%% Everything will have gone now because #pubs == #acks
2021-10-11 17:49:11 +08:00
{0, 0, Qi18} = IndexMod:bounds(Qi17),
%% should get length back as 0 because all persistent
%% msgs have been acked
{0, 0, Qi19} = restart_test_queue(Qi18, QName),
Qi19
end),
%% These next bits are just to hit the auto deletion of segment files.
%% First, partials:
2021-10-11 18:30:39 +08:00
%% a) partial pub+del+ack, then move to new segment
with_empty_test_queue(
fun (Qi0, _QName) ->
{Qi1, _SeqIdsMsgIdsC} = queue_index_publish(SeqIdsC,
false, Qi0),
2021-10-11 18:30:39 +08:00
Qi2 = case IndexMod of
rabbit_queue_index -> IndexMod:deliver(SeqIdsC, Qi1);
_ -> Qi1
end,
{_DeletedSegments, Qi3} = IndexMod:ack(SeqIdsC, Qi2),
2021-10-11 17:49:11 +08:00
Qi4 = IndexMod:flush(Qi3),
{Qi5, _SeqIdsMsgIdsC1} = queue_index_publish([SegmentSize],
false, Qi4),
Qi5
end),
2021-10-11 18:30:39 +08:00
%% b) partial pub+del, then move to new segment, then ack all in old segment
with_empty_test_queue(
fun (Qi0, _QName) ->
{Qi1, _SeqIdsMsgIdsC2} = queue_index_publish(SeqIdsC,
false, Qi0),
2021-10-11 18:30:39 +08:00
Qi2 = case IndexMod of
rabbit_queue_index -> IndexMod:deliver(SeqIdsC, Qi1);
_ -> Qi1
end,
{Qi3, _SeqIdsMsgIdsC3} = queue_index_publish([SegmentSize],
2021-10-11 18:30:39 +08:00
false, Qi2),
2021-10-11 17:49:11 +08:00
{_DeletedSegments, Qi4} = IndexMod:ack(SeqIdsC, Qi3),
IndexMod:flush(Qi4)
end),
2021-08-30 20:10:47 +08:00
%% c) just fill up several segments of all pubs, then +acks
with_empty_test_queue(
fun (Qi0, _QName) ->
{Qi1, _SeqIdsMsgIdsD} = queue_index_publish(SeqIdsD,
false, Qi0),
2021-10-11 18:30:39 +08:00
Qi2 = case IndexMod of
rabbit_queue_index -> IndexMod:deliver(SeqIdsD, Qi1);
_ -> Qi1
end,
{_DeletedSegments, Qi3} = IndexMod:ack(SeqIdsD, Qi2),
2021-10-11 17:49:11 +08:00
IndexMod:flush(Qi3)
end),
%% d) get messages in all states to a segment, then flush, then do
%% the same again, don't flush and read.
with_empty_test_queue(
fun (Qi0, _QName) ->
{Qi1, [Seven,Five,Four|_]} = queue_index_publish([0,1,2,4,5,7],
false, Qi0),
2021-10-11 17:49:11 +08:00
Qi2 = case IndexMod of
rabbit_queue_index -> IndexMod:deliver([0,1,4], Qi1);
_ -> Qi1
end,
{_DeletedSegments3, Qi3} = IndexMod:ack([0], Qi2),
Qi4 = IndexMod:flush(Qi3),
{Qi5, [Eight,Six|_]} = queue_index_publish([3,6,8], false, Qi4),
2021-10-11 17:49:11 +08:00
Qi6 = case IndexMod of
rabbit_queue_index -> IndexMod:deliver([2,3,5,6], Qi5);
_ -> Qi5
end,
{_DeletedSegments7, Qi7} = IndexMod:ack([1,2,3], Qi6),
{[], Qi8} = IndexMod:read(0, 4, Qi7),
{ReadD, Qi9} = IndexMod:read(4, 7, Qi8),
ok = VerifyReadWithPublishedFun(false, ReadD,
[Four, Five, Six]),
2021-10-11 17:49:11 +08:00
{ReadE, Qi10} = IndexMod:read(7, 9, Qi9),
ok = VerifyReadWithPublishedFun(false, ReadE,
[Seven, Eight]),
Qi10
end),
%% e) as for (d), but use terminate instead of read.
with_empty_test_queue(
fun (Qi0, QName) ->
{Qi1, _SeqIdsMsgIdsE} = queue_index_publish([0,1,2,4,5,7],
true, Qi0),
2021-10-11 17:49:11 +08:00
Qi2 = case IndexMod of
rabbit_queue_index -> IndexMod:deliver([0,1,4], Qi1);
_ -> Qi1
end,
{_DeletedSegments3, Qi3} = IndexMod:ack([0], Qi2),
{5, 50, Qi4} = restart_test_queue(Qi3, QName),
{Qi5, _SeqIdsMsgIdsF} = queue_index_publish([3,6,8], true, Qi4),
2021-10-11 17:49:11 +08:00
Qi6 = case IndexMod of
rabbit_queue_index -> IndexMod:deliver([2,3,5,6], Qi5);
_ -> Qi5
end,
{_DeletedSegments7, Qi7} = IndexMod:ack([1,2,3], Qi6),
{5, 50, Qi8} = restart_test_queue(Qi7, QName),
Qi8
end),
ok = rabbit_variable_queue:stop(?VHOST),
{ok, _} = rabbit_variable_queue:start(?VHOST, []),
passed.
%% The v2 index does not store the MsgId unless required.
%% We therefore do not check it.
verify_read_with_published_v2(_Persistent, [], _) ->
ok;
verify_read_with_published_v2(Persistent,
[{_MsgId1, SeqId, _Location, _Props, Persistent}|Read],
[{SeqId, _MsgId2}|Published]) ->
verify_read_with_published_v2(Persistent, Read, Published);
verify_read_with_published_v2(_Persistent, _Read, _Published) ->
ko.
bq_queue_index_props(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, bq_queue_index_props1, [Config]).
bq_queue_index_props1(_Config) ->
2021-10-11 17:49:11 +08:00
IndexMod = index_mod(),
with_empty_test_queue(
fun(Qi0, _QName) ->
MsgId = rabbit_guid:gen(),
Props = #message_properties{expiry=12345, size = 10},
2021-10-11 17:49:11 +08:00
Qi1 = IndexMod:publish(
MsgId, 0, memory, Props, true, infinity, Qi0),
2021-08-30 20:10:47 +08:00
{[{MsgId, 0, _, Props, _}], Qi2} =
2021-10-11 17:49:11 +08:00
IndexMod:read(0, 1, Qi1),
Qi2
end),
ok = rabbit_variable_queue:stop(?VHOST),
{ok, _} = rabbit_variable_queue:start(?VHOST, []),
passed.
v2_delete_segment_file_completely_acked(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, v2_delete_segment_file_completely_acked1, [Config]).
v2_delete_segment_file_completely_acked1(_Config) ->
IndexMod = rabbit_classic_queue_index_v2,
SegmentSize = IndexMod:next_segment_boundary(0),
SeqIds = lists:seq(0, SegmentSize - 1),
with_empty_test_queue(
fun (Qi0, _QName) ->
%% Publish a full segment file.
{Qi1, SeqIdsMsgIds} = queue_index_publish(SeqIds, true, Qi0),
SegmentSize = length(SeqIdsMsgIds),
{0, SegmentSize, Qi2} = IndexMod:bounds(Qi1),
%% Confirm that the file exists on disk.
Path = IndexMod:segment_file(0, Qi2),
true = filelib:is_file(Path),
%% Ack the full segment file.
{[0], Qi3} = IndexMod:ack(SeqIds, Qi2),
%% Confirm that the file was deleted.
false = filelib:is_file(Path),
Qi3
end),
passed.
v2_delete_segment_file_partially_acked(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, v2_delete_segment_file_partially_acked1, [Config]).
v2_delete_segment_file_partially_acked1(_Config) ->
IndexMod = rabbit_classic_queue_index_v2,
SegmentSize = IndexMod:next_segment_boundary(0),
SeqIds = lists:seq(0, SegmentSize div 2),
SeqIdsLen = length(SeqIds),
with_empty_test_queue(
fun (Qi0, _QName) ->
%% Publish a partial segment file.
{Qi1, SeqIdsMsgIds} = queue_index_publish(SeqIds, true, Qi0),
SeqIdsLen = length(SeqIdsMsgIds),
{0, SegmentSize, Qi2} = IndexMod:bounds(Qi1),
%% Confirm that the file exists on disk.
Path = IndexMod:segment_file(0, Qi2),
true = filelib:is_file(Path),
%% Ack the partial segment file.
{[0], Qi3} = IndexMod:ack(SeqIds, Qi2),
%% Confirm that the file was deleted.
false = filelib:is_file(Path),
Qi3
end),
passed.
v2_delete_segment_file_partially_acked_with_holes(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, v2_delete_segment_file_partially_acked_with_holes1, [Config]).
v2_delete_segment_file_partially_acked_with_holes1(_Config) ->
IndexMod = rabbit_classic_queue_index_v2,
SegmentSize = IndexMod:next_segment_boundary(0),
SeqIdsA = lists:seq(0, SegmentSize div 2),
SeqIdsB = lists:seq(11 + SegmentSize div 2, SegmentSize - 1),
SeqIdsLen = length(SeqIdsA) + length(SeqIdsB),
with_empty_test_queue(
fun (Qi0, _QName) ->
%% Publish a partial segment file with holes.
{Qi1, SeqIdsMsgIdsA} = queue_index_publish(SeqIdsA, true, Qi0),
{Qi2, SeqIdsMsgIdsB} = queue_index_publish(SeqIdsB, true, Qi1),
SeqIdsLen = length(SeqIdsMsgIdsA) + length(SeqIdsMsgIdsB),
{0, SegmentSize, Qi3} = IndexMod:bounds(Qi2),
%% Confirm that the file exists on disk.
Path = IndexMod:segment_file(0, Qi3),
true = filelib:is_file(Path),
%% Ack the partial segment file with holes.
{[], Qi4} = IndexMod:ack(SeqIdsA, Qi3),
{[0], Qi5} = IndexMod:ack(SeqIdsB, Qi4),
%% Confirm that the file was deleted.
false = filelib:is_file(Path),
Qi5
end),
passed.
bq_variable_queue_delete_msg_store_files_callback(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, bq_variable_queue_delete_msg_store_files_callback1, [Config]).
bq_variable_queue_delete_msg_store_files_callback1(Config) ->
ok = restart_msg_store_empty(),
Make #amqqueue{} a private record + add a feature flag The #amqqueue{} record is now isolated in the `amqqueue` module: all accesses and modifications to the record are made through this helper. The type `rabbit_types:amqqueue()` is now `amqqueue:amqqueue()`. `amqqueue` knows about the new #amqqueue{} record introduced with quorum queues. If the actual record is the old one (as defined in RabbitMQ 3.7.x), it calls `amqqueue_v1` which knows about the old definition. When it needs to produce an #amqqueue{} record, it verifies if it is allowed to, using the new Feature flags subsystem and the `quorum_queue` feature flag specifically, and proceeds. If an old format #amqqueue{} is required, the creation is deferred to `amqqueue_v1`. The new modules come with a couple headers: they provide macros to replace pattern matching with `when` conditions. The `amqqueue_v*.hrl` headers are generated using the `macros()` functions in `amqqueue` and `amqqueue_v1` modules. Some operations now depend on the state of the `quorum_queue` feature flag. In particular, creating a quorum queue is denied if the feature flag is disabled. In the process, the following modules were moved from rabbitmq-common to rabbitmq-server: * rabbit_backing_queue * rabbit_queue_master_locator Likewise, the few functions were copied from `rabbit_misc` to `rabbit_amqqueue` because they depend on `amqqueue`. They don't really need to live in rabbitmq-common and they prevent further progress with the feature flags: feature flags need informations about the cluster. The `quorum_queue` feature flag comes with a migration function which is responsible of converting the #amqqueue{} records in Mnesia tables `rabbit_queue` and `rabbit_durable_queue`. Therefore, the upgrade function which did this is removed in this commit. [#159298729]
2018-10-11 18:12:39 +08:00
QName0 = queue_name(Config, <<"bq_variable_queue_delete_msg_store_files_callback-q">>),
{new, Q} = rabbit_amqqueue:declare(QName0, true, false, [], none, <<"acting-user">>),
QName = amqqueue:get_name(Q),
QPid = amqqueue:get_pid(Q),
Payload = <<0:8388608>>, %% 1MB
Count = 30,
Stream Queue This is an aggregated commit of all changes related to the initial implementation of queue types and on top of that the stream queue type. The varios commit messages have simply been included mostly un-edited below. Make rabbit_amqqueue:not_found_or_absent_dirty/1 visible For use in the stream plugin. Use bigger retention policy on max-age test Set coordinator timeout to 30s Handle coordinator unavailable error Handle operator policies as maps when checking if is applicable Add is_policy_applicable/2 to classic queues Ignore restart commands if the stream has been deleted It could happen that after termination some of the monitors are still up and trigger writer/replica restarts Policy support on stream queues Remove subscription events on stream coordinator Ensure old leaders are removed from monitors Introduce delay when retrying a failed phase Note that this ensures monitor is setup, there was a bug where no monitor was really started when re-trying the same phase Restart replicas after leader election instead of relying on old monitors Use timer for stream coordinator retries Fix stream stats for members/online Multiple fixes for replica monitoring and restart Ensure pending commands are appended at the end and re-run Ensure phase is reset with the state Remove duplicates from replica list Restart current phase on state_enter Remove unused import Ensure rabbit is running when checking for stream quorum Restart replicas Add a close/1 function to queue types So that we can get a chance of cleaning up resources if needed. Stream queues close their osiris logs at this point. fix compiler errors stream-queue: take retention into account When calculating ready messages metrics. Add osiris to the list of rabbit deps Retry restart of replicas Do not restart replicas or leaders after receiving a delete cluster command Add more logging to the stream coordinator Monitor subscribed processes on the stream coordinator Memory breakdown for stream queues Update quorum queue event formatter rabbit_msg_record fixes Refactor channel confirms Remove old unconfirmed_messages module that was designed to handle multiple queue fan in logic including all ha mirrors etc. Replaced with simpler rabbit_confirms module that handles the fan out and leaves any queue specific logic (such as confirms from mirrors) to the queue type implemention. Also this module has a dedicated test module. Which is nice. Backward compatibility with 3.8.x events Supports mixed version cluster upgrades Match specification when stream queue already exists Max age retention for stream queues Stop all replicas before starting leader election stream: disallow global qos remove IS_CLASSIC|QUORUM macros Ensure only classic queues are notified on channel down This also removes the delivering_queues map in the channel state as it should not be needed for this and just cause additional unecessary accounting. Polish AMQP 1.0/0.9.1 properties conversion Support byte in application properties, handle 1-bit representation for booleans. Use binary in header for long AMQP 1.0 ID Fix AMQP 1.0 to 0.9.1 conversion Fix test due to incorrect type Convert timestamp application properties to/from seconds AMQP 1.0 uses milliseconds for timestamp and AMQP 0.9.1 uses seconds, so conversion needed. Dialyzer fixes Handle all message-id types AMQP 1.0 is more liberal in it's allowed types of message-id and correlation-id - this adds headers to describe the type of the data in the message_id / correlation_id properties and also handles the case where the data cannot fit by again using headers. Resize stream coordinator cluster when broker configuration changes convert timestamp to and fro seconds user_id should be a binary message annotations keys need to be symbols stream-queue: default exchange and routing key As these won't be present for data written using the rabbitmq-stream plugin. Add exchange, routing key as message annotations To the AMQP 1.0 formatted data to enable roundtrip. Add osiris logging module config And update logging config test suite. Restart election when start of new leader fails The node might have just gone down so we need to try another one Only aux keeps track of phase now, as it might change if the leader election fails Stream coordinator refactor - all state is kept on the ra machine Ensure any ra cluster not a qq is not cleaned up Fixes to recovery and monitoring Add AMQP 1.0 common to dependencies Add rabbit_msg_record module To handle conversions into internal stream storage format. Use rabbitmq-common stream-queue branch Use SSH for osiris dependency Stream coordinator: delete replica Stream coordinator: add replica Stream coordinator: leader failover Stream coordinator: declare and delete Test consuming from a random offset Previous offsets should not be delivered to consumers Consume from stream replicas and multiple test fixes Use max-length-bytes and add new max-segment-size Use SSH for osiris dependency Basic cancel for stream queues Publish stream queues and settle/reject/requeue refactor Consume from stream queues Fix recovery Publish stream messages Add/delete stream replicas Use safe queue names Set retention policy for stream queues Required by the ctl command [#171207092] Stream queue delete queue fix missing callback impl Stream queue declare Queue type abstraction And use the implementing module as the value of the amqqueue record `type` field. This will allow for easy dispatch to the queue type implementation. Queue type abstraction Move queue declare into rabbit_queue_type Move queue delete into queue type implementation Queue type: dequeue/basic_get Move info inside queue type abstraction Move policy change into queue type interface Add purge to queue type Add recovery to the queue type interface Rename amqqueue quorum_nodes field To a more generic an extensible opaque queue type specific map. Fix tests and handle classic API response Fix HA queue confirm bug All mirrors need to be present as queue names. This introduces context linking allowing additional queue refs to be linked to a single "master" queue ref contining the actual queue context. Fix issue with events of deleted queues Also update queue type smoke test to use a cluster by default. correct default value of amqqueue getter Move classic queues further inside queue type interface why [TrackerId] Dialyzer fixes
2020-09-29 18:43:24 +08:00
QTState = publish_and_confirm(Q, Payload, Count),
{ok, Limiter} = rabbit_limiter:start_link(no_id),
CountMinusOne = Count - 1,
Stream Queue This is an aggregated commit of all changes related to the initial implementation of queue types and on top of that the stream queue type. The varios commit messages have simply been included mostly un-edited below. Make rabbit_amqqueue:not_found_or_absent_dirty/1 visible For use in the stream plugin. Use bigger retention policy on max-age test Set coordinator timeout to 30s Handle coordinator unavailable error Handle operator policies as maps when checking if is applicable Add is_policy_applicable/2 to classic queues Ignore restart commands if the stream has been deleted It could happen that after termination some of the monitors are still up and trigger writer/replica restarts Policy support on stream queues Remove subscription events on stream coordinator Ensure old leaders are removed from monitors Introduce delay when retrying a failed phase Note that this ensures monitor is setup, there was a bug where no monitor was really started when re-trying the same phase Restart replicas after leader election instead of relying on old monitors Use timer for stream coordinator retries Fix stream stats for members/online Multiple fixes for replica monitoring and restart Ensure pending commands are appended at the end and re-run Ensure phase is reset with the state Remove duplicates from replica list Restart current phase on state_enter Remove unused import Ensure rabbit is running when checking for stream quorum Restart replicas Add a close/1 function to queue types So that we can get a chance of cleaning up resources if needed. Stream queues close their osiris logs at this point. fix compiler errors stream-queue: take retention into account When calculating ready messages metrics. Add osiris to the list of rabbit deps Retry restart of replicas Do not restart replicas or leaders after receiving a delete cluster command Add more logging to the stream coordinator Monitor subscribed processes on the stream coordinator Memory breakdown for stream queues Update quorum queue event formatter rabbit_msg_record fixes Refactor channel confirms Remove old unconfirmed_messages module that was designed to handle multiple queue fan in logic including all ha mirrors etc. Replaced with simpler rabbit_confirms module that handles the fan out and leaves any queue specific logic (such as confirms from mirrors) to the queue type implemention. Also this module has a dedicated test module. Which is nice. Backward compatibility with 3.8.x events Supports mixed version cluster upgrades Match specification when stream queue already exists Max age retention for stream queues Stop all replicas before starting leader election stream: disallow global qos remove IS_CLASSIC|QUORUM macros Ensure only classic queues are notified on channel down This also removes the delivering_queues map in the channel state as it should not be needed for this and just cause additional unecessary accounting. Polish AMQP 1.0/0.9.1 properties conversion Support byte in application properties, handle 1-bit representation for booleans. Use binary in header for long AMQP 1.0 ID Fix AMQP 1.0 to 0.9.1 conversion Fix test due to incorrect type Convert timestamp application properties to/from seconds AMQP 1.0 uses milliseconds for timestamp and AMQP 0.9.1 uses seconds, so conversion needed. Dialyzer fixes Handle all message-id types AMQP 1.0 is more liberal in it's allowed types of message-id and correlation-id - this adds headers to describe the type of the data in the message_id / correlation_id properties and also handles the case where the data cannot fit by again using headers. Resize stream coordinator cluster when broker configuration changes convert timestamp to and fro seconds user_id should be a binary message annotations keys need to be symbols stream-queue: default exchange and routing key As these won't be present for data written using the rabbitmq-stream plugin. Add exchange, routing key as message annotations To the AMQP 1.0 formatted data to enable roundtrip. Add osiris logging module config And update logging config test suite. Restart election when start of new leader fails The node might have just gone down so we need to try another one Only aux keeps track of phase now, as it might change if the leader election fails Stream coordinator refactor - all state is kept on the ra machine Ensure any ra cluster not a qq is not cleaned up Fixes to recovery and monitoring Add AMQP 1.0 common to dependencies Add rabbit_msg_record module To handle conversions into internal stream storage format. Use rabbitmq-common stream-queue branch Use SSH for osiris dependency Stream coordinator: delete replica Stream coordinator: add replica Stream coordinator: leader failover Stream coordinator: declare and delete Test consuming from a random offset Previous offsets should not be delivered to consumers Consume from stream replicas and multiple test fixes Use max-length-bytes and add new max-segment-size Use SSH for osiris dependency Basic cancel for stream queues Publish stream queues and settle/reject/requeue refactor Consume from stream queues Fix recovery Publish stream messages Add/delete stream replicas Use safe queue names Set retention policy for stream queues Required by the ctl command [#171207092] Stream queue delete queue fix missing callback impl Stream queue declare Queue type abstraction And use the implementing module as the value of the amqqueue record `type` field. This will allow for easy dispatch to the queue type implementation. Queue type abstraction Move queue declare into rabbit_queue_type Move queue delete into queue type implementation Queue type: dequeue/basic_get Move info inside queue type abstraction Move policy change into queue type interface Add purge to queue type Add recovery to the queue type interface Rename amqqueue quorum_nodes field To a more generic an extensible opaque queue type specific map. Fix tests and handle classic API response Fix HA queue confirm bug All mirrors need to be present as queue names. This introduces context linking allowing additional queue refs to be linked to a single "master" queue ref contining the actual queue context. Fix issue with events of deleted queues Also update queue type smoke test to use a cluster by default. correct default value of amqqueue getter Move classic queues further inside queue type interface why [TrackerId] Dialyzer fixes
2020-09-29 18:43:24 +08:00
{ok, CountMinusOne, {QName, QPid, _AckTag, false, _Msg}, _} =
rabbit_amqqueue:basic_get(Q, true, Limiter,
Quorum queues (#1706) * Test queue.declare method with quorum type [#154472130] * Cosmetics [#154472130] * Start quorum queue Includes ra as a rabbit dependency [#154472152] * Update info and list operations to use quorum queues Basic implementation. Might need an update when more functionality is added to the quorum queues. [#154472152] * Stop quorum queue [#154472158] * Restart quorum queue [#154472164] * Introduce UId in ra config to support newer version of ra Improved ra stop [#154472158] * Put data inside VHost specific subdirs [#154472164] * Include ra in rabbit deps to support stop_app/start_app command [#154472164] * Stop quorum queues in `rabbit_amqqueue:stop/1` [#154472158] * Revert creation of fifo ets table inside rabbit Now supported by ra [#154472158] * Filter quorum queues [#154472158] * Test restart node with quorum queues [#154472164] * Publish to quorum queues [#154472174] * Use `ra:restart_node/1` [#154472164] * Wait for stats to be published when querying quorum queues [#154472174] * Test publish and queue length after restart [#154472174] * Consume messages from quorum queues with basic.get [#154472211] * Autoack messages from quorum queues on basic.get [#154472211] * Fix no_ack meaning no_ack = true is equivalent to autoack [#154472211] * Use data_dir as provided in the config If we modify the data_dir, ra is not able to delete the data when a queue is deleted [#154472158] * Remove unused code/variables [#154472158] * Subscribe to a quorum queue Supports auto-ack [#154472215] * Ack messages consumed from quorum queues [#154472221] * Nack messages consumed from quorum queues [#154804608] * Use delivery tag as consumer tag for basic.get in quorum queues [#154472221] * Support for publisher confirms in quorum queues [#154472198] * Integrate with ra_fifo_client * Clear queue state on queue.delete [#154472158] * Fix quorum nack [#154804608] * Test redelivery after nack [#154804608] * Nack without requeueing [#154472225] * Test multiple acks [#154804208] * Test multiple nacks [#154804314] * Configure dead letter exchange with queue declare [#155076661] * Use a per-vhost process to handle dead-lettering Needs to hold state for quorum queues [#155401802] * Implement dead-lettering on nack'ed messages [#154804620] * Use queue name as a resource on message delivery Fixes a previously introduced bug [#154804608] * Handle ra events on dead letter process [#155401802] * Pass empty queue states to queue delete Queue deletion on vhost deletion calls directly to rabbit_amqqueue. Queue states are not available, but we can provide an empty map as in deletion the states are only needed for cleanup. * Generate quorum queue stats and events Consumer delete events are still pending, as depend on basic.cancel (not implemented yet), ra terminating or ra detecting channel down [#154472241] * Ensure quorum mapping entries are available before metric emission [#154472241] * Configure data_dir, uses new RABBITMQ_QUORUM_BASE env var [#154472152] * Use untracked enqueues when sending wihtout channel Updated several other calls missed during the quorum implementation * Revert "Configure data_dir, uses new RABBITMQ_QUORUM_BASE env var" This reverts commit f2261212410affecb238fcbd1fb451381aee4036. * Configure data_dir, uses new RABBITMQ_QUORUM_DIR based on mnesia dir [#154472152] * Fix get_quorum_state * Fix calculation of quorum pids * Move all quorum queues code to its own module [#154472241] * Return an error when declaring a quorum queue with an incompatible argument [#154521696] * Cleanup of quorum queue state after queue delete Also fixes some existing problems where the state wasn't properly stored [#155458625] * Revert Revert "Declare a quorum queue using the queue.declare method" * Remove duplicated state info [#154472241] * Start/stop multi-node quorum queue [#154472231] [#154472236] * Restart nodes in a multi-node quorum cluster [#154472238] * Test restart and leadership takeover on multiple nodes [#154472238] * Wait for leader down after deleting a quorum cluster It ensures an smooth delete-declare sequence without race conditions. The test included here detected the situation before the fix. [#154472236] * Populate quorum_mapping from mnesia when not available Ensures that leader nodes that don't have direct requests can get the mapping ra name -> queue name * Cosmetics * Do not emit core metrics if queue has just been deleted * Use rabbit_mnesia:is_process_alive Fixes bug introduced by cac9583e1bb2705be7f06c2ab7f416a75d11c875 [#154472231] * Only try to report stats if quorum process is alive * Implement cancel consumer callback Deletes metrics and sends consumer deleted event * Remove unnecessary trigger election call ra:restart_node has already been called during the recovery * Apply cancellation callback on node hosting the channel * Cosmetics * Read new fifo metrics which store directly total, ready and unack * Implement basic.cancel for quorum queues * Store leader in amqqueue record, report all in stats [#154472407] * Declare quorum queue in mnesia before starting the ra cluster Record needs to be stored first to update the leader on ra effects * Revert * Purge quorum queues [#154472182] * Improve use of untracked_enqueue Choose the persisted leader id instead of just using the id of the leader at point of creation. * Store quorum leader in the pid field of amqqueue record Same as mirrored queues, no real need for an additional field * Improve recovery When a ra node has never been started on a rabbit node ensure it doesn't fail but instead rebuilds the config and starts the node as a new node. Also fix issue when a quorum queue is declared when one of it's rabbit nodes are unavailable. [#157054606] * Cleanup core metrics after leader change [#157054473] * Return an error on sync_queue on quorum queues [#154472334] * Return an error on cancel_sync_queue on quorum queues [#154472337] * Fix basic_cancel and basic_consume return values Ensure the quorum queue state is always returned by these functions. * Restore arity of amqqeueu delete and purge functions. This avoids some breaking changes in the cli. * Fix bug returning consumers. * remove rogue debug log * Integrate ingress flow control with quorum queues [#157000583] * Configure commands soft limit [#157000583] * Support quorum pids on rabbit_mnesia:is_process_alive * Publish consumers metric for quorum queues * Whitelist quorum directory in is_virgin_node Allow the quorum directoy to exist without affecting the status of the Rabbit node. * Delete queue_metrics on leader change. Also run the become_leader handler in a separate process to avoid blocking. [#157424225] * Report cluster status in quorum queue infos. New per node status command. Related to [#157146500] * Remove quorum_mapping table As we can store the full queue name resource as the cluster id of the ra_fifo_client state we can avoid needed the quorum_mapping table. * Fix xref issue * Provide quorum members information in stats [#157146500] * fix unused variable * quorum queue multiple declare handling Extend rabbit_amqqueue:internal_declare/2 to indicate if the queue record was created or exisiting. From this we can then provide a code path that should handle concurrent queue declares of the same quorum queue. * Return an error when declaring exclusive/auto-delete quorum queue [#157472160] * Restore lost changes from 79c9bd201e1eac006a42bd162e7c86df96496629 * recover another part of commit * fixup cherry pick * Ra io/file metrics handler and stats publishing [#157193081] * Revert "Ra io/file metrics handler and stats publishing" This reverts commit 05d15c786540322583fc655709825db215b70952. * Do not issue confirms on node down for quorum queues. Only a ra_event should be used to issue positive confirms for a quorum queue. * Ra stats publishing [#157193081] * Pick consumer utilisation from ra data [#155402726] * Handle error when deleting a quorum queue and all nodes are already down This is in fact a successful deletion as all raft nodes are already 'stopped' [#158656366] * Return an error when declaring non-durable quorum queues [#158656454] * Rename dirty_query to committed_query * Delete stats on leader node [#158661152] * Give full list of nodes to fifo client * Handle timeout in quorum basic_get * Fix unused variable error * Handle timeout in basic get [#158656366] * Force GC after purge [#158789389] * Increase `ra:delete_cluster` timeout to 120s * Revert "Force GC after purge" This reverts commit 5c98bf22994eb39004760799d3a2c5041d16e9d4. * Add quorum member command [#157481599] * Delete quorum member command [#157481599] * Implement basic.recover for quorum queues [#157597411] * Change concumer utilisation to use the new ra_fifo table and api. * Set max quorum queue size limit Defaults to 7, can be configured per queue on queue.declare Nodes are selected randomly from the list of nodes, but the one that is executing the queue.declare command [#159338081] * remove potentially unrelated changes to rabbit_networking * Move ra_fifo to rabbit Copied ra_fifo to rabbit and renamed it rabbit_fifo. [#159338031] * rabbit_fifo tidy up * rabbit_fifo tidy up * rabbit_fifo: customer -> consumer rename * Move ra_fifo tests [#159338031] * Tweak quorum_queue defaults * quorum_queue test reliability * Optimise quorum_queue test suite. By only starting a rabbit cluster per group rather than test. [#160612638] * Renamings in line with ra API changes * rabbit_fifo fixes * Update with ra API changes Ra has consolidated and simplified it's api. These changes update to confirm to that. * Update rabbit_fifo with latest ra changes * Clean up out of date comment * Return map of states * Add test case for basic.get on an empty queue Before the previous patch, any subsequent basic.get would crash as the map of states had been replaced by a single state. * Clarify use of deliver tags on record_sent * Clean up queues after testcase * Remove erlang monitor of quorum queues in rabbit_channel The eol event can be used instead * Use macros to make clearer distinctions between quorum/classic queues Cosmetic only * Erase queue stats on 'eol' event * Update to follow Ra's cluster_id -> cluster_name rename. * Rename qourum-cluster-size To quorum-initial-group-size * Issue confirms on quorum queue eol Also avoid creating quorum queue session state on queue operation methods. * Only classic queues should be notified on channel down * Quorum queues do not support global qos Exit with protocol error of a basic.consume for a quorum queue is issued on a channel with global qos enabled. * unused variable name * Refactoring Strictly enfornce that channels do not monitor quorum queues. * Refactor foreach_per_queue in the channel. To make it call classic and quorum queues the same way. [#161314899] * rename function * Query classic and quorum queues separately during recovery as they should not be marked as stopped during failed vhost recovery. * Remove force_event_refresh function As the only user of this function, the management API no longer requires it. * fix errors * Remove created_at from amqqueue record [#161343680] * rabbit_fifo: support AMQP 1.0 consumer credit This change implements an alternative consumer credit mechanism similar to AMQP 1.0 link credit where the credit (prefetch) isn't automatically topped up as deliveries are settled and instead needs to be manually increased using a credit command. This is to be integrated with the AMQP 1.0 plugin. [#161256187] * Add basic.credit support for quorum queues. Added support for AMQP 1.0 transfer flow control. [#161256187] * Make quorum queue recover idempotent So that if a vhost crashes and runs the recover steps it doesn't fail because ra servers are still running. [#161343651] * Add tests for vhost deletion To ensure quorum queues are cleaned up on vhost removal. Also fix xref issue. [#161343673] * remove unused clause * always return latest value of queue * Add rabbitmq-queues scripts. Remove ra config from .bat scripts. * Return error if trying to get quorum status of a classic queue.
2018-10-29 17:47:29 +08:00
<<"bq_variable_queue_delete_msg_store_files_callback1">>,
Stream Queue This is an aggregated commit of all changes related to the initial implementation of queue types and on top of that the stream queue type. The varios commit messages have simply been included mostly un-edited below. Make rabbit_amqqueue:not_found_or_absent_dirty/1 visible For use in the stream plugin. Use bigger retention policy on max-age test Set coordinator timeout to 30s Handle coordinator unavailable error Handle operator policies as maps when checking if is applicable Add is_policy_applicable/2 to classic queues Ignore restart commands if the stream has been deleted It could happen that after termination some of the monitors are still up and trigger writer/replica restarts Policy support on stream queues Remove subscription events on stream coordinator Ensure old leaders are removed from monitors Introduce delay when retrying a failed phase Note that this ensures monitor is setup, there was a bug where no monitor was really started when re-trying the same phase Restart replicas after leader election instead of relying on old monitors Use timer for stream coordinator retries Fix stream stats for members/online Multiple fixes for replica monitoring and restart Ensure pending commands are appended at the end and re-run Ensure phase is reset with the state Remove duplicates from replica list Restart current phase on state_enter Remove unused import Ensure rabbit is running when checking for stream quorum Restart replicas Add a close/1 function to queue types So that we can get a chance of cleaning up resources if needed. Stream queues close their osiris logs at this point. fix compiler errors stream-queue: take retention into account When calculating ready messages metrics. Add osiris to the list of rabbit deps Retry restart of replicas Do not restart replicas or leaders after receiving a delete cluster command Add more logging to the stream coordinator Monitor subscribed processes on the stream coordinator Memory breakdown for stream queues Update quorum queue event formatter rabbit_msg_record fixes Refactor channel confirms Remove old unconfirmed_messages module that was designed to handle multiple queue fan in logic including all ha mirrors etc. Replaced with simpler rabbit_confirms module that handles the fan out and leaves any queue specific logic (such as confirms from mirrors) to the queue type implemention. Also this module has a dedicated test module. Which is nice. Backward compatibility with 3.8.x events Supports mixed version cluster upgrades Match specification when stream queue already exists Max age retention for stream queues Stop all replicas before starting leader election stream: disallow global qos remove IS_CLASSIC|QUORUM macros Ensure only classic queues are notified on channel down This also removes the delivering_queues map in the channel state as it should not be needed for this and just cause additional unecessary accounting. Polish AMQP 1.0/0.9.1 properties conversion Support byte in application properties, handle 1-bit representation for booleans. Use binary in header for long AMQP 1.0 ID Fix AMQP 1.0 to 0.9.1 conversion Fix test due to incorrect type Convert timestamp application properties to/from seconds AMQP 1.0 uses milliseconds for timestamp and AMQP 0.9.1 uses seconds, so conversion needed. Dialyzer fixes Handle all message-id types AMQP 1.0 is more liberal in it's allowed types of message-id and correlation-id - this adds headers to describe the type of the data in the message_id / correlation_id properties and also handles the case where the data cannot fit by again using headers. Resize stream coordinator cluster when broker configuration changes convert timestamp to and fro seconds user_id should be a binary message annotations keys need to be symbols stream-queue: default exchange and routing key As these won't be present for data written using the rabbitmq-stream plugin. Add exchange, routing key as message annotations To the AMQP 1.0 formatted data to enable roundtrip. Add osiris logging module config And update logging config test suite. Restart election when start of new leader fails The node might have just gone down so we need to try another one Only aux keeps track of phase now, as it might change if the leader election fails Stream coordinator refactor - all state is kept on the ra machine Ensure any ra cluster not a qq is not cleaned up Fixes to recovery and monitoring Add AMQP 1.0 common to dependencies Add rabbit_msg_record module To handle conversions into internal stream storage format. Use rabbitmq-common stream-queue branch Use SSH for osiris dependency Stream coordinator: delete replica Stream coordinator: add replica Stream coordinator: leader failover Stream coordinator: declare and delete Test consuming from a random offset Previous offsets should not be delivered to consumers Consume from stream replicas and multiple test fixes Use max-length-bytes and add new max-segment-size Use SSH for osiris dependency Basic cancel for stream queues Publish stream queues and settle/reject/requeue refactor Consume from stream queues Fix recovery Publish stream messages Add/delete stream replicas Use safe queue names Set retention policy for stream queues Required by the ctl command [#171207092] Stream queue delete queue fix missing callback impl Stream queue declare Queue type abstraction And use the implementing module as the value of the amqqueue record `type` field. This will allow for easy dispatch to the queue type implementation. Queue type abstraction Move queue declare into rabbit_queue_type Move queue delete into queue type implementation Queue type: dequeue/basic_get Move info inside queue type abstraction Move policy change into queue type interface Add purge to queue type Add recovery to the queue type interface Rename amqqueue quorum_nodes field To a more generic an extensible opaque queue type specific map. Fix tests and handle classic API response Fix HA queue confirm bug All mirrors need to be present as queue names. This introduces context linking allowing additional queue refs to be linked to a single "master" queue ref contining the actual queue context. Fix issue with events of deleted queues Also update queue type smoke test to use a cluster by default. correct default value of amqqueue getter Move classic queues further inside queue type interface why [TrackerId] Dialyzer fixes
2020-09-29 18:43:24 +08:00
QTState),
{ok, CountMinusOne} = rabbit_amqqueue:purge(Q),
%% give the queue a second to receive the close_fds callback msg
timer:sleep(1000),
2017-03-09 23:05:37 +08:00
rabbit_amqqueue:delete(Q, false, false, <<"acting-user">>),
passed.
bq_queue_recover(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, bq_queue_recover1, [Config]).
bq_queue_recover1(Config) ->
init_queue_index(),
2021-10-11 17:49:11 +08:00
IndexMod = index_mod(),
Count = 2 * IndexMod:next_segment_boundary(0),
Make #amqqueue{} a private record + add a feature flag The #amqqueue{} record is now isolated in the `amqqueue` module: all accesses and modifications to the record are made through this helper. The type `rabbit_types:amqqueue()` is now `amqqueue:amqqueue()`. `amqqueue` knows about the new #amqqueue{} record introduced with quorum queues. If the actual record is the old one (as defined in RabbitMQ 3.7.x), it calls `amqqueue_v1` which knows about the old definition. When it needs to produce an #amqqueue{} record, it verifies if it is allowed to, using the new Feature flags subsystem and the `quorum_queue` feature flag specifically, and proceeds. If an old format #amqqueue{} is required, the creation is deferred to `amqqueue_v1`. The new modules come with a couple headers: they provide macros to replace pattern matching with `when` conditions. The `amqqueue_v*.hrl` headers are generated using the `macros()` functions in `amqqueue` and `amqqueue_v1` modules. Some operations now depend on the state of the `quorum_queue` feature flag. In particular, creating a quorum queue is denied if the feature flag is disabled. In the process, the following modules were moved from rabbitmq-common to rabbitmq-server: * rabbit_backing_queue * rabbit_queue_master_locator Likewise, the few functions were copied from `rabbit_misc` to `rabbit_amqqueue` because they depend on `amqqueue`. They don't really need to live in rabbitmq-common and they prevent further progress with the feature flags: feature flags need informations about the cluster. The `quorum_queue` feature flag comes with a migration function which is responsible of converting the #amqqueue{} records in Mnesia tables `rabbit_queue` and `rabbit_durable_queue`. Therefore, the upgrade function which did this is removed in this commit. [#159298729]
2018-10-11 18:12:39 +08:00
QName0 = queue_name(Config, <<"bq_queue_recover-q">>),
{new, Q} = rabbit_amqqueue:declare(QName0, true, false, [], none, <<"acting-user">>),
QName = amqqueue:get_name(Q),
QPid = amqqueue:get_pid(Q),
Stream Queue This is an aggregated commit of all changes related to the initial implementation of queue types and on top of that the stream queue type. The varios commit messages have simply been included mostly un-edited below. Make rabbit_amqqueue:not_found_or_absent_dirty/1 visible For use in the stream plugin. Use bigger retention policy on max-age test Set coordinator timeout to 30s Handle coordinator unavailable error Handle operator policies as maps when checking if is applicable Add is_policy_applicable/2 to classic queues Ignore restart commands if the stream has been deleted It could happen that after termination some of the monitors are still up and trigger writer/replica restarts Policy support on stream queues Remove subscription events on stream coordinator Ensure old leaders are removed from monitors Introduce delay when retrying a failed phase Note that this ensures monitor is setup, there was a bug where no monitor was really started when re-trying the same phase Restart replicas after leader election instead of relying on old monitors Use timer for stream coordinator retries Fix stream stats for members/online Multiple fixes for replica monitoring and restart Ensure pending commands are appended at the end and re-run Ensure phase is reset with the state Remove duplicates from replica list Restart current phase on state_enter Remove unused import Ensure rabbit is running when checking for stream quorum Restart replicas Add a close/1 function to queue types So that we can get a chance of cleaning up resources if needed. Stream queues close their osiris logs at this point. fix compiler errors stream-queue: take retention into account When calculating ready messages metrics. Add osiris to the list of rabbit deps Retry restart of replicas Do not restart replicas or leaders after receiving a delete cluster command Add more logging to the stream coordinator Monitor subscribed processes on the stream coordinator Memory breakdown for stream queues Update quorum queue event formatter rabbit_msg_record fixes Refactor channel confirms Remove old unconfirmed_messages module that was designed to handle multiple queue fan in logic including all ha mirrors etc. Replaced with simpler rabbit_confirms module that handles the fan out and leaves any queue specific logic (such as confirms from mirrors) to the queue type implemention. Also this module has a dedicated test module. Which is nice. Backward compatibility with 3.8.x events Supports mixed version cluster upgrades Match specification when stream queue already exists Max age retention for stream queues Stop all replicas before starting leader election stream: disallow global qos remove IS_CLASSIC|QUORUM macros Ensure only classic queues are notified on channel down This also removes the delivering_queues map in the channel state as it should not be needed for this and just cause additional unecessary accounting. Polish AMQP 1.0/0.9.1 properties conversion Support byte in application properties, handle 1-bit representation for booleans. Use binary in header for long AMQP 1.0 ID Fix AMQP 1.0 to 0.9.1 conversion Fix test due to incorrect type Convert timestamp application properties to/from seconds AMQP 1.0 uses milliseconds for timestamp and AMQP 0.9.1 uses seconds, so conversion needed. Dialyzer fixes Handle all message-id types AMQP 1.0 is more liberal in it's allowed types of message-id and correlation-id - this adds headers to describe the type of the data in the message_id / correlation_id properties and also handles the case where the data cannot fit by again using headers. Resize stream coordinator cluster when broker configuration changes convert timestamp to and fro seconds user_id should be a binary message annotations keys need to be symbols stream-queue: default exchange and routing key As these won't be present for data written using the rabbitmq-stream plugin. Add exchange, routing key as message annotations To the AMQP 1.0 formatted data to enable roundtrip. Add osiris logging module config And update logging config test suite. Restart election when start of new leader fails The node might have just gone down so we need to try another one Only aux keeps track of phase now, as it might change if the leader election fails Stream coordinator refactor - all state is kept on the ra machine Ensure any ra cluster not a qq is not cleaned up Fixes to recovery and monitoring Add AMQP 1.0 common to dependencies Add rabbit_msg_record module To handle conversions into internal stream storage format. Use rabbitmq-common stream-queue branch Use SSH for osiris dependency Stream coordinator: delete replica Stream coordinator: add replica Stream coordinator: leader failover Stream coordinator: declare and delete Test consuming from a random offset Previous offsets should not be delivered to consumers Consume from stream replicas and multiple test fixes Use max-length-bytes and add new max-segment-size Use SSH for osiris dependency Basic cancel for stream queues Publish stream queues and settle/reject/requeue refactor Consume from stream queues Fix recovery Publish stream messages Add/delete stream replicas Use safe queue names Set retention policy for stream queues Required by the ctl command [#171207092] Stream queue delete queue fix missing callback impl Stream queue declare Queue type abstraction And use the implementing module as the value of the amqqueue record `type` field. This will allow for easy dispatch to the queue type implementation. Queue type abstraction Move queue declare into rabbit_queue_type Move queue delete into queue type implementation Queue type: dequeue/basic_get Move info inside queue type abstraction Move policy change into queue type interface Add purge to queue type Add recovery to the queue type interface Rename amqqueue quorum_nodes field To a more generic an extensible opaque queue type specific map. Fix tests and handle classic API response Fix HA queue confirm bug All mirrors need to be present as queue names. This introduces context linking allowing additional queue refs to be linked to a single "master" queue ref contining the actual queue context. Fix issue with events of deleted queues Also update queue type smoke test to use a cluster by default. correct default value of amqqueue getter Move classic queues further inside queue type interface why [TrackerId] Dialyzer fixes
2020-09-29 18:43:24 +08:00
QT = publish_and_confirm(Q, <<>>, Count),
SupPid = get_queue_sup_pid(Q),
true = is_pid(SupPid),
exit(SupPid, kill),
exit(QPid, kill),
MRef = erlang:monitor(process, QPid),
receive {'DOWN', MRef, process, QPid, _Info} -> ok
after 10000 -> exit(timeout_waiting_for_queue_death)
end,
rabbit_amqqueue:stop(?VHOST),
Stream Queue This is an aggregated commit of all changes related to the initial implementation of queue types and on top of that the stream queue type. The varios commit messages have simply been included mostly un-edited below. Make rabbit_amqqueue:not_found_or_absent_dirty/1 visible For use in the stream plugin. Use bigger retention policy on max-age test Set coordinator timeout to 30s Handle coordinator unavailable error Handle operator policies as maps when checking if is applicable Add is_policy_applicable/2 to classic queues Ignore restart commands if the stream has been deleted It could happen that after termination some of the monitors are still up and trigger writer/replica restarts Policy support on stream queues Remove subscription events on stream coordinator Ensure old leaders are removed from monitors Introduce delay when retrying a failed phase Note that this ensures monitor is setup, there was a bug where no monitor was really started when re-trying the same phase Restart replicas after leader election instead of relying on old monitors Use timer for stream coordinator retries Fix stream stats for members/online Multiple fixes for replica monitoring and restart Ensure pending commands are appended at the end and re-run Ensure phase is reset with the state Remove duplicates from replica list Restart current phase on state_enter Remove unused import Ensure rabbit is running when checking for stream quorum Restart replicas Add a close/1 function to queue types So that we can get a chance of cleaning up resources if needed. Stream queues close their osiris logs at this point. fix compiler errors stream-queue: take retention into account When calculating ready messages metrics. Add osiris to the list of rabbit deps Retry restart of replicas Do not restart replicas or leaders after receiving a delete cluster command Add more logging to the stream coordinator Monitor subscribed processes on the stream coordinator Memory breakdown for stream queues Update quorum queue event formatter rabbit_msg_record fixes Refactor channel confirms Remove old unconfirmed_messages module that was designed to handle multiple queue fan in logic including all ha mirrors etc. Replaced with simpler rabbit_confirms module that handles the fan out and leaves any queue specific logic (such as confirms from mirrors) to the queue type implemention. Also this module has a dedicated test module. Which is nice. Backward compatibility with 3.8.x events Supports mixed version cluster upgrades Match specification when stream queue already exists Max age retention for stream queues Stop all replicas before starting leader election stream: disallow global qos remove IS_CLASSIC|QUORUM macros Ensure only classic queues are notified on channel down This also removes the delivering_queues map in the channel state as it should not be needed for this and just cause additional unecessary accounting. Polish AMQP 1.0/0.9.1 properties conversion Support byte in application properties, handle 1-bit representation for booleans. Use binary in header for long AMQP 1.0 ID Fix AMQP 1.0 to 0.9.1 conversion Fix test due to incorrect type Convert timestamp application properties to/from seconds AMQP 1.0 uses milliseconds for timestamp and AMQP 0.9.1 uses seconds, so conversion needed. Dialyzer fixes Handle all message-id types AMQP 1.0 is more liberal in it's allowed types of message-id and correlation-id - this adds headers to describe the type of the data in the message_id / correlation_id properties and also handles the case where the data cannot fit by again using headers. Resize stream coordinator cluster when broker configuration changes convert timestamp to and fro seconds user_id should be a binary message annotations keys need to be symbols stream-queue: default exchange and routing key As these won't be present for data written using the rabbitmq-stream plugin. Add exchange, routing key as message annotations To the AMQP 1.0 formatted data to enable roundtrip. Add osiris logging module config And update logging config test suite. Restart election when start of new leader fails The node might have just gone down so we need to try another one Only aux keeps track of phase now, as it might change if the leader election fails Stream coordinator refactor - all state is kept on the ra machine Ensure any ra cluster not a qq is not cleaned up Fixes to recovery and monitoring Add AMQP 1.0 common to dependencies Add rabbit_msg_record module To handle conversions into internal stream storage format. Use rabbitmq-common stream-queue branch Use SSH for osiris dependency Stream coordinator: delete replica Stream coordinator: add replica Stream coordinator: leader failover Stream coordinator: declare and delete Test consuming from a random offset Previous offsets should not be delivered to consumers Consume from stream replicas and multiple test fixes Use max-length-bytes and add new max-segment-size Use SSH for osiris dependency Basic cancel for stream queues Publish stream queues and settle/reject/requeue refactor Consume from stream queues Fix recovery Publish stream messages Add/delete stream replicas Use safe queue names Set retention policy for stream queues Required by the ctl command [#171207092] Stream queue delete queue fix missing callback impl Stream queue declare Queue type abstraction And use the implementing module as the value of the amqqueue record `type` field. This will allow for easy dispatch to the queue type implementation. Queue type abstraction Move queue declare into rabbit_queue_type Move queue delete into queue type implementation Queue type: dequeue/basic_get Move info inside queue type abstraction Move policy change into queue type interface Add purge to queue type Add recovery to the queue type interface Rename amqqueue quorum_nodes field To a more generic an extensible opaque queue type specific map. Fix tests and handle classic API response Fix HA queue confirm bug All mirrors need to be present as queue names. This introduces context linking allowing additional queue refs to be linked to a single "master" queue ref contining the actual queue context. Fix issue with events of deleted queues Also update queue type smoke test to use a cluster by default. correct default value of amqqueue getter Move classic queues further inside queue type interface why [TrackerId] Dialyzer fixes
2020-09-29 18:43:24 +08:00
{Recovered, []} = rabbit_amqqueue:recover(?VHOST),
rabbit_amqqueue:start(Recovered),
{ok, Limiter} = rabbit_limiter:start_link(no_id),
rabbit_amqqueue:with_or_die(
QName,
Make #amqqueue{} a private record + add a feature flag The #amqqueue{} record is now isolated in the `amqqueue` module: all accesses and modifications to the record are made through this helper. The type `rabbit_types:amqqueue()` is now `amqqueue:amqqueue()`. `amqqueue` knows about the new #amqqueue{} record introduced with quorum queues. If the actual record is the old one (as defined in RabbitMQ 3.7.x), it calls `amqqueue_v1` which knows about the old definition. When it needs to produce an #amqqueue{} record, it verifies if it is allowed to, using the new Feature flags subsystem and the `quorum_queue` feature flag specifically, and proceeds. If an old format #amqqueue{} is required, the creation is deferred to `amqqueue_v1`. The new modules come with a couple headers: they provide macros to replace pattern matching with `when` conditions. The `amqqueue_v*.hrl` headers are generated using the `macros()` functions in `amqqueue` and `amqqueue_v1` modules. Some operations now depend on the state of the `quorum_queue` feature flag. In particular, creating a quorum queue is denied if the feature flag is disabled. In the process, the following modules were moved from rabbitmq-common to rabbitmq-server: * rabbit_backing_queue * rabbit_queue_master_locator Likewise, the few functions were copied from `rabbit_misc` to `rabbit_amqqueue` because they depend on `amqqueue`. They don't really need to live in rabbitmq-common and they prevent further progress with the feature flags: feature flags need informations about the cluster. The `quorum_queue` feature flag comes with a migration function which is responsible of converting the #amqqueue{} records in Mnesia tables `rabbit_queue` and `rabbit_durable_queue`. Therefore, the upgrade function which did this is removed in this commit. [#159298729]
2018-10-11 18:12:39 +08:00
fun (Q1) when ?is_amqqueue(Q1) ->
QPid1 = amqqueue:get_pid(Q1),
CountMinusOne = Count - 1,
Stream Queue This is an aggregated commit of all changes related to the initial implementation of queue types and on top of that the stream queue type. The varios commit messages have simply been included mostly un-edited below. Make rabbit_amqqueue:not_found_or_absent_dirty/1 visible For use in the stream plugin. Use bigger retention policy on max-age test Set coordinator timeout to 30s Handle coordinator unavailable error Handle operator policies as maps when checking if is applicable Add is_policy_applicable/2 to classic queues Ignore restart commands if the stream has been deleted It could happen that after termination some of the monitors are still up and trigger writer/replica restarts Policy support on stream queues Remove subscription events on stream coordinator Ensure old leaders are removed from monitors Introduce delay when retrying a failed phase Note that this ensures monitor is setup, there was a bug where no monitor was really started when re-trying the same phase Restart replicas after leader election instead of relying on old monitors Use timer for stream coordinator retries Fix stream stats for members/online Multiple fixes for replica monitoring and restart Ensure pending commands are appended at the end and re-run Ensure phase is reset with the state Remove duplicates from replica list Restart current phase on state_enter Remove unused import Ensure rabbit is running when checking for stream quorum Restart replicas Add a close/1 function to queue types So that we can get a chance of cleaning up resources if needed. Stream queues close their osiris logs at this point. fix compiler errors stream-queue: take retention into account When calculating ready messages metrics. Add osiris to the list of rabbit deps Retry restart of replicas Do not restart replicas or leaders after receiving a delete cluster command Add more logging to the stream coordinator Monitor subscribed processes on the stream coordinator Memory breakdown for stream queues Update quorum queue event formatter rabbit_msg_record fixes Refactor channel confirms Remove old unconfirmed_messages module that was designed to handle multiple queue fan in logic including all ha mirrors etc. Replaced with simpler rabbit_confirms module that handles the fan out and leaves any queue specific logic (such as confirms from mirrors) to the queue type implemention. Also this module has a dedicated test module. Which is nice. Backward compatibility with 3.8.x events Supports mixed version cluster upgrades Match specification when stream queue already exists Max age retention for stream queues Stop all replicas before starting leader election stream: disallow global qos remove IS_CLASSIC|QUORUM macros Ensure only classic queues are notified on channel down This also removes the delivering_queues map in the channel state as it should not be needed for this and just cause additional unecessary accounting. Polish AMQP 1.0/0.9.1 properties conversion Support byte in application properties, handle 1-bit representation for booleans. Use binary in header for long AMQP 1.0 ID Fix AMQP 1.0 to 0.9.1 conversion Fix test due to incorrect type Convert timestamp application properties to/from seconds AMQP 1.0 uses milliseconds for timestamp and AMQP 0.9.1 uses seconds, so conversion needed. Dialyzer fixes Handle all message-id types AMQP 1.0 is more liberal in it's allowed types of message-id and correlation-id - this adds headers to describe the type of the data in the message_id / correlation_id properties and also handles the case where the data cannot fit by again using headers. Resize stream coordinator cluster when broker configuration changes convert timestamp to and fro seconds user_id should be a binary message annotations keys need to be symbols stream-queue: default exchange and routing key As these won't be present for data written using the rabbitmq-stream plugin. Add exchange, routing key as message annotations To the AMQP 1.0 formatted data to enable roundtrip. Add osiris logging module config And update logging config test suite. Restart election when start of new leader fails The node might have just gone down so we need to try another one Only aux keeps track of phase now, as it might change if the leader election fails Stream coordinator refactor - all state is kept on the ra machine Ensure any ra cluster not a qq is not cleaned up Fixes to recovery and monitoring Add AMQP 1.0 common to dependencies Add rabbit_msg_record module To handle conversions into internal stream storage format. Use rabbitmq-common stream-queue branch Use SSH for osiris dependency Stream coordinator: delete replica Stream coordinator: add replica Stream coordinator: leader failover Stream coordinator: declare and delete Test consuming from a random offset Previous offsets should not be delivered to consumers Consume from stream replicas and multiple test fixes Use max-length-bytes and add new max-segment-size Use SSH for osiris dependency Basic cancel for stream queues Publish stream queues and settle/reject/requeue refactor Consume from stream queues Fix recovery Publish stream messages Add/delete stream replicas Use safe queue names Set retention policy for stream queues Required by the ctl command [#171207092] Stream queue delete queue fix missing callback impl Stream queue declare Queue type abstraction And use the implementing module as the value of the amqqueue record `type` field. This will allow for easy dispatch to the queue type implementation. Queue type abstraction Move queue declare into rabbit_queue_type Move queue delete into queue type implementation Queue type: dequeue/basic_get Move info inside queue type abstraction Move policy change into queue type interface Add purge to queue type Add recovery to the queue type interface Rename amqqueue quorum_nodes field To a more generic an extensible opaque queue type specific map. Fix tests and handle classic API response Fix HA queue confirm bug All mirrors need to be present as queue names. This introduces context linking allowing additional queue refs to be linked to a single "master" queue ref contining the actual queue context. Fix issue with events of deleted queues Also update queue type smoke test to use a cluster by default. correct default value of amqqueue getter Move classic queues further inside queue type interface why [TrackerId] Dialyzer fixes
2020-09-29 18:43:24 +08:00
{ok, CountMinusOne, {QName, QPid1, _AckTag, true, _Msg}, _} =
rabbit_amqqueue:basic_get(Q1, false, Limiter,
<<"bq_queue_recover1">>, QT),
exit(QPid1, shutdown),
VQ1 = variable_queue_init(Q, true),
{{_Msg1, true, _AckTag1}, VQ2} =
rabbit_variable_queue:fetch(true, VQ1),
CountMinusOne = rabbit_variable_queue:len(VQ2),
_VQ3 = rabbit_variable_queue:delete_and_terminate(shutdown, VQ2),
ok = rabbit_amqqueue:internal_delete(Q1, <<"acting-user">>)
end),
passed.
%% Return the PID of the given queue's supervisor.
Make #amqqueue{} a private record + add a feature flag The #amqqueue{} record is now isolated in the `amqqueue` module: all accesses and modifications to the record are made through this helper. The type `rabbit_types:amqqueue()` is now `amqqueue:amqqueue()`. `amqqueue` knows about the new #amqqueue{} record introduced with quorum queues. If the actual record is the old one (as defined in RabbitMQ 3.7.x), it calls `amqqueue_v1` which knows about the old definition. When it needs to produce an #amqqueue{} record, it verifies if it is allowed to, using the new Feature flags subsystem and the `quorum_queue` feature flag specifically, and proceeds. If an old format #amqqueue{} is required, the creation is deferred to `amqqueue_v1`. The new modules come with a couple headers: they provide macros to replace pattern matching with `when` conditions. The `amqqueue_v*.hrl` headers are generated using the `macros()` functions in `amqqueue` and `amqqueue_v1` modules. Some operations now depend on the state of the `quorum_queue` feature flag. In particular, creating a quorum queue is denied if the feature flag is disabled. In the process, the following modules were moved from rabbitmq-common to rabbitmq-server: * rabbit_backing_queue * rabbit_queue_master_locator Likewise, the few functions were copied from `rabbit_misc` to `rabbit_amqqueue` because they depend on `amqqueue`. They don't really need to live in rabbitmq-common and they prevent further progress with the feature flags: feature flags need informations about the cluster. The `quorum_queue` feature flag comes with a migration function which is responsible of converting the #amqqueue{} records in Mnesia tables `rabbit_queue` and `rabbit_durable_queue`. Therefore, the upgrade function which did this is removed in this commit. [#159298729]
2018-10-11 18:12:39 +08:00
get_queue_sup_pid(Q) when ?is_amqqueue(Q) ->
QName = amqqueue:get_name(Q),
QPid = amqqueue:get_pid(Q),
VHost = QName#resource.virtual_host,
{ok, AmqSup} = rabbit_amqqueue_sup_sup:find_for_vhost(VHost, node(QPid)),
Sups = supervisor:which_children(AmqSup),
get_queue_sup_pid(Sups, QPid).
get_queue_sup_pid([{_, SupPid, _, _} | Rest], QueuePid) ->
WorkerPids = [Pid || {_, Pid, _, _} <- supervisor:which_children(SupPid)],
case lists:member(QueuePid, WorkerPids) of
true -> SupPid;
false -> get_queue_sup_pid(Rest, QueuePid)
end;
get_queue_sup_pid([], _QueuePid) ->
undefined.
variable_queue_partial_segments_delta_thing(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, variable_queue_partial_segments_delta_thing1, [Config]).
variable_queue_partial_segments_delta_thing1(Config) ->
with_fresh_variable_queue(
fun variable_queue_partial_segments_delta_thing2/2,
?config(variable_queue_type, Config)).
variable_queue_partial_segments_delta_thing2(VQ0, _QName) ->
2021-10-11 17:49:11 +08:00
IndexMod = index_mod(),
SegmentSize = IndexMod:next_segment_boundary(0),
HalfSegment = SegmentSize div 2,
OneAndAHalfSegment = SegmentSize + HalfSegment,
VQ1 = variable_queue_publish(true, OneAndAHalfSegment, VQ0),
VQ2 = rabbit_variable_queue:update_rates(VQ1),
VQ3 = check_variable_queue_status(
VQ2,
%% We only have one message in memory because the amount in memory
%% depends on the consume rate, which is nil in this test.
[{delta, {delta, 1, OneAndAHalfSegment - 1, 0, OneAndAHalfSegment}},
{q3, 1},
{len, OneAndAHalfSegment}]),
VQ5 = check_variable_queue_status(
variable_queue_publish(true, 1, VQ3),
%% one alpha, but it's in the same segment as the deltas
%% @todo That's wrong now! v1/v2
[{delta, {delta, 1, OneAndAHalfSegment, 0, OneAndAHalfSegment + 1}},
{q3, 1},
{len, OneAndAHalfSegment + 1}]),
{VQ6, AckTags} = variable_queue_fetch(SegmentSize, true, false,
SegmentSize + HalfSegment + 1, VQ5),
VQ7 = check_variable_queue_status(
VQ6,
%% We only read from delta up to the end of the segment, so
%% after fetching exactly one segment, we should have no
%% messages in memory.
[{delta, {delta, SegmentSize, HalfSegment + 1, 0, OneAndAHalfSegment + 1}},
{q3, 0},
{len, HalfSegment + 1}]),
{VQ8, AckTags1} = variable_queue_fetch(HalfSegment + 1, true, false,
HalfSegment + 1, VQ7),
{_Guids, VQ9} = rabbit_variable_queue:ack(AckTags ++ AckTags1, VQ8),
%% should be empty now
{empty, VQ10} = rabbit_variable_queue:fetch(true, VQ9),
VQ10.
variable_queue_all_the_bits_not_covered_elsewhere_A(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, variable_queue_all_the_bits_not_covered_elsewhere_A1, [Config]).
variable_queue_all_the_bits_not_covered_elsewhere_A1(Config) ->
with_fresh_variable_queue(
fun variable_queue_all_the_bits_not_covered_elsewhere_A2/2,
?config(variable_queue_type, Config)).
variable_queue_all_the_bits_not_covered_elsewhere_A2(VQ0, QName) ->
2021-10-11 17:49:11 +08:00
IndexMod = index_mod(),
Count = 2 * IndexMod:next_segment_boundary(0),
VQ1 = variable_queue_publish(true, Count, VQ0),
VQ2 = variable_queue_publish(false, Count, VQ1),
{VQ4, _AckTags} = variable_queue_fetch(Count, true, false,
Count + Count, VQ2),
{VQ5, _AckTags1} = variable_queue_fetch(Count, false, false,
Count, VQ4),
_VQ6 = rabbit_variable_queue:terminate(shutdown, VQ5),
VQ7 = variable_queue_init(test_amqqueue(QName, true), true),
{{_Msg1, true, _AckTag1}, VQ8} = rabbit_variable_queue:fetch(true, VQ7),
Count1 = rabbit_variable_queue:len(VQ8),
VQ9 = variable_queue_publish(false, 1, VQ8),
{VQ11, _AckTags2} = variable_queue_fetch(Count1, true, true, Count, VQ9),
{VQ12, _AckTags3} = variable_queue_fetch(1, false, false, 1, VQ11),
VQ12.
variable_queue_all_the_bits_not_covered_elsewhere_B(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, variable_queue_all_the_bits_not_covered_elsewhere_B1, [Config]).
variable_queue_all_the_bits_not_covered_elsewhere_B1(Config) ->
with_fresh_variable_queue(
fun variable_queue_all_the_bits_not_covered_elsewhere_B2/2,
?config(variable_queue_type, Config)).
variable_queue_all_the_bits_not_covered_elsewhere_B2(VQ1, QName) ->
VQ2 = variable_queue_publish(false, 4, VQ1),
{VQ3, AckTags} = variable_queue_fetch(2, false, false, 4, VQ2),
{_Guids, VQ4} =
rabbit_variable_queue:requeue(AckTags, VQ3),
VQ5 = rabbit_variable_queue:timeout(VQ4),
_VQ6 = rabbit_variable_queue:terminate(shutdown, VQ5),
VQ7 = variable_queue_init(test_amqqueue(QName, true), true),
{empty, VQ8} = rabbit_variable_queue:fetch(false, VQ7),
VQ8.
variable_queue_drop(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, variable_queue_drop1, [Config]).
variable_queue_drop1(Config) ->
with_fresh_variable_queue(
fun variable_queue_drop2/2,
?config(variable_queue_type, Config)).
variable_queue_drop2(VQ0, _QName) ->
%% start by sending a messages
VQ1 = variable_queue_publish(false, 1, VQ0),
%% drop message with AckRequired = true
{{MsgId, AckTag}, VQ2} = rabbit_variable_queue:drop(true, VQ1),
true = rabbit_variable_queue:is_empty(VQ2),
true = AckTag =/= undefinded,
%% drop again -> empty
{empty, VQ3} = rabbit_variable_queue:drop(false, VQ2),
%% requeue
{[MsgId], VQ4} = rabbit_variable_queue:requeue([AckTag], VQ3),
%% drop message with AckRequired = false
{{MsgId, undefined}, VQ5} = rabbit_variable_queue:drop(false, VQ4),
true = rabbit_variable_queue:is_empty(VQ5),
VQ5.
variable_queue_fold_msg_on_disk(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, variable_queue_fold_msg_on_disk1, [Config]).
variable_queue_fold_msg_on_disk1(Config) ->
with_fresh_variable_queue(
fun variable_queue_fold_msg_on_disk2/2,
?config(variable_queue_type, Config)).
variable_queue_fold_msg_on_disk2(VQ0, _QName) ->
VQ1 = variable_queue_publish(true, 1, VQ0),
{VQ2, AckTags} = variable_queue_fetch(1, true, false, 1, VQ1),
{ok, VQ3} = rabbit_variable_queue:ackfold(fun (_M, _A, ok) -> ok end,
ok, VQ2, AckTags),
VQ3.
variable_queue_dropfetchwhile(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, variable_queue_dropfetchwhile1, [Config]).
variable_queue_dropfetchwhile1(Config) ->
with_fresh_variable_queue(
fun variable_queue_dropfetchwhile2/2,
?config(variable_queue_type, Config)).
variable_queue_dropfetchwhile2(VQ0, _QName) ->
Count = 10,
%% add messages with sequential expiry
VQ1 = variable_queue_publish(
false, 1, Count,
fun (N, Props) -> Props#message_properties{expiry = N} end,
fun erlang:term_to_binary/1, VQ0),
%% fetch the first 5 messages
{#message_properties{expiry = 6}, {Msgs, AckTags}, VQ2} =
rabbit_variable_queue:fetchwhile(
fun (#message_properties{expiry = Expiry}) -> Expiry =< 5 end,
fun (Msg, AckTag, {MsgAcc, AckAcc}) ->
{[Msg | MsgAcc], [AckTag | AckAcc]}
end, {[], []}, VQ1),
true = lists:seq(1, 5) == [msg2int(M) || M <- lists:reverse(Msgs)],
%% requeue them
{_MsgIds, VQ3} = rabbit_variable_queue:requeue(AckTags, VQ2),
%% drop the first 5 messages
{#message_properties{expiry = 6}, VQ4} =
rabbit_variable_queue:dropwhile(
fun (#message_properties {expiry = Expiry}) -> Expiry =< 5 end, VQ3),
%% fetch 5
VQ5 = lists:foldl(fun (N, VQN) ->
{{Msg, _, _}, VQM} =
rabbit_variable_queue:fetch(false, VQN),
true = msg2int(Msg) == N,
VQM
end, VQ4, lists:seq(6, Count)),
%% should be empty now
true = rabbit_variable_queue:is_empty(VQ5),
VQ5.
variable_queue_dropwhile_restart(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, variable_queue_dropwhile_restart1, [Config]).
variable_queue_dropwhile_restart1(Config) ->
with_fresh_variable_queue(
fun variable_queue_dropwhile_restart2/2,
?config(variable_queue_type, Config)).
variable_queue_dropwhile_restart2(VQ0, QName) ->
Count = 10000,
%% add messages with sequential expiry
VQ1 = variable_queue_publish(
true, 1, Count,
fun (N, Props) -> Props#message_properties{expiry = N} end,
fun erlang:term_to_binary/1, VQ0),
%% drop the first 5 messages
{#message_properties{expiry = 6}, VQ2} =
rabbit_variable_queue:dropwhile(
fun (#message_properties {expiry = Expiry}) -> Expiry =< 5 end, VQ1),
_VQ3 = rabbit_variable_queue:terminate(shutdown, VQ2),
Terms = variable_queue_read_terms(QName),
VQ4 = variable_queue_init(test_amqqueue(QName, true), Terms),
%% fetch 5
VQ5 = lists:foldl(fun (_, VQN) ->
{{_, _, _}, VQM} =
rabbit_variable_queue:fetch(false, VQN),
VQM
end, VQ4, lists:seq(6, Count)),
%% should be empty now
true = rabbit_variable_queue:is_empty(VQ5),
VQ5.
variable_queue_dropwhile_sync_restart(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, variable_queue_dropwhile_sync_restart1, [Config]).
variable_queue_dropwhile_sync_restart1(Config) ->
with_fresh_variable_queue(
fun variable_queue_dropwhile_sync_restart2/2,
?config(variable_queue_type, Config)).
variable_queue_dropwhile_sync_restart2(VQ0, QName) ->
Count = 10000,
%% add messages with sequential expiry
VQ1 = variable_queue_publish(
true, 1, Count,
fun (N, Props) -> Props#message_properties{expiry = N} end,
fun erlang:term_to_binary/1, VQ0),
%% drop the first 5 messages
{#message_properties{expiry = 6}, VQ2} =
rabbit_variable_queue:dropwhile(
fun (#message_properties {expiry = Expiry}) -> Expiry =< 5 end, VQ1),
%% Queue index sync.
VQ2b = rabbit_variable_queue:handle_pre_hibernate(VQ2),
_VQ3 = rabbit_variable_queue:terminate(shutdown, VQ2b),
Terms = variable_queue_read_terms(QName),
VQ4 = variable_queue_init(test_amqqueue(QName, true), Terms),
%% fetch 5
VQ5 = lists:foldl(fun (_, VQN) ->
{{_, _, _}, VQM} =
rabbit_variable_queue:fetch(false, VQN),
VQM
end, VQ4, lists:seq(6, Count)),
%% should be empty now
true = rabbit_variable_queue:is_empty(VQ5),
VQ5.
variable_queue_ack_limiting(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, variable_queue_ack_limiting1, [Config]).
variable_queue_ack_limiting1(Config) ->
with_fresh_variable_queue(
fun variable_queue_ack_limiting2/2,
?config(variable_queue_type, Config)).
variable_queue_ack_limiting2(VQ0, _Config) ->
%% start by sending in a bunch of messages
Len = 1024,
VQ1 = variable_queue_publish(false, Len, VQ0),
%% squeeze and relax queue
Churn = Len div 32,
VQ2 = publish_fetch_and_ack(Churn, Len, VQ1),
%% update stats
VQ3 = rabbit_variable_queue:update_rates(VQ2),
%% fetch half the messages
{VQ4, _AckTags} = variable_queue_fetch(Len div 2, false, false, Len, VQ3),
%% We only check the length anymore because
%% that's the only predictable stats we got.
VQ5 = check_variable_queue_status(VQ4, [{len, Len div 2}]),
VQ5.
variable_queue_purge(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, variable_queue_purge1, [Config]).
variable_queue_purge1(Config) ->
with_fresh_variable_queue(
fun variable_queue_purge2/2,
?config(variable_queue_type, Config)).
variable_queue_purge2(VQ0, _Config) ->
LenDepth = fun (VQ) ->
{rabbit_variable_queue:len(VQ),
rabbit_variable_queue:depth(VQ)}
end,
VQ1 = variable_queue_publish(false, 10, VQ0),
{VQ2, Acks} = variable_queue_fetch(6, false, false, 10, VQ1),
{4, VQ3} = rabbit_variable_queue:purge(VQ2),
{0, 6} = LenDepth(VQ3),
{_, VQ4} = rabbit_variable_queue:requeue(lists:sublist(Acks, 2), VQ3),
{2, 6} = LenDepth(VQ4),
VQ5 = rabbit_variable_queue:purge_acks(VQ4),
{2, 2} = LenDepth(VQ5),
VQ5.
variable_queue_requeue(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, variable_queue_requeue1, [Config]).
variable_queue_requeue1(Config) ->
with_fresh_variable_queue(
fun variable_queue_requeue2/2,
?config(variable_queue_type, Config)).
variable_queue_requeue2(VQ0, _Config) ->
{_PendingMsgs, RequeuedMsgs, FreshMsgs, VQ1} =
variable_queue_with_holes(VQ0),
Msgs =
lists:zip(RequeuedMsgs,
lists:duplicate(length(RequeuedMsgs), true)) ++
lists:zip(FreshMsgs,
lists:duplicate(length(FreshMsgs), false)),
VQ2 = lists:foldl(fun ({I, Requeued}, VQa) ->
{{M, MRequeued, _}, VQb} =
rabbit_variable_queue:fetch(true, VQa),
Requeued = MRequeued, %% assertion
I = msg2int(M), %% assertion
VQb
end, VQ1, Msgs),
{empty, VQ3} = rabbit_variable_queue:fetch(true, VQ2),
VQ3.
%% requeue from ram_pending_ack into q3, move to delta and then empty queue
variable_queue_requeue_ram_beta(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, variable_queue_requeue_ram_beta1, [Config]).
variable_queue_requeue_ram_beta1(Config) ->
with_fresh_variable_queue(
fun variable_queue_requeue_ram_beta2/2,
?config(variable_queue_type, Config)).
variable_queue_requeue_ram_beta2(VQ0, _Config) ->
2021-10-11 17:49:11 +08:00
IndexMod = index_mod(),
Count = IndexMod:next_segment_boundary(0)*2 + 2,
VQ1 = variable_queue_publish(false, Count, VQ0),
{VQ2, AcksR} = variable_queue_fetch(Count, false, false, Count, VQ1),
{Back, Front} = lists:split(Count div 2, AcksR),
{_, VQ3} = rabbit_variable_queue:requeue(erlang:tl(Back), VQ2),
{_, VQ5} = rabbit_variable_queue:requeue([erlang:hd(Back)], VQ3),
VQ6 = requeue_one_by_one(Front, VQ5),
{VQ7, AcksAll} = variable_queue_fetch(Count, false, true, Count, VQ6),
{_, VQ8} = rabbit_variable_queue:ack(AcksAll, VQ7),
VQ8.
variable_queue_fold(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, variable_queue_fold1, [Config]).
variable_queue_fold1(Config) ->
with_fresh_variable_queue(
fun variable_queue_fold2/2,
?config(variable_queue_type, Config)).
variable_queue_fold2(VQ0, _Config) ->
{PendingMsgs, RequeuedMsgs, FreshMsgs, VQ1} =
variable_queue_with_holes(VQ0),
Count = rabbit_variable_queue:depth(VQ1),
Msgs = lists:sort(PendingMsgs ++ RequeuedMsgs ++ FreshMsgs),
lists:foldl(fun (Cut, VQ2) ->
test_variable_queue_fold(Cut, Msgs, PendingMsgs, VQ2)
end, VQ1, [0, 1, 2, Count div 2,
Count - 1, Count, Count + 1, Count * 2]).
test_variable_queue_fold(Cut, Msgs, PendingMsgs, VQ0) ->
{Acc, VQ1} = rabbit_variable_queue:fold(
fun (M, _, Pending, A) ->
MInt = msg2int(M),
Pending = lists:member(MInt, PendingMsgs), %% assert
case MInt =< Cut of
true -> {cont, [MInt | A]};
false -> {stop, A}
end
end, [], VQ0),
Expected = lists:takewhile(fun (I) -> I =< Cut end, Msgs),
Expected = lists:reverse(Acc), %% assertion
VQ1.
%% same as test_variable_queue_requeue_ram_beta but randomly changing
%% the queue mode after every step.
variable_queue_mode_change(Config) ->
passed = rabbit_ct_broker_helpers:rpc(Config, 0,
?MODULE, variable_queue_mode_change1, [Config]).
variable_queue_mode_change1(Config) ->
with_fresh_variable_queue(
fun variable_queue_mode_change2/2,
?config(variable_queue_type, Config)).
variable_queue_mode_change2(VQ0, _Config) ->
2021-10-11 17:49:11 +08:00
IndexMod = index_mod(),
Count = IndexMod:next_segment_boundary(0)*2 + 2,
VQ1 = variable_queue_publish(false, Count, VQ0),
VQ2 = maybe_switch_queue_mode(VQ1),
{VQ3, AcksR} = variable_queue_fetch(Count, false, false, Count, VQ2),
VQ4 = maybe_switch_queue_mode(VQ3),
{Back, Front} = lists:split(Count div 2, AcksR),
{_, VQ5} = rabbit_variable_queue:requeue(erlang:tl(Back), VQ4),
VQ6 = maybe_switch_queue_mode(VQ5),
VQ8 = maybe_switch_queue_mode(VQ6),
{_, VQ9} = rabbit_variable_queue:requeue([erlang:hd(Back)], VQ8),
VQ10 = maybe_switch_queue_mode(VQ9),
VQ11 = requeue_one_by_one(Front, VQ10),
VQ12 = maybe_switch_queue_mode(VQ11),
{VQ13, AcksAll} = variable_queue_fetch(Count, false, true, Count, VQ12),
VQ14 = maybe_switch_queue_mode(VQ13),
{_, VQ15} = rabbit_variable_queue:ack(AcksAll, VQ14),
VQ16 = maybe_switch_queue_mode(VQ15),
VQ16.
maybe_switch_queue_mode(VQ) ->
Mode = random_queue_mode(),
set_queue_mode(Mode, VQ).
random_queue_mode() ->
Modes = [lazy, default],
2017-03-09 23:05:37 +08:00
lists:nth(rand:uniform(length(Modes)), Modes).
pub_res({_, VQS}) ->
VQS;
pub_res(VQS) ->
VQS.
make_publish(IsPersistent, PayloadFun, PropFun, N) ->
Message Containers (#5077) This PR implements an approach for a "protocol (data format) agnostic core" where the format of the message isn't converted at point of reception. Currently all non AMQP 0.9.1 originating messages are converted into a AMQP 0.9.1 flavoured basic_message record before sent to a queue. If the messages are then consumed by the originating protocol they are converted back from AMQP 0.9.1. For some protocols such as MQTT 3.1 this isn't too expensive as MQTT is mostly a fairly easily mapped subset of AMQP 0.9.1 but for others such as AMQP 1.0 the conversions are awkward and in some cases lossy even if consuming from the originating protocol. This PR instead wraps all incoming messages in their originating form into a generic, extensible message container type (mc). The container module exposes an API to get common message details such as size and various properties (ttl, priority etc) directly from the source data type. Each protocol needs to implement the mc behaviour such that when a message originating form one protocol is consumed by another protocol we convert it to the target protocol at that point. The message container also contains annotations, dead letter records and other meta data we need to record during the lifetime of a message. The original protocol message is never modified unless it is consumed. This includes conversion modules to and from amqp, amqpl (AMQP 0.9.1) and mqtt. COMMIT HISTORY: * Refactor away from using the delivery{} record In many places including exchange types. This should make it easier to move towards using a message container type instead of basic_message. Add mc module and move direct replies outside of exchange Lots of changes incl classic queues Implement stream support incl amqp conversions simplify mc state record move mc.erl mc dlx stuff recent history exchange Make tracking work But doesn't take a protocol agnostic approach as we just convert everything into AMQP legacy and back. Might be good enough for now. Tracing as a whole may want a bit of a re-vamp at some point. tidy make quorum queue peek work by legacy conversion dead lettering fixes dead lettering fixes CMQ fixes rabbit_trace type fixes fixes fix Fix classic queue props test assertion fix feature flag and backwards compat Enable message_container feature flag in some SUITEs Dialyzer fixes fixes fix test fixes Various Manually update a gazelle generated file until a gazelle enhancement can be made https://github.com/rabbitmq/rules_erlang/issues/185 Add message_containers_SUITE to bazel and regen bazel files with gazelle from rules_erlang@main Simplify essential proprty access Such as durable, ttl and priority by extracting them into annotations at message container init time. Move type to remove dependenc on amqp10 stuff in mc.erl mostly because I don't know how to make bazel do the right thing add more stuff Refine routing header stuff wip Cosmetics Do not use "maybe" as type name as "maybe" is a keyword since OTP 25 which makes Erlang LS complain. * Dedup death queue names * Fix function clause crashes Fix failing tests in the MQTT shared_SUITE: A classic queue message ID can be undefined as set in https://github.com/rabbitmq/rabbitmq-server/blob/fbe79ff47b4edbc0fd95457e623d6593161ad198/deps/rabbit/src/rabbit_classic_queue_index_v2.erl#L1048 Fix failing tests in the MQTT shared_SUITE-mixed: When feature flag message_containers is disabled, the message is not an #mc{} record, but a #basic_message{} record. * Fix is_utf8_no_null crash Prior to this commit, the function crashed if invalid UTF-8 was provided, e.g.: ``` 1> rabbit_misc:is_valid_shortstr(<<"😇"/utf16>>). ** exception error: no function clause matching rabbit_misc:is_utf8_no_null(<<216,61,222,7>>) (rabbit_misc.erl, line 1481) ``` * Implement mqtt mc behaviour For now via amqp translation. This is still work in progress, but the following SUITEs pass: ``` make -C deps/rabbitmq_mqtt ct-shared t=[mqtt,v5,cluster_size_1] FULL=1 make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_1] FULL=1 ``` * Shorten mc file names Module name length matters because for each persistent message the #mc{} record is persisted to disk. ``` 1> iolist_size(term_to_iovec({mc, rabbit_mc_amqp_legacy})). 30 2> iolist_size(term_to_iovec({mc, mc_amqpl})). 17 ``` This commit renames the mc modules: ``` ag -l rabbit_mc_amqp_legacy | xargs sed -i 's/rabbit_mc_amqp_legacy/mc_amqpl/g' ag -l rabbit_mc_amqp | xargs sed -i 's/rabbit_mc_amqp/mc_amqp/g' ag -l rabbit_mqtt_mc | xargs sed -i 's/rabbit_mqtt_mc/mc_mqtt/g' ``` * mc: make deaths an annotation + fixes * Fix mc_mqtt protocol_state callback * Fix test will_delay_node_restart ``` make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_3]:will_delay_node_restart FULL=1 ``` * Bazel run gazelle * mix format rabbitmqctl.ex * Ensure ttl annotation is refelected in amqp legacy protocol state * Fix id access in message store * Fix rabbit_message_interceptor_SUITE * dializer fixes * Fix rabbit:rabbit_message_interceptor_SUITE-mixed set_annotation/3 should not result in duplicate keys * Fix MQTT shared_SUITE-mixed Up to 3.12 non-MQTT publishes were always QoS 1 regardless of delivery_mode. https://github.com/rabbitmq/rabbitmq-server/blob/75a953ce286a10aca910c098805a4f545989af38/deps/rabbitmq_mqtt/src/rabbit_mqtt_processor.erl#L2075-L2076 From now on, non-MQTT publishes are QoS 1 if durable. This makes more sense. The MQTT plugin must send a #basic_message{} to an old node that does not understand message containers. * Field content of 'v1_0.data' can be binary Fix ``` bazel test //deps/rabbitmq_mqtt:shared_SUITE-mixed \ --test_env FOCUS="-group [mqtt,v4,cluster_size_1] -case trace" \ -t- --test_sharding_strategy=disabled ``` * Remove route/2 and implement route/3 for all exchange types. This removes the route/2 callback from rabbit_exchange_type and makes route/3 mandatory instead. This is a breaking change and will require all implementations of exchange types to update their code, however this is necessary anyway for them to correctly handle the mc type. stream filtering fixes * Translate directly from MQTT to AMQP 0.9.1 * handle undecoded properties in mc_compat amqpl: put clause in right order recover death deatails from amqp data * Replace callback init_amqp with convert_from * Fix return value of lists:keyfind/3 * Translate directly from AMQP 0.9.1 to MQTT * Fix MQTT payload size MQTT payload can be a list when converted from AMQP 0.9.1 for example First conversions tests Plus some other conversion related fixes. bazel bazel translate amqp 1.0 null to undefined mc: property/2 and correlation_id/message_id return type tagged values. To ensure we can support a variety of types better. The type type tags are AMQP 1.0 flavoured. fix death recovery mc_mqtt: impl new api Add callbacks to allow protocols to compact data before storage And make readable if needing to query things repeatedly. bazel fix * more decoding * tracking mixed versions compat * mc: flip default of `durable` annotation to save some data. Assuming most messages are durable and that in memory messages suffer less from persistence overhead it makes sense for a non existent `durable` annotation to mean durable=true. * mc conversion tests and tidy up * mc make x_header unstrict again * amqpl: death record fixes * bazel * amqp -> amqpl conversion test * Fix crash in mc_amqp:size/1 Body can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0/ ct-system t=java ``` on branch native-amqp. * Fix crash in lists:flatten/1 Data can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp. * Fix crash in rabbit_writer Running test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp resulted in the following crash: ``` crasher: initial call: rabbit_writer:enter_mainloop/2 pid: <0.711.0> registered_name: [] exception error: bad argument in function size/1 called as size([<<0>>,<<"Sw">>,[<<160,2>>,<<"hi">>]]) *** argument 1: not tuple or binary in call from rabbit_binary_generator:build_content_frames/7 (rabbit_binary_generator.erl, line 89) in call from rabbit_binary_generator:build_simple_content_frames/4 (rabbit_binary_generator.erl, line 61) in call from rabbit_writer:assemble_frames/5 (rabbit_writer.erl, line 334) in call from rabbit_writer:internal_send_command_async/3 (rabbit_writer.erl, line 365) in call from rabbit_writer:handle_message/2 (rabbit_writer.erl, line 265) in call from rabbit_writer:handle_message/3 (rabbit_writer.erl, line 232) in call from rabbit_writer:mainloop1/2 (rabbit_writer.erl, line 223) ``` because #content.payload_fragments_rev is currently supposed to be a flat list of binaries instead of being an iolist. This commit fixes this crash inefficiently by calling iolist_to_binary/1. A better solution would be to allow AMQP legacy's #content.payload_fragments_rev to be an iolist. * Add accidentally deleted line back * mc: optimise mc_amqp internal format By removint the outer records for message and delivery annotations as well as application properties and footers. * mc: optimis mc_amqp map_add by using upsert * mc: refactoring and bug fixes * mc_SUITE routingheader assertions * mc remove serialize/1 callback as only used by amqp * mc_amqp: avoid returning a nested list from protocol_state * test and bug fix * move infer_type to mc_util * mc fixes and additiona assertions * Support headers exchange routing for MQTT messages When a headers exchange is bound to the MQTT topic exchange, routing will be performend based on both MQTT topic (by the topic exchange) and MQTT User Property (by the headers exchange). This combines the best worlds of both MQTT 5.0 and AMQP 0.9.1 and enables powerful routing topologies. When the User Property contains the same name multiple times, only the last name (and value) will be considered by the headers exchange. * Fix crash when sending from stream to amqpl When publishing a message via the stream protocol and consuming it via AMQP 0.9.1, the following crash occurred prior to this commit: ``` crasher: initial call: rabbit_channel:init/1 pid: <0.818.0> registered_name: [] exception exit: {{badmatch,undefined}, [{rabbit_channel,handle_deliver0,4, [{file,"rabbit_channel.erl"}, {line,2728}]}, {lists,foldl,3,[{file,"lists.erl"},{line,1594}]}, {rabbit_channel,handle_cast,2, [{file,"rabbit_channel.erl"}, {line,728}]}, {gen_server2,handle_msg,2, [{file,"gen_server2.erl"},{line,1056}]}, {proc_lib,wake_up,3, [{file,"proc_lib.erl"},{line,251}]}]} ``` This commit first gives `mc:init/3` the chance to set exchange and routing_keys annotations. If not set, `rabbit_stream_queue` will set these annotations assuming the message was originally published via the stream protocol. * Support consistent hash exchange routing for MQTT 5.0 When a consistent hash exchange is bound to the MQTT topic exchange, MQTT 5.0 messages can be routed to queues consistently based on the Correlation-Data in the PUBLISH packet. * Convert MQTT 5.0 User Property * to AMQP 0.9.1 headers * from AMQP 0.9.1 headers * to AMQP 1.0 application properties and message annotations * from AMQP 1.0 application properties and message annotations * Make use of Annotations in mc_mqtt:protocol_state/2 mc_mqtt:protocol_state/2 includes Annotations as parameter. It's cleaner to make use of these Annotations when computing the protocol state instead of relying on the caller (rabbitmq_mqtt_processor) to compute the protocol state. * Enforce AMQP 0.9.1 field name length limit The AMQP 0.9.1 spec prohibits field names longer than 128 characters. Therefore, when converting AMQP 1.0 message annotations, application properties or MQTT 5.0 User Property to AMQP 0.9.1 headers, drop any names longer than 128 characters. * Fix type specs Apply feedback from Michael Davis Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * Add mc_mqtt unit test suite Implement mc_mqtt:x_header/2 * Translate indicator that payload is UTF-8 encoded when converting between MQTT 5.0 and AMQP 1.0 * Translate single amqp-value section from AMQP 1.0 to MQTT Convert to a text representation, if possible, and indicate to MQTT client that the payload is UTF-8 encoded. This way, the MQTT client will be able to parse the payload. If conversion to text representation is not possible, encode the payload using the AMQP 1.0 type system and indiate the encoding via Content-Type message/vnd.rabbitmq.amqp. This Content-Type is not registered. Type "message" makes sense since it's a message. Vendor tree "vnd.rabbitmq.amqp" makes sense since merely subtype "amqp" is not registered. * Fix payload conversion * Translate Response Topic between MQTT and AMQP Translate MQTT 5.0 Response Topic to AMQP 1.0 reply-to address and vice versa. The Response Topic must be a UTF-8 encoded string. This commit re-uses the already defined RabbitMQ target addresses: ``` "/topic/" RK Publish to amq.topic with routing key RK "/exchange/" X "/" RK Publish to exchange X with routing key RK ``` By default, the MQTT topic exchange is configure dto be amq.topic using the 1st target address. When an operator modifies the mqtt.exchange, the 2nd target address is used. * Apply PR feedback and fix formatting Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * tidy up * Add MQTT message_containers test * consistent hash exchange: avoid amqp legacy conversion When hashing on a header value. * Avoid converting to amqp legacy when using exchange federation * Fix test flake * test and dialyzer fixes * dialyzer fix * Add MQTT protocol interoperability tests Test receiving from and sending to MQTT 5.0 and * AMQP 0.9.1 * AMQP 1.0 * STOMP * Streams * Regenerate portions of deps/rabbit/app.bzl with gazelle I'm not exactly sure how this happened, but gazell seems to have been run with an older version of the rules_erlang gazelle extension at some point. This caused generation of a structure that is no longer used. This commit updates the structure to the current pattern. * mc: refactoring * mc_amqpl: handle delivery annotations Just in case they are included. Also use iolist_to_iovec to create flat list of binaries when converting from amqp with amqp encoded payload. --------- Co-authored-by: David Ansari <david.ansari@gmx.de> Co-authored-by: Michael Davis <mcarsondavis@gmail.com> Co-authored-by: Rin Kuryloski <kuryloskip@vmware.com>
2023-08-31 18:27:13 +08:00
{message(IsPersistent, PayloadFun, N),
PropFun(N, #message_properties{size = 10}),
false}.
make_publish_delivered(IsPersistent, PayloadFun, PropFun, N) ->
Message Containers (#5077) This PR implements an approach for a "protocol (data format) agnostic core" where the format of the message isn't converted at point of reception. Currently all non AMQP 0.9.1 originating messages are converted into a AMQP 0.9.1 flavoured basic_message record before sent to a queue. If the messages are then consumed by the originating protocol they are converted back from AMQP 0.9.1. For some protocols such as MQTT 3.1 this isn't too expensive as MQTT is mostly a fairly easily mapped subset of AMQP 0.9.1 but for others such as AMQP 1.0 the conversions are awkward and in some cases lossy even if consuming from the originating protocol. This PR instead wraps all incoming messages in their originating form into a generic, extensible message container type (mc). The container module exposes an API to get common message details such as size and various properties (ttl, priority etc) directly from the source data type. Each protocol needs to implement the mc behaviour such that when a message originating form one protocol is consumed by another protocol we convert it to the target protocol at that point. The message container also contains annotations, dead letter records and other meta data we need to record during the lifetime of a message. The original protocol message is never modified unless it is consumed. This includes conversion modules to and from amqp, amqpl (AMQP 0.9.1) and mqtt. COMMIT HISTORY: * Refactor away from using the delivery{} record In many places including exchange types. This should make it easier to move towards using a message container type instead of basic_message. Add mc module and move direct replies outside of exchange Lots of changes incl classic queues Implement stream support incl amqp conversions simplify mc state record move mc.erl mc dlx stuff recent history exchange Make tracking work But doesn't take a protocol agnostic approach as we just convert everything into AMQP legacy and back. Might be good enough for now. Tracing as a whole may want a bit of a re-vamp at some point. tidy make quorum queue peek work by legacy conversion dead lettering fixes dead lettering fixes CMQ fixes rabbit_trace type fixes fixes fix Fix classic queue props test assertion fix feature flag and backwards compat Enable message_container feature flag in some SUITEs Dialyzer fixes fixes fix test fixes Various Manually update a gazelle generated file until a gazelle enhancement can be made https://github.com/rabbitmq/rules_erlang/issues/185 Add message_containers_SUITE to bazel and regen bazel files with gazelle from rules_erlang@main Simplify essential proprty access Such as durable, ttl and priority by extracting them into annotations at message container init time. Move type to remove dependenc on amqp10 stuff in mc.erl mostly because I don't know how to make bazel do the right thing add more stuff Refine routing header stuff wip Cosmetics Do not use "maybe" as type name as "maybe" is a keyword since OTP 25 which makes Erlang LS complain. * Dedup death queue names * Fix function clause crashes Fix failing tests in the MQTT shared_SUITE: A classic queue message ID can be undefined as set in https://github.com/rabbitmq/rabbitmq-server/blob/fbe79ff47b4edbc0fd95457e623d6593161ad198/deps/rabbit/src/rabbit_classic_queue_index_v2.erl#L1048 Fix failing tests in the MQTT shared_SUITE-mixed: When feature flag message_containers is disabled, the message is not an #mc{} record, but a #basic_message{} record. * Fix is_utf8_no_null crash Prior to this commit, the function crashed if invalid UTF-8 was provided, e.g.: ``` 1> rabbit_misc:is_valid_shortstr(<<"😇"/utf16>>). ** exception error: no function clause matching rabbit_misc:is_utf8_no_null(<<216,61,222,7>>) (rabbit_misc.erl, line 1481) ``` * Implement mqtt mc behaviour For now via amqp translation. This is still work in progress, but the following SUITEs pass: ``` make -C deps/rabbitmq_mqtt ct-shared t=[mqtt,v5,cluster_size_1] FULL=1 make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_1] FULL=1 ``` * Shorten mc file names Module name length matters because for each persistent message the #mc{} record is persisted to disk. ``` 1> iolist_size(term_to_iovec({mc, rabbit_mc_amqp_legacy})). 30 2> iolist_size(term_to_iovec({mc, mc_amqpl})). 17 ``` This commit renames the mc modules: ``` ag -l rabbit_mc_amqp_legacy | xargs sed -i 's/rabbit_mc_amqp_legacy/mc_amqpl/g' ag -l rabbit_mc_amqp | xargs sed -i 's/rabbit_mc_amqp/mc_amqp/g' ag -l rabbit_mqtt_mc | xargs sed -i 's/rabbit_mqtt_mc/mc_mqtt/g' ``` * mc: make deaths an annotation + fixes * Fix mc_mqtt protocol_state callback * Fix test will_delay_node_restart ``` make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_3]:will_delay_node_restart FULL=1 ``` * Bazel run gazelle * mix format rabbitmqctl.ex * Ensure ttl annotation is refelected in amqp legacy protocol state * Fix id access in message store * Fix rabbit_message_interceptor_SUITE * dializer fixes * Fix rabbit:rabbit_message_interceptor_SUITE-mixed set_annotation/3 should not result in duplicate keys * Fix MQTT shared_SUITE-mixed Up to 3.12 non-MQTT publishes were always QoS 1 regardless of delivery_mode. https://github.com/rabbitmq/rabbitmq-server/blob/75a953ce286a10aca910c098805a4f545989af38/deps/rabbitmq_mqtt/src/rabbit_mqtt_processor.erl#L2075-L2076 From now on, non-MQTT publishes are QoS 1 if durable. This makes more sense. The MQTT plugin must send a #basic_message{} to an old node that does not understand message containers. * Field content of 'v1_0.data' can be binary Fix ``` bazel test //deps/rabbitmq_mqtt:shared_SUITE-mixed \ --test_env FOCUS="-group [mqtt,v4,cluster_size_1] -case trace" \ -t- --test_sharding_strategy=disabled ``` * Remove route/2 and implement route/3 for all exchange types. This removes the route/2 callback from rabbit_exchange_type and makes route/3 mandatory instead. This is a breaking change and will require all implementations of exchange types to update their code, however this is necessary anyway for them to correctly handle the mc type. stream filtering fixes * Translate directly from MQTT to AMQP 0.9.1 * handle undecoded properties in mc_compat amqpl: put clause in right order recover death deatails from amqp data * Replace callback init_amqp with convert_from * Fix return value of lists:keyfind/3 * Translate directly from AMQP 0.9.1 to MQTT * Fix MQTT payload size MQTT payload can be a list when converted from AMQP 0.9.1 for example First conversions tests Plus some other conversion related fixes. bazel bazel translate amqp 1.0 null to undefined mc: property/2 and correlation_id/message_id return type tagged values. To ensure we can support a variety of types better. The type type tags are AMQP 1.0 flavoured. fix death recovery mc_mqtt: impl new api Add callbacks to allow protocols to compact data before storage And make readable if needing to query things repeatedly. bazel fix * more decoding * tracking mixed versions compat * mc: flip default of `durable` annotation to save some data. Assuming most messages are durable and that in memory messages suffer less from persistence overhead it makes sense for a non existent `durable` annotation to mean durable=true. * mc conversion tests and tidy up * mc make x_header unstrict again * amqpl: death record fixes * bazel * amqp -> amqpl conversion test * Fix crash in mc_amqp:size/1 Body can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0/ ct-system t=java ``` on branch native-amqp. * Fix crash in lists:flatten/1 Data can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp. * Fix crash in rabbit_writer Running test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp resulted in the following crash: ``` crasher: initial call: rabbit_writer:enter_mainloop/2 pid: <0.711.0> registered_name: [] exception error: bad argument in function size/1 called as size([<<0>>,<<"Sw">>,[<<160,2>>,<<"hi">>]]) *** argument 1: not tuple or binary in call from rabbit_binary_generator:build_content_frames/7 (rabbit_binary_generator.erl, line 89) in call from rabbit_binary_generator:build_simple_content_frames/4 (rabbit_binary_generator.erl, line 61) in call from rabbit_writer:assemble_frames/5 (rabbit_writer.erl, line 334) in call from rabbit_writer:internal_send_command_async/3 (rabbit_writer.erl, line 365) in call from rabbit_writer:handle_message/2 (rabbit_writer.erl, line 265) in call from rabbit_writer:handle_message/3 (rabbit_writer.erl, line 232) in call from rabbit_writer:mainloop1/2 (rabbit_writer.erl, line 223) ``` because #content.payload_fragments_rev is currently supposed to be a flat list of binaries instead of being an iolist. This commit fixes this crash inefficiently by calling iolist_to_binary/1. A better solution would be to allow AMQP legacy's #content.payload_fragments_rev to be an iolist. * Add accidentally deleted line back * mc: optimise mc_amqp internal format By removint the outer records for message and delivery annotations as well as application properties and footers. * mc: optimis mc_amqp map_add by using upsert * mc: refactoring and bug fixes * mc_SUITE routingheader assertions * mc remove serialize/1 callback as only used by amqp * mc_amqp: avoid returning a nested list from protocol_state * test and bug fix * move infer_type to mc_util * mc fixes and additiona assertions * Support headers exchange routing for MQTT messages When a headers exchange is bound to the MQTT topic exchange, routing will be performend based on both MQTT topic (by the topic exchange) and MQTT User Property (by the headers exchange). This combines the best worlds of both MQTT 5.0 and AMQP 0.9.1 and enables powerful routing topologies. When the User Property contains the same name multiple times, only the last name (and value) will be considered by the headers exchange. * Fix crash when sending from stream to amqpl When publishing a message via the stream protocol and consuming it via AMQP 0.9.1, the following crash occurred prior to this commit: ``` crasher: initial call: rabbit_channel:init/1 pid: <0.818.0> registered_name: [] exception exit: {{badmatch,undefined}, [{rabbit_channel,handle_deliver0,4, [{file,"rabbit_channel.erl"}, {line,2728}]}, {lists,foldl,3,[{file,"lists.erl"},{line,1594}]}, {rabbit_channel,handle_cast,2, [{file,"rabbit_channel.erl"}, {line,728}]}, {gen_server2,handle_msg,2, [{file,"gen_server2.erl"},{line,1056}]}, {proc_lib,wake_up,3, [{file,"proc_lib.erl"},{line,251}]}]} ``` This commit first gives `mc:init/3` the chance to set exchange and routing_keys annotations. If not set, `rabbit_stream_queue` will set these annotations assuming the message was originally published via the stream protocol. * Support consistent hash exchange routing for MQTT 5.0 When a consistent hash exchange is bound to the MQTT topic exchange, MQTT 5.0 messages can be routed to queues consistently based on the Correlation-Data in the PUBLISH packet. * Convert MQTT 5.0 User Property * to AMQP 0.9.1 headers * from AMQP 0.9.1 headers * to AMQP 1.0 application properties and message annotations * from AMQP 1.0 application properties and message annotations * Make use of Annotations in mc_mqtt:protocol_state/2 mc_mqtt:protocol_state/2 includes Annotations as parameter. It's cleaner to make use of these Annotations when computing the protocol state instead of relying on the caller (rabbitmq_mqtt_processor) to compute the protocol state. * Enforce AMQP 0.9.1 field name length limit The AMQP 0.9.1 spec prohibits field names longer than 128 characters. Therefore, when converting AMQP 1.0 message annotations, application properties or MQTT 5.0 User Property to AMQP 0.9.1 headers, drop any names longer than 128 characters. * Fix type specs Apply feedback from Michael Davis Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * Add mc_mqtt unit test suite Implement mc_mqtt:x_header/2 * Translate indicator that payload is UTF-8 encoded when converting between MQTT 5.0 and AMQP 1.0 * Translate single amqp-value section from AMQP 1.0 to MQTT Convert to a text representation, if possible, and indicate to MQTT client that the payload is UTF-8 encoded. This way, the MQTT client will be able to parse the payload. If conversion to text representation is not possible, encode the payload using the AMQP 1.0 type system and indiate the encoding via Content-Type message/vnd.rabbitmq.amqp. This Content-Type is not registered. Type "message" makes sense since it's a message. Vendor tree "vnd.rabbitmq.amqp" makes sense since merely subtype "amqp" is not registered. * Fix payload conversion * Translate Response Topic between MQTT and AMQP Translate MQTT 5.0 Response Topic to AMQP 1.0 reply-to address and vice versa. The Response Topic must be a UTF-8 encoded string. This commit re-uses the already defined RabbitMQ target addresses: ``` "/topic/" RK Publish to amq.topic with routing key RK "/exchange/" X "/" RK Publish to exchange X with routing key RK ``` By default, the MQTT topic exchange is configure dto be amq.topic using the 1st target address. When an operator modifies the mqtt.exchange, the 2nd target address is used. * Apply PR feedback and fix formatting Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * tidy up * Add MQTT message_containers test * consistent hash exchange: avoid amqp legacy conversion When hashing on a header value. * Avoid converting to amqp legacy when using exchange federation * Fix test flake * test and dialyzer fixes * dialyzer fix * Add MQTT protocol interoperability tests Test receiving from and sending to MQTT 5.0 and * AMQP 0.9.1 * AMQP 1.0 * STOMP * Streams * Regenerate portions of deps/rabbit/app.bzl with gazelle I'm not exactly sure how this happened, but gazell seems to have been run with an older version of the rules_erlang gazelle extension at some point. This caused generation of a structure that is no longer used. This commit updates the structure to the current pattern. * mc: refactoring * mc_amqpl: handle delivery annotations Just in case they are included. Also use iolist_to_iovec to create flat list of binaries when converting from amqp with amqp encoded payload. --------- Co-authored-by: David Ansari <david.ansari@gmx.de> Co-authored-by: Michael Davis <mcarsondavis@gmail.com> Co-authored-by: Rin Kuryloski <kuryloskip@vmware.com>
2023-08-31 18:27:13 +08:00
{message(IsPersistent, PayloadFun, N),
PropFun(N, #message_properties{size = 10})}.
queue_name(Config, Name) ->
Name1 = iolist_to_binary(rabbit_ct_helpers:config_to_testcase_name(Config, Name)),
queue_name(Name1).
queue_name(Name) ->
rabbit_misc:r(<<"/">>, queue, Name).
test_queue() ->
queue_name(rabbit_guid:gen()).
init_test_queue(QName) ->
PRef = rabbit_guid:gen(),
PersistentClient = msg_store_client_init(?PERSISTENT_MSG_STORE, PRef),
2021-10-11 17:49:11 +08:00
IndexMod = index_mod(),
Res = IndexMod:recover(
QName, [], false,
fun (MsgId) ->
rabbit_msg_store:contains(MsgId, PersistentClient)
end,
fun nop/1, fun nop/1,
main),
ok = rabbit_msg_store:client_delete_and_terminate(PersistentClient),
Res.
restart_test_queue(Qi, QName) ->
2021-10-11 17:49:11 +08:00
IndexMod = index_mod(),
_ = IndexMod:terminate(?VHOST, [], Qi),
ok = rabbit_variable_queue:stop(?VHOST),
{ok, _} = rabbit_variable_queue:start(?VHOST, [QName]),
init_test_queue(QName).
empty_test_queue(QName) ->
ok = rabbit_variable_queue:stop(?VHOST),
{ok, _} = rabbit_variable_queue:start(?VHOST, []),
{0, 0, Qi} = init_test_queue(QName),
2021-10-11 17:49:11 +08:00
IndexMod = index_mod(),
_ = IndexMod:delete_and_terminate(Qi),
ok.
unin_empty_test_queue(QName) ->
{0, 0, Qi} = init_test_queue(QName),
2021-10-11 17:49:11 +08:00
IndexMod = index_mod(),
_ = IndexMod:delete_and_terminate(Qi),
ok.
with_empty_test_queue(Fun) ->
QName = test_queue(),
ok = empty_test_queue(QName),
{0, 0, Qi} = init_test_queue(QName),
2021-10-11 17:49:11 +08:00
IndexMod = index_mod(),
IndexMod:delete_and_terminate(Fun(Qi, QName)).
init_queue_index() ->
%% We must set the segment entry count in the process dictionary
%% for tests that call the v1 queue index directly to have a correct
%% value.
put(segment_entry_count, 2048),
ok.
restart_app() ->
rabbit:stop(),
rabbit:start().
queue_index_publish(SeqIds, Persistent, Qi) ->
2021-10-11 17:49:11 +08:00
IndexMod = index_mod(),
Ref = rabbit_guid:gen(),
MsgStore = case Persistent of
true -> ?PERSISTENT_MSG_STORE;
false -> ?TRANSIENT_MSG_STORE
end,
MSCState = msg_store_client_init(MsgStore, Ref),
{A, B = [{_SeqId, LastMsgIdWritten} | _]} =
lists:foldl(
fun (SeqId, {QiN, SeqIdsMsgIdsAcc}) ->
MsgId = rabbit_guid:gen(),
2021-10-11 17:49:11 +08:00
QiM = IndexMod:publish(
2021-08-30 20:10:47 +08:00
MsgId, SeqId, rabbit_msg_store,
#message_properties{size = 10},
Persistent, infinity, QiN),
ok = rabbit_msg_store:write(SeqId, MsgId, MsgId, MSCState),
{QiM, [{SeqId, MsgId} | SeqIdsMsgIdsAcc]}
end, {Qi, []}, SeqIds),
%% do this just to force all of the publishes through to the msg_store:
true = rabbit_msg_store:contains(LastMsgIdWritten, MSCState),
ok = rabbit_msg_store:client_delete_and_terminate(MSCState),
{A, B}.
nop(_) -> ok.
nop(_, _) -> ok.
msg_store_client_init(MsgStore, Ref) ->
rabbit_vhost_msg_store:client_init(?VHOST, MsgStore, Ref, undefined).
variable_queue_init(Q, Recover) ->
rabbit_variable_queue:init(
Q, case Recover of
true -> non_clean_shutdown;
false -> new;
Terms -> Terms
end, fun nop/2, fun nop/1, fun nop/1).
variable_queue_read_terms(QName) ->
#resource { kind = queue,
virtual_host = VHost,
name = Name } = QName,
<<Num:128>> = erlang:md5(<<"queue", VHost/binary, Name/binary>>),
DirName = rabbit_misc:format("~.36B", [Num]),
{ok, Terms} = rabbit_recovery_terms:read(VHost, DirName),
Terms.
publish_and_confirm(Q, Payload, Count) ->
Seqs = lists:seq(1, Count),
Stream Queue This is an aggregated commit of all changes related to the initial implementation of queue types and on top of that the stream queue type. The varios commit messages have simply been included mostly un-edited below. Make rabbit_amqqueue:not_found_or_absent_dirty/1 visible For use in the stream plugin. Use bigger retention policy on max-age test Set coordinator timeout to 30s Handle coordinator unavailable error Handle operator policies as maps when checking if is applicable Add is_policy_applicable/2 to classic queues Ignore restart commands if the stream has been deleted It could happen that after termination some of the monitors are still up and trigger writer/replica restarts Policy support on stream queues Remove subscription events on stream coordinator Ensure old leaders are removed from monitors Introduce delay when retrying a failed phase Note that this ensures monitor is setup, there was a bug where no monitor was really started when re-trying the same phase Restart replicas after leader election instead of relying on old monitors Use timer for stream coordinator retries Fix stream stats for members/online Multiple fixes for replica monitoring and restart Ensure pending commands are appended at the end and re-run Ensure phase is reset with the state Remove duplicates from replica list Restart current phase on state_enter Remove unused import Ensure rabbit is running when checking for stream quorum Restart replicas Add a close/1 function to queue types So that we can get a chance of cleaning up resources if needed. Stream queues close their osiris logs at this point. fix compiler errors stream-queue: take retention into account When calculating ready messages metrics. Add osiris to the list of rabbit deps Retry restart of replicas Do not restart replicas or leaders after receiving a delete cluster command Add more logging to the stream coordinator Monitor subscribed processes on the stream coordinator Memory breakdown for stream queues Update quorum queue event formatter rabbit_msg_record fixes Refactor channel confirms Remove old unconfirmed_messages module that was designed to handle multiple queue fan in logic including all ha mirrors etc. Replaced with simpler rabbit_confirms module that handles the fan out and leaves any queue specific logic (such as confirms from mirrors) to the queue type implemention. Also this module has a dedicated test module. Which is nice. Backward compatibility with 3.8.x events Supports mixed version cluster upgrades Match specification when stream queue already exists Max age retention for stream queues Stop all replicas before starting leader election stream: disallow global qos remove IS_CLASSIC|QUORUM macros Ensure only classic queues are notified on channel down This also removes the delivering_queues map in the channel state as it should not be needed for this and just cause additional unecessary accounting. Polish AMQP 1.0/0.9.1 properties conversion Support byte in application properties, handle 1-bit representation for booleans. Use binary in header for long AMQP 1.0 ID Fix AMQP 1.0 to 0.9.1 conversion Fix test due to incorrect type Convert timestamp application properties to/from seconds AMQP 1.0 uses milliseconds for timestamp and AMQP 0.9.1 uses seconds, so conversion needed. Dialyzer fixes Handle all message-id types AMQP 1.0 is more liberal in it's allowed types of message-id and correlation-id - this adds headers to describe the type of the data in the message_id / correlation_id properties and also handles the case where the data cannot fit by again using headers. Resize stream coordinator cluster when broker configuration changes convert timestamp to and fro seconds user_id should be a binary message annotations keys need to be symbols stream-queue: default exchange and routing key As these won't be present for data written using the rabbitmq-stream plugin. Add exchange, routing key as message annotations To the AMQP 1.0 formatted data to enable roundtrip. Add osiris logging module config And update logging config test suite. Restart election when start of new leader fails The node might have just gone down so we need to try another one Only aux keeps track of phase now, as it might change if the leader election fails Stream coordinator refactor - all state is kept on the ra machine Ensure any ra cluster not a qq is not cleaned up Fixes to recovery and monitoring Add AMQP 1.0 common to dependencies Add rabbit_msg_record module To handle conversions into internal stream storage format. Use rabbitmq-common stream-queue branch Use SSH for osiris dependency Stream coordinator: delete replica Stream coordinator: add replica Stream coordinator: leader failover Stream coordinator: declare and delete Test consuming from a random offset Previous offsets should not be delivered to consumers Consume from stream replicas and multiple test fixes Use max-length-bytes and add new max-segment-size Use SSH for osiris dependency Basic cancel for stream queues Publish stream queues and settle/reject/requeue refactor Consume from stream queues Fix recovery Publish stream messages Add/delete stream replicas Use safe queue names Set retention policy for stream queues Required by the ctl command [#171207092] Stream queue delete queue fix missing callback impl Stream queue declare Queue type abstraction And use the implementing module as the value of the amqqueue record `type` field. This will allow for easy dispatch to the queue type implementation. Queue type abstraction Move queue declare into rabbit_queue_type Move queue delete into queue type implementation Queue type: dequeue/basic_get Move info inside queue type abstraction Move policy change into queue type interface Add purge to queue type Add recovery to the queue type interface Rename amqqueue quorum_nodes field To a more generic an extensible opaque queue type specific map. Fix tests and handle classic API response Fix HA queue confirm bug All mirrors need to be present as queue names. This introduces context linking allowing additional queue refs to be linked to a single "master" queue ref contining the actual queue context. Fix issue with events of deleted queues Also update queue type smoke test to use a cluster by default. correct default value of amqqueue getter Move classic queues further inside queue type interface why [TrackerId] Dialyzer fixes
2020-09-29 18:43:24 +08:00
QTState0 = rabbit_queue_type:new(Q, rabbit_queue_type:init()),
QTState =
lists:foldl(
fun (Seq, Acc0) ->
Message Containers (#5077) This PR implements an approach for a "protocol (data format) agnostic core" where the format of the message isn't converted at point of reception. Currently all non AMQP 0.9.1 originating messages are converted into a AMQP 0.9.1 flavoured basic_message record before sent to a queue. If the messages are then consumed by the originating protocol they are converted back from AMQP 0.9.1. For some protocols such as MQTT 3.1 this isn't too expensive as MQTT is mostly a fairly easily mapped subset of AMQP 0.9.1 but for others such as AMQP 1.0 the conversions are awkward and in some cases lossy even if consuming from the originating protocol. This PR instead wraps all incoming messages in their originating form into a generic, extensible message container type (mc). The container module exposes an API to get common message details such as size and various properties (ttl, priority etc) directly from the source data type. Each protocol needs to implement the mc behaviour such that when a message originating form one protocol is consumed by another protocol we convert it to the target protocol at that point. The message container also contains annotations, dead letter records and other meta data we need to record during the lifetime of a message. The original protocol message is never modified unless it is consumed. This includes conversion modules to and from amqp, amqpl (AMQP 0.9.1) and mqtt. COMMIT HISTORY: * Refactor away from using the delivery{} record In many places including exchange types. This should make it easier to move towards using a message container type instead of basic_message. Add mc module and move direct replies outside of exchange Lots of changes incl classic queues Implement stream support incl amqp conversions simplify mc state record move mc.erl mc dlx stuff recent history exchange Make tracking work But doesn't take a protocol agnostic approach as we just convert everything into AMQP legacy and back. Might be good enough for now. Tracing as a whole may want a bit of a re-vamp at some point. tidy make quorum queue peek work by legacy conversion dead lettering fixes dead lettering fixes CMQ fixes rabbit_trace type fixes fixes fix Fix classic queue props test assertion fix feature flag and backwards compat Enable message_container feature flag in some SUITEs Dialyzer fixes fixes fix test fixes Various Manually update a gazelle generated file until a gazelle enhancement can be made https://github.com/rabbitmq/rules_erlang/issues/185 Add message_containers_SUITE to bazel and regen bazel files with gazelle from rules_erlang@main Simplify essential proprty access Such as durable, ttl and priority by extracting them into annotations at message container init time. Move type to remove dependenc on amqp10 stuff in mc.erl mostly because I don't know how to make bazel do the right thing add more stuff Refine routing header stuff wip Cosmetics Do not use "maybe" as type name as "maybe" is a keyword since OTP 25 which makes Erlang LS complain. * Dedup death queue names * Fix function clause crashes Fix failing tests in the MQTT shared_SUITE: A classic queue message ID can be undefined as set in https://github.com/rabbitmq/rabbitmq-server/blob/fbe79ff47b4edbc0fd95457e623d6593161ad198/deps/rabbit/src/rabbit_classic_queue_index_v2.erl#L1048 Fix failing tests in the MQTT shared_SUITE-mixed: When feature flag message_containers is disabled, the message is not an #mc{} record, but a #basic_message{} record. * Fix is_utf8_no_null crash Prior to this commit, the function crashed if invalid UTF-8 was provided, e.g.: ``` 1> rabbit_misc:is_valid_shortstr(<<"😇"/utf16>>). ** exception error: no function clause matching rabbit_misc:is_utf8_no_null(<<216,61,222,7>>) (rabbit_misc.erl, line 1481) ``` * Implement mqtt mc behaviour For now via amqp translation. This is still work in progress, but the following SUITEs pass: ``` make -C deps/rabbitmq_mqtt ct-shared t=[mqtt,v5,cluster_size_1] FULL=1 make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_1] FULL=1 ``` * Shorten mc file names Module name length matters because for each persistent message the #mc{} record is persisted to disk. ``` 1> iolist_size(term_to_iovec({mc, rabbit_mc_amqp_legacy})). 30 2> iolist_size(term_to_iovec({mc, mc_amqpl})). 17 ``` This commit renames the mc modules: ``` ag -l rabbit_mc_amqp_legacy | xargs sed -i 's/rabbit_mc_amqp_legacy/mc_amqpl/g' ag -l rabbit_mc_amqp | xargs sed -i 's/rabbit_mc_amqp/mc_amqp/g' ag -l rabbit_mqtt_mc | xargs sed -i 's/rabbit_mqtt_mc/mc_mqtt/g' ``` * mc: make deaths an annotation + fixes * Fix mc_mqtt protocol_state callback * Fix test will_delay_node_restart ``` make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_3]:will_delay_node_restart FULL=1 ``` * Bazel run gazelle * mix format rabbitmqctl.ex * Ensure ttl annotation is refelected in amqp legacy protocol state * Fix id access in message store * Fix rabbit_message_interceptor_SUITE * dializer fixes * Fix rabbit:rabbit_message_interceptor_SUITE-mixed set_annotation/3 should not result in duplicate keys * Fix MQTT shared_SUITE-mixed Up to 3.12 non-MQTT publishes were always QoS 1 regardless of delivery_mode. https://github.com/rabbitmq/rabbitmq-server/blob/75a953ce286a10aca910c098805a4f545989af38/deps/rabbitmq_mqtt/src/rabbit_mqtt_processor.erl#L2075-L2076 From now on, non-MQTT publishes are QoS 1 if durable. This makes more sense. The MQTT plugin must send a #basic_message{} to an old node that does not understand message containers. * Field content of 'v1_0.data' can be binary Fix ``` bazel test //deps/rabbitmq_mqtt:shared_SUITE-mixed \ --test_env FOCUS="-group [mqtt,v4,cluster_size_1] -case trace" \ -t- --test_sharding_strategy=disabled ``` * Remove route/2 and implement route/3 for all exchange types. This removes the route/2 callback from rabbit_exchange_type and makes route/3 mandatory instead. This is a breaking change and will require all implementations of exchange types to update their code, however this is necessary anyway for them to correctly handle the mc type. stream filtering fixes * Translate directly from MQTT to AMQP 0.9.1 * handle undecoded properties in mc_compat amqpl: put clause in right order recover death deatails from amqp data * Replace callback init_amqp with convert_from * Fix return value of lists:keyfind/3 * Translate directly from AMQP 0.9.1 to MQTT * Fix MQTT payload size MQTT payload can be a list when converted from AMQP 0.9.1 for example First conversions tests Plus some other conversion related fixes. bazel bazel translate amqp 1.0 null to undefined mc: property/2 and correlation_id/message_id return type tagged values. To ensure we can support a variety of types better. The type type tags are AMQP 1.0 flavoured. fix death recovery mc_mqtt: impl new api Add callbacks to allow protocols to compact data before storage And make readable if needing to query things repeatedly. bazel fix * more decoding * tracking mixed versions compat * mc: flip default of `durable` annotation to save some data. Assuming most messages are durable and that in memory messages suffer less from persistence overhead it makes sense for a non existent `durable` annotation to mean durable=true. * mc conversion tests and tidy up * mc make x_header unstrict again * amqpl: death record fixes * bazel * amqp -> amqpl conversion test * Fix crash in mc_amqp:size/1 Body can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0/ ct-system t=java ``` on branch native-amqp. * Fix crash in lists:flatten/1 Data can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp. * Fix crash in rabbit_writer Running test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp resulted in the following crash: ``` crasher: initial call: rabbit_writer:enter_mainloop/2 pid: <0.711.0> registered_name: [] exception error: bad argument in function size/1 called as size([<<0>>,<<"Sw">>,[<<160,2>>,<<"hi">>]]) *** argument 1: not tuple or binary in call from rabbit_binary_generator:build_content_frames/7 (rabbit_binary_generator.erl, line 89) in call from rabbit_binary_generator:build_simple_content_frames/4 (rabbit_binary_generator.erl, line 61) in call from rabbit_writer:assemble_frames/5 (rabbit_writer.erl, line 334) in call from rabbit_writer:internal_send_command_async/3 (rabbit_writer.erl, line 365) in call from rabbit_writer:handle_message/2 (rabbit_writer.erl, line 265) in call from rabbit_writer:handle_message/3 (rabbit_writer.erl, line 232) in call from rabbit_writer:mainloop1/2 (rabbit_writer.erl, line 223) ``` because #content.payload_fragments_rev is currently supposed to be a flat list of binaries instead of being an iolist. This commit fixes this crash inefficiently by calling iolist_to_binary/1. A better solution would be to allow AMQP legacy's #content.payload_fragments_rev to be an iolist. * Add accidentally deleted line back * mc: optimise mc_amqp internal format By removint the outer records for message and delivery annotations as well as application properties and footers. * mc: optimis mc_amqp map_add by using upsert * mc: refactoring and bug fixes * mc_SUITE routingheader assertions * mc remove serialize/1 callback as only used by amqp * mc_amqp: avoid returning a nested list from protocol_state * test and bug fix * move infer_type to mc_util * mc fixes and additiona assertions * Support headers exchange routing for MQTT messages When a headers exchange is bound to the MQTT topic exchange, routing will be performend based on both MQTT topic (by the topic exchange) and MQTT User Property (by the headers exchange). This combines the best worlds of both MQTT 5.0 and AMQP 0.9.1 and enables powerful routing topologies. When the User Property contains the same name multiple times, only the last name (and value) will be considered by the headers exchange. * Fix crash when sending from stream to amqpl When publishing a message via the stream protocol and consuming it via AMQP 0.9.1, the following crash occurred prior to this commit: ``` crasher: initial call: rabbit_channel:init/1 pid: <0.818.0> registered_name: [] exception exit: {{badmatch,undefined}, [{rabbit_channel,handle_deliver0,4, [{file,"rabbit_channel.erl"}, {line,2728}]}, {lists,foldl,3,[{file,"lists.erl"},{line,1594}]}, {rabbit_channel,handle_cast,2, [{file,"rabbit_channel.erl"}, {line,728}]}, {gen_server2,handle_msg,2, [{file,"gen_server2.erl"},{line,1056}]}, {proc_lib,wake_up,3, [{file,"proc_lib.erl"},{line,251}]}]} ``` This commit first gives `mc:init/3` the chance to set exchange and routing_keys annotations. If not set, `rabbit_stream_queue` will set these annotations assuming the message was originally published via the stream protocol. * Support consistent hash exchange routing for MQTT 5.0 When a consistent hash exchange is bound to the MQTT topic exchange, MQTT 5.0 messages can be routed to queues consistently based on the Correlation-Data in the PUBLISH packet. * Convert MQTT 5.0 User Property * to AMQP 0.9.1 headers * from AMQP 0.9.1 headers * to AMQP 1.0 application properties and message annotations * from AMQP 1.0 application properties and message annotations * Make use of Annotations in mc_mqtt:protocol_state/2 mc_mqtt:protocol_state/2 includes Annotations as parameter. It's cleaner to make use of these Annotations when computing the protocol state instead of relying on the caller (rabbitmq_mqtt_processor) to compute the protocol state. * Enforce AMQP 0.9.1 field name length limit The AMQP 0.9.1 spec prohibits field names longer than 128 characters. Therefore, when converting AMQP 1.0 message annotations, application properties or MQTT 5.0 User Property to AMQP 0.9.1 headers, drop any names longer than 128 characters. * Fix type specs Apply feedback from Michael Davis Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * Add mc_mqtt unit test suite Implement mc_mqtt:x_header/2 * Translate indicator that payload is UTF-8 encoded when converting between MQTT 5.0 and AMQP 1.0 * Translate single amqp-value section from AMQP 1.0 to MQTT Convert to a text representation, if possible, and indicate to MQTT client that the payload is UTF-8 encoded. This way, the MQTT client will be able to parse the payload. If conversion to text representation is not possible, encode the payload using the AMQP 1.0 type system and indiate the encoding via Content-Type message/vnd.rabbitmq.amqp. This Content-Type is not registered. Type "message" makes sense since it's a message. Vendor tree "vnd.rabbitmq.amqp" makes sense since merely subtype "amqp" is not registered. * Fix payload conversion * Translate Response Topic between MQTT and AMQP Translate MQTT 5.0 Response Topic to AMQP 1.0 reply-to address and vice versa. The Response Topic must be a UTF-8 encoded string. This commit re-uses the already defined RabbitMQ target addresses: ``` "/topic/" RK Publish to amq.topic with routing key RK "/exchange/" X "/" RK Publish to exchange X with routing key RK ``` By default, the MQTT topic exchange is configure dto be amq.topic using the 1st target address. When an operator modifies the mqtt.exchange, the 2nd target address is used. * Apply PR feedback and fix formatting Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * tidy up * Add MQTT message_containers test * consistent hash exchange: avoid amqp legacy conversion When hashing on a header value. * Avoid converting to amqp legacy when using exchange federation * Fix test flake * test and dialyzer fixes * dialyzer fix * Add MQTT protocol interoperability tests Test receiving from and sending to MQTT 5.0 and * AMQP 0.9.1 * AMQP 1.0 * STOMP * Streams * Regenerate portions of deps/rabbit/app.bzl with gazelle I'm not exactly sure how this happened, but gazell seems to have been run with an older version of the rules_erlang gazelle extension at some point. This caused generation of a structure that is no longer used. This commit updates the structure to the current pattern. * mc: refactoring * mc_amqpl: handle delivery annotations Just in case they are included. Also use iolist_to_iovec to create flat list of binaries when converting from amqp with amqp encoded payload. --------- Co-authored-by: David Ansari <david.ansari@gmx.de> Co-authored-by: Michael Davis <mcarsondavis@gmail.com> Co-authored-by: Rin Kuryloski <kuryloskip@vmware.com>
2023-08-31 18:27:13 +08:00
BMsg = rabbit_basic:message(rabbit_misc:r(<<>>, exchange, <<>>),
Stream Queue This is an aggregated commit of all changes related to the initial implementation of queue types and on top of that the stream queue type. The varios commit messages have simply been included mostly un-edited below. Make rabbit_amqqueue:not_found_or_absent_dirty/1 visible For use in the stream plugin. Use bigger retention policy on max-age test Set coordinator timeout to 30s Handle coordinator unavailable error Handle operator policies as maps when checking if is applicable Add is_policy_applicable/2 to classic queues Ignore restart commands if the stream has been deleted It could happen that after termination some of the monitors are still up and trigger writer/replica restarts Policy support on stream queues Remove subscription events on stream coordinator Ensure old leaders are removed from monitors Introduce delay when retrying a failed phase Note that this ensures monitor is setup, there was a bug where no monitor was really started when re-trying the same phase Restart replicas after leader election instead of relying on old monitors Use timer for stream coordinator retries Fix stream stats for members/online Multiple fixes for replica monitoring and restart Ensure pending commands are appended at the end and re-run Ensure phase is reset with the state Remove duplicates from replica list Restart current phase on state_enter Remove unused import Ensure rabbit is running when checking for stream quorum Restart replicas Add a close/1 function to queue types So that we can get a chance of cleaning up resources if needed. Stream queues close their osiris logs at this point. fix compiler errors stream-queue: take retention into account When calculating ready messages metrics. Add osiris to the list of rabbit deps Retry restart of replicas Do not restart replicas or leaders after receiving a delete cluster command Add more logging to the stream coordinator Monitor subscribed processes on the stream coordinator Memory breakdown for stream queues Update quorum queue event formatter rabbit_msg_record fixes Refactor channel confirms Remove old unconfirmed_messages module that was designed to handle multiple queue fan in logic including all ha mirrors etc. Replaced with simpler rabbit_confirms module that handles the fan out and leaves any queue specific logic (such as confirms from mirrors) to the queue type implemention. Also this module has a dedicated test module. Which is nice. Backward compatibility with 3.8.x events Supports mixed version cluster upgrades Match specification when stream queue already exists Max age retention for stream queues Stop all replicas before starting leader election stream: disallow global qos remove IS_CLASSIC|QUORUM macros Ensure only classic queues are notified on channel down This also removes the delivering_queues map in the channel state as it should not be needed for this and just cause additional unecessary accounting. Polish AMQP 1.0/0.9.1 properties conversion Support byte in application properties, handle 1-bit representation for booleans. Use binary in header for long AMQP 1.0 ID Fix AMQP 1.0 to 0.9.1 conversion Fix test due to incorrect type Convert timestamp application properties to/from seconds AMQP 1.0 uses milliseconds for timestamp and AMQP 0.9.1 uses seconds, so conversion needed. Dialyzer fixes Handle all message-id types AMQP 1.0 is more liberal in it's allowed types of message-id and correlation-id - this adds headers to describe the type of the data in the message_id / correlation_id properties and also handles the case where the data cannot fit by again using headers. Resize stream coordinator cluster when broker configuration changes convert timestamp to and fro seconds user_id should be a binary message annotations keys need to be symbols stream-queue: default exchange and routing key As these won't be present for data written using the rabbitmq-stream plugin. Add exchange, routing key as message annotations To the AMQP 1.0 formatted data to enable roundtrip. Add osiris logging module config And update logging config test suite. Restart election when start of new leader fails The node might have just gone down so we need to try another one Only aux keeps track of phase now, as it might change if the leader election fails Stream coordinator refactor - all state is kept on the ra machine Ensure any ra cluster not a qq is not cleaned up Fixes to recovery and monitoring Add AMQP 1.0 common to dependencies Add rabbit_msg_record module To handle conversions into internal stream storage format. Use rabbitmq-common stream-queue branch Use SSH for osiris dependency Stream coordinator: delete replica Stream coordinator: add replica Stream coordinator: leader failover Stream coordinator: declare and delete Test consuming from a random offset Previous offsets should not be delivered to consumers Consume from stream replicas and multiple test fixes Use max-length-bytes and add new max-segment-size Use SSH for osiris dependency Basic cancel for stream queues Publish stream queues and settle/reject/requeue refactor Consume from stream queues Fix recovery Publish stream messages Add/delete stream replicas Use safe queue names Set retention policy for stream queues Required by the ctl command [#171207092] Stream queue delete queue fix missing callback impl Stream queue declare Queue type abstraction And use the implementing module as the value of the amqqueue record `type` field. This will allow for easy dispatch to the queue type implementation. Queue type abstraction Move queue declare into rabbit_queue_type Move queue delete into queue type implementation Queue type: dequeue/basic_get Move info inside queue type abstraction Move policy change into queue type interface Add purge to queue type Add recovery to the queue type interface Rename amqqueue quorum_nodes field To a more generic an extensible opaque queue type specific map. Fix tests and handle classic API response Fix HA queue confirm bug All mirrors need to be present as queue names. This introduces context linking allowing additional queue refs to be linked to a single "master" queue ref contining the actual queue context. Fix issue with events of deleted queues Also update queue type smoke test to use a cluster by default. correct default value of amqqueue getter Move classic queues further inside queue type interface why [TrackerId] Dialyzer fixes
2020-09-29 18:43:24 +08:00
<<>>, #'P_basic'{delivery_mode = 2},
Payload),
Message Containers (#5077) This PR implements an approach for a "protocol (data format) agnostic core" where the format of the message isn't converted at point of reception. Currently all non AMQP 0.9.1 originating messages are converted into a AMQP 0.9.1 flavoured basic_message record before sent to a queue. If the messages are then consumed by the originating protocol they are converted back from AMQP 0.9.1. For some protocols such as MQTT 3.1 this isn't too expensive as MQTT is mostly a fairly easily mapped subset of AMQP 0.9.1 but for others such as AMQP 1.0 the conversions are awkward and in some cases lossy even if consuming from the originating protocol. This PR instead wraps all incoming messages in their originating form into a generic, extensible message container type (mc). The container module exposes an API to get common message details such as size and various properties (ttl, priority etc) directly from the source data type. Each protocol needs to implement the mc behaviour such that when a message originating form one protocol is consumed by another protocol we convert it to the target protocol at that point. The message container also contains annotations, dead letter records and other meta data we need to record during the lifetime of a message. The original protocol message is never modified unless it is consumed. This includes conversion modules to and from amqp, amqpl (AMQP 0.9.1) and mqtt. COMMIT HISTORY: * Refactor away from using the delivery{} record In many places including exchange types. This should make it easier to move towards using a message container type instead of basic_message. Add mc module and move direct replies outside of exchange Lots of changes incl classic queues Implement stream support incl amqp conversions simplify mc state record move mc.erl mc dlx stuff recent history exchange Make tracking work But doesn't take a protocol agnostic approach as we just convert everything into AMQP legacy and back. Might be good enough for now. Tracing as a whole may want a bit of a re-vamp at some point. tidy make quorum queue peek work by legacy conversion dead lettering fixes dead lettering fixes CMQ fixes rabbit_trace type fixes fixes fix Fix classic queue props test assertion fix feature flag and backwards compat Enable message_container feature flag in some SUITEs Dialyzer fixes fixes fix test fixes Various Manually update a gazelle generated file until a gazelle enhancement can be made https://github.com/rabbitmq/rules_erlang/issues/185 Add message_containers_SUITE to bazel and regen bazel files with gazelle from rules_erlang@main Simplify essential proprty access Such as durable, ttl and priority by extracting them into annotations at message container init time. Move type to remove dependenc on amqp10 stuff in mc.erl mostly because I don't know how to make bazel do the right thing add more stuff Refine routing header stuff wip Cosmetics Do not use "maybe" as type name as "maybe" is a keyword since OTP 25 which makes Erlang LS complain. * Dedup death queue names * Fix function clause crashes Fix failing tests in the MQTT shared_SUITE: A classic queue message ID can be undefined as set in https://github.com/rabbitmq/rabbitmq-server/blob/fbe79ff47b4edbc0fd95457e623d6593161ad198/deps/rabbit/src/rabbit_classic_queue_index_v2.erl#L1048 Fix failing tests in the MQTT shared_SUITE-mixed: When feature flag message_containers is disabled, the message is not an #mc{} record, but a #basic_message{} record. * Fix is_utf8_no_null crash Prior to this commit, the function crashed if invalid UTF-8 was provided, e.g.: ``` 1> rabbit_misc:is_valid_shortstr(<<"😇"/utf16>>). ** exception error: no function clause matching rabbit_misc:is_utf8_no_null(<<216,61,222,7>>) (rabbit_misc.erl, line 1481) ``` * Implement mqtt mc behaviour For now via amqp translation. This is still work in progress, but the following SUITEs pass: ``` make -C deps/rabbitmq_mqtt ct-shared t=[mqtt,v5,cluster_size_1] FULL=1 make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_1] FULL=1 ``` * Shorten mc file names Module name length matters because for each persistent message the #mc{} record is persisted to disk. ``` 1> iolist_size(term_to_iovec({mc, rabbit_mc_amqp_legacy})). 30 2> iolist_size(term_to_iovec({mc, mc_amqpl})). 17 ``` This commit renames the mc modules: ``` ag -l rabbit_mc_amqp_legacy | xargs sed -i 's/rabbit_mc_amqp_legacy/mc_amqpl/g' ag -l rabbit_mc_amqp | xargs sed -i 's/rabbit_mc_amqp/mc_amqp/g' ag -l rabbit_mqtt_mc | xargs sed -i 's/rabbit_mqtt_mc/mc_mqtt/g' ``` * mc: make deaths an annotation + fixes * Fix mc_mqtt protocol_state callback * Fix test will_delay_node_restart ``` make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_3]:will_delay_node_restart FULL=1 ``` * Bazel run gazelle * mix format rabbitmqctl.ex * Ensure ttl annotation is refelected in amqp legacy protocol state * Fix id access in message store * Fix rabbit_message_interceptor_SUITE * dializer fixes * Fix rabbit:rabbit_message_interceptor_SUITE-mixed set_annotation/3 should not result in duplicate keys * Fix MQTT shared_SUITE-mixed Up to 3.12 non-MQTT publishes were always QoS 1 regardless of delivery_mode. https://github.com/rabbitmq/rabbitmq-server/blob/75a953ce286a10aca910c098805a4f545989af38/deps/rabbitmq_mqtt/src/rabbit_mqtt_processor.erl#L2075-L2076 From now on, non-MQTT publishes are QoS 1 if durable. This makes more sense. The MQTT plugin must send a #basic_message{} to an old node that does not understand message containers. * Field content of 'v1_0.data' can be binary Fix ``` bazel test //deps/rabbitmq_mqtt:shared_SUITE-mixed \ --test_env FOCUS="-group [mqtt,v4,cluster_size_1] -case trace" \ -t- --test_sharding_strategy=disabled ``` * Remove route/2 and implement route/3 for all exchange types. This removes the route/2 callback from rabbit_exchange_type and makes route/3 mandatory instead. This is a breaking change and will require all implementations of exchange types to update their code, however this is necessary anyway for them to correctly handle the mc type. stream filtering fixes * Translate directly from MQTT to AMQP 0.9.1 * handle undecoded properties in mc_compat amqpl: put clause in right order recover death deatails from amqp data * Replace callback init_amqp with convert_from * Fix return value of lists:keyfind/3 * Translate directly from AMQP 0.9.1 to MQTT * Fix MQTT payload size MQTT payload can be a list when converted from AMQP 0.9.1 for example First conversions tests Plus some other conversion related fixes. bazel bazel translate amqp 1.0 null to undefined mc: property/2 and correlation_id/message_id return type tagged values. To ensure we can support a variety of types better. The type type tags are AMQP 1.0 flavoured. fix death recovery mc_mqtt: impl new api Add callbacks to allow protocols to compact data before storage And make readable if needing to query things repeatedly. bazel fix * more decoding * tracking mixed versions compat * mc: flip default of `durable` annotation to save some data. Assuming most messages are durable and that in memory messages suffer less from persistence overhead it makes sense for a non existent `durable` annotation to mean durable=true. * mc conversion tests and tidy up * mc make x_header unstrict again * amqpl: death record fixes * bazel * amqp -> amqpl conversion test * Fix crash in mc_amqp:size/1 Body can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0/ ct-system t=java ``` on branch native-amqp. * Fix crash in lists:flatten/1 Data can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp. * Fix crash in rabbit_writer Running test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp resulted in the following crash: ``` crasher: initial call: rabbit_writer:enter_mainloop/2 pid: <0.711.0> registered_name: [] exception error: bad argument in function size/1 called as size([<<0>>,<<"Sw">>,[<<160,2>>,<<"hi">>]]) *** argument 1: not tuple or binary in call from rabbit_binary_generator:build_content_frames/7 (rabbit_binary_generator.erl, line 89) in call from rabbit_binary_generator:build_simple_content_frames/4 (rabbit_binary_generator.erl, line 61) in call from rabbit_writer:assemble_frames/5 (rabbit_writer.erl, line 334) in call from rabbit_writer:internal_send_command_async/3 (rabbit_writer.erl, line 365) in call from rabbit_writer:handle_message/2 (rabbit_writer.erl, line 265) in call from rabbit_writer:handle_message/3 (rabbit_writer.erl, line 232) in call from rabbit_writer:mainloop1/2 (rabbit_writer.erl, line 223) ``` because #content.payload_fragments_rev is currently supposed to be a flat list of binaries instead of being an iolist. This commit fixes this crash inefficiently by calling iolist_to_binary/1. A better solution would be to allow AMQP legacy's #content.payload_fragments_rev to be an iolist. * Add accidentally deleted line back * mc: optimise mc_amqp internal format By removint the outer records for message and delivery annotations as well as application properties and footers. * mc: optimis mc_amqp map_add by using upsert * mc: refactoring and bug fixes * mc_SUITE routingheader assertions * mc remove serialize/1 callback as only used by amqp * mc_amqp: avoid returning a nested list from protocol_state * test and bug fix * move infer_type to mc_util * mc fixes and additiona assertions * Support headers exchange routing for MQTT messages When a headers exchange is bound to the MQTT topic exchange, routing will be performend based on both MQTT topic (by the topic exchange) and MQTT User Property (by the headers exchange). This combines the best worlds of both MQTT 5.0 and AMQP 0.9.1 and enables powerful routing topologies. When the User Property contains the same name multiple times, only the last name (and value) will be considered by the headers exchange. * Fix crash when sending from stream to amqpl When publishing a message via the stream protocol and consuming it via AMQP 0.9.1, the following crash occurred prior to this commit: ``` crasher: initial call: rabbit_channel:init/1 pid: <0.818.0> registered_name: [] exception exit: {{badmatch,undefined}, [{rabbit_channel,handle_deliver0,4, [{file,"rabbit_channel.erl"}, {line,2728}]}, {lists,foldl,3,[{file,"lists.erl"},{line,1594}]}, {rabbit_channel,handle_cast,2, [{file,"rabbit_channel.erl"}, {line,728}]}, {gen_server2,handle_msg,2, [{file,"gen_server2.erl"},{line,1056}]}, {proc_lib,wake_up,3, [{file,"proc_lib.erl"},{line,251}]}]} ``` This commit first gives `mc:init/3` the chance to set exchange and routing_keys annotations. If not set, `rabbit_stream_queue` will set these annotations assuming the message was originally published via the stream protocol. * Support consistent hash exchange routing for MQTT 5.0 When a consistent hash exchange is bound to the MQTT topic exchange, MQTT 5.0 messages can be routed to queues consistently based on the Correlation-Data in the PUBLISH packet. * Convert MQTT 5.0 User Property * to AMQP 0.9.1 headers * from AMQP 0.9.1 headers * to AMQP 1.0 application properties and message annotations * from AMQP 1.0 application properties and message annotations * Make use of Annotations in mc_mqtt:protocol_state/2 mc_mqtt:protocol_state/2 includes Annotations as parameter. It's cleaner to make use of these Annotations when computing the protocol state instead of relying on the caller (rabbitmq_mqtt_processor) to compute the protocol state. * Enforce AMQP 0.9.1 field name length limit The AMQP 0.9.1 spec prohibits field names longer than 128 characters. Therefore, when converting AMQP 1.0 message annotations, application properties or MQTT 5.0 User Property to AMQP 0.9.1 headers, drop any names longer than 128 characters. * Fix type specs Apply feedback from Michael Davis Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * Add mc_mqtt unit test suite Implement mc_mqtt:x_header/2 * Translate indicator that payload is UTF-8 encoded when converting between MQTT 5.0 and AMQP 1.0 * Translate single amqp-value section from AMQP 1.0 to MQTT Convert to a text representation, if possible, and indicate to MQTT client that the payload is UTF-8 encoded. This way, the MQTT client will be able to parse the payload. If conversion to text representation is not possible, encode the payload using the AMQP 1.0 type system and indiate the encoding via Content-Type message/vnd.rabbitmq.amqp. This Content-Type is not registered. Type "message" makes sense since it's a message. Vendor tree "vnd.rabbitmq.amqp" makes sense since merely subtype "amqp" is not registered. * Fix payload conversion * Translate Response Topic between MQTT and AMQP Translate MQTT 5.0 Response Topic to AMQP 1.0 reply-to address and vice versa. The Response Topic must be a UTF-8 encoded string. This commit re-uses the already defined RabbitMQ target addresses: ``` "/topic/" RK Publish to amq.topic with routing key RK "/exchange/" X "/" RK Publish to exchange X with routing key RK ``` By default, the MQTT topic exchange is configure dto be amq.topic using the 1st target address. When an operator modifies the mqtt.exchange, the 2nd target address is used. * Apply PR feedback and fix formatting Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * tidy up * Add MQTT message_containers test * consistent hash exchange: avoid amqp legacy conversion When hashing on a header value. * Avoid converting to amqp legacy when using exchange federation * Fix test flake * test and dialyzer fixes * dialyzer fix * Add MQTT protocol interoperability tests Test receiving from and sending to MQTT 5.0 and * AMQP 0.9.1 * AMQP 1.0 * STOMP * Streams * Regenerate portions of deps/rabbit/app.bzl with gazelle I'm not exactly sure how this happened, but gazell seems to have been run with an older version of the rules_erlang gazelle extension at some point. This caused generation of a structure that is no longer used. This commit updates the structure to the current pattern. * mc: refactoring * mc_amqpl: handle delivery annotations Just in case they are included. Also use iolist_to_iovec to create flat list of binaries when converting from amqp with amqp encoded payload. --------- Co-authored-by: David Ansari <david.ansari@gmx.de> Co-authored-by: Michael Davis <mcarsondavis@gmail.com> Co-authored-by: Rin Kuryloski <kuryloskip@vmware.com>
2023-08-31 18:27:13 +08:00
Content = BMsg#basic_message.content,
Ex = BMsg#basic_message.exchange_name,
{ok, Msg} = mc_amqpl:message(Ex, <<>>, Content),
Message Containers (#5077) This PR implements an approach for a "protocol (data format) agnostic core" where the format of the message isn't converted at point of reception. Currently all non AMQP 0.9.1 originating messages are converted into a AMQP 0.9.1 flavoured basic_message record before sent to a queue. If the messages are then consumed by the originating protocol they are converted back from AMQP 0.9.1. For some protocols such as MQTT 3.1 this isn't too expensive as MQTT is mostly a fairly easily mapped subset of AMQP 0.9.1 but for others such as AMQP 1.0 the conversions are awkward and in some cases lossy even if consuming from the originating protocol. This PR instead wraps all incoming messages in their originating form into a generic, extensible message container type (mc). The container module exposes an API to get common message details such as size and various properties (ttl, priority etc) directly from the source data type. Each protocol needs to implement the mc behaviour such that when a message originating form one protocol is consumed by another protocol we convert it to the target protocol at that point. The message container also contains annotations, dead letter records and other meta data we need to record during the lifetime of a message. The original protocol message is never modified unless it is consumed. This includes conversion modules to and from amqp, amqpl (AMQP 0.9.1) and mqtt. COMMIT HISTORY: * Refactor away from using the delivery{} record In many places including exchange types. This should make it easier to move towards using a message container type instead of basic_message. Add mc module and move direct replies outside of exchange Lots of changes incl classic queues Implement stream support incl amqp conversions simplify mc state record move mc.erl mc dlx stuff recent history exchange Make tracking work But doesn't take a protocol agnostic approach as we just convert everything into AMQP legacy and back. Might be good enough for now. Tracing as a whole may want a bit of a re-vamp at some point. tidy make quorum queue peek work by legacy conversion dead lettering fixes dead lettering fixes CMQ fixes rabbit_trace type fixes fixes fix Fix classic queue props test assertion fix feature flag and backwards compat Enable message_container feature flag in some SUITEs Dialyzer fixes fixes fix test fixes Various Manually update a gazelle generated file until a gazelle enhancement can be made https://github.com/rabbitmq/rules_erlang/issues/185 Add message_containers_SUITE to bazel and regen bazel files with gazelle from rules_erlang@main Simplify essential proprty access Such as durable, ttl and priority by extracting them into annotations at message container init time. Move type to remove dependenc on amqp10 stuff in mc.erl mostly because I don't know how to make bazel do the right thing add more stuff Refine routing header stuff wip Cosmetics Do not use "maybe" as type name as "maybe" is a keyword since OTP 25 which makes Erlang LS complain. * Dedup death queue names * Fix function clause crashes Fix failing tests in the MQTT shared_SUITE: A classic queue message ID can be undefined as set in https://github.com/rabbitmq/rabbitmq-server/blob/fbe79ff47b4edbc0fd95457e623d6593161ad198/deps/rabbit/src/rabbit_classic_queue_index_v2.erl#L1048 Fix failing tests in the MQTT shared_SUITE-mixed: When feature flag message_containers is disabled, the message is not an #mc{} record, but a #basic_message{} record. * Fix is_utf8_no_null crash Prior to this commit, the function crashed if invalid UTF-8 was provided, e.g.: ``` 1> rabbit_misc:is_valid_shortstr(<<"😇"/utf16>>). ** exception error: no function clause matching rabbit_misc:is_utf8_no_null(<<216,61,222,7>>) (rabbit_misc.erl, line 1481) ``` * Implement mqtt mc behaviour For now via amqp translation. This is still work in progress, but the following SUITEs pass: ``` make -C deps/rabbitmq_mqtt ct-shared t=[mqtt,v5,cluster_size_1] FULL=1 make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_1] FULL=1 ``` * Shorten mc file names Module name length matters because for each persistent message the #mc{} record is persisted to disk. ``` 1> iolist_size(term_to_iovec({mc, rabbit_mc_amqp_legacy})). 30 2> iolist_size(term_to_iovec({mc, mc_amqpl})). 17 ``` This commit renames the mc modules: ``` ag -l rabbit_mc_amqp_legacy | xargs sed -i 's/rabbit_mc_amqp_legacy/mc_amqpl/g' ag -l rabbit_mc_amqp | xargs sed -i 's/rabbit_mc_amqp/mc_amqp/g' ag -l rabbit_mqtt_mc | xargs sed -i 's/rabbit_mqtt_mc/mc_mqtt/g' ``` * mc: make deaths an annotation + fixes * Fix mc_mqtt protocol_state callback * Fix test will_delay_node_restart ``` make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_3]:will_delay_node_restart FULL=1 ``` * Bazel run gazelle * mix format rabbitmqctl.ex * Ensure ttl annotation is refelected in amqp legacy protocol state * Fix id access in message store * Fix rabbit_message_interceptor_SUITE * dializer fixes * Fix rabbit:rabbit_message_interceptor_SUITE-mixed set_annotation/3 should not result in duplicate keys * Fix MQTT shared_SUITE-mixed Up to 3.12 non-MQTT publishes were always QoS 1 regardless of delivery_mode. https://github.com/rabbitmq/rabbitmq-server/blob/75a953ce286a10aca910c098805a4f545989af38/deps/rabbitmq_mqtt/src/rabbit_mqtt_processor.erl#L2075-L2076 From now on, non-MQTT publishes are QoS 1 if durable. This makes more sense. The MQTT plugin must send a #basic_message{} to an old node that does not understand message containers. * Field content of 'v1_0.data' can be binary Fix ``` bazel test //deps/rabbitmq_mqtt:shared_SUITE-mixed \ --test_env FOCUS="-group [mqtt,v4,cluster_size_1] -case trace" \ -t- --test_sharding_strategy=disabled ``` * Remove route/2 and implement route/3 for all exchange types. This removes the route/2 callback from rabbit_exchange_type and makes route/3 mandatory instead. This is a breaking change and will require all implementations of exchange types to update their code, however this is necessary anyway for them to correctly handle the mc type. stream filtering fixes * Translate directly from MQTT to AMQP 0.9.1 * handle undecoded properties in mc_compat amqpl: put clause in right order recover death deatails from amqp data * Replace callback init_amqp with convert_from * Fix return value of lists:keyfind/3 * Translate directly from AMQP 0.9.1 to MQTT * Fix MQTT payload size MQTT payload can be a list when converted from AMQP 0.9.1 for example First conversions tests Plus some other conversion related fixes. bazel bazel translate amqp 1.0 null to undefined mc: property/2 and correlation_id/message_id return type tagged values. To ensure we can support a variety of types better. The type type tags are AMQP 1.0 flavoured. fix death recovery mc_mqtt: impl new api Add callbacks to allow protocols to compact data before storage And make readable if needing to query things repeatedly. bazel fix * more decoding * tracking mixed versions compat * mc: flip default of `durable` annotation to save some data. Assuming most messages are durable and that in memory messages suffer less from persistence overhead it makes sense for a non existent `durable` annotation to mean durable=true. * mc conversion tests and tidy up * mc make x_header unstrict again * amqpl: death record fixes * bazel * amqp -> amqpl conversion test * Fix crash in mc_amqp:size/1 Body can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0/ ct-system t=java ``` on branch native-amqp. * Fix crash in lists:flatten/1 Data can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp. * Fix crash in rabbit_writer Running test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp resulted in the following crash: ``` crasher: initial call: rabbit_writer:enter_mainloop/2 pid: <0.711.0> registered_name: [] exception error: bad argument in function size/1 called as size([<<0>>,<<"Sw">>,[<<160,2>>,<<"hi">>]]) *** argument 1: not tuple or binary in call from rabbit_binary_generator:build_content_frames/7 (rabbit_binary_generator.erl, line 89) in call from rabbit_binary_generator:build_simple_content_frames/4 (rabbit_binary_generator.erl, line 61) in call from rabbit_writer:assemble_frames/5 (rabbit_writer.erl, line 334) in call from rabbit_writer:internal_send_command_async/3 (rabbit_writer.erl, line 365) in call from rabbit_writer:handle_message/2 (rabbit_writer.erl, line 265) in call from rabbit_writer:handle_message/3 (rabbit_writer.erl, line 232) in call from rabbit_writer:mainloop1/2 (rabbit_writer.erl, line 223) ``` because #content.payload_fragments_rev is currently supposed to be a flat list of binaries instead of being an iolist. This commit fixes this crash inefficiently by calling iolist_to_binary/1. A better solution would be to allow AMQP legacy's #content.payload_fragments_rev to be an iolist. * Add accidentally deleted line back * mc: optimise mc_amqp internal format By removint the outer records for message and delivery annotations as well as application properties and footers. * mc: optimis mc_amqp map_add by using upsert * mc: refactoring and bug fixes * mc_SUITE routingheader assertions * mc remove serialize/1 callback as only used by amqp * mc_amqp: avoid returning a nested list from protocol_state * test and bug fix * move infer_type to mc_util * mc fixes and additiona assertions * Support headers exchange routing for MQTT messages When a headers exchange is bound to the MQTT topic exchange, routing will be performend based on both MQTT topic (by the topic exchange) and MQTT User Property (by the headers exchange). This combines the best worlds of both MQTT 5.0 and AMQP 0.9.1 and enables powerful routing topologies. When the User Property contains the same name multiple times, only the last name (and value) will be considered by the headers exchange. * Fix crash when sending from stream to amqpl When publishing a message via the stream protocol and consuming it via AMQP 0.9.1, the following crash occurred prior to this commit: ``` crasher: initial call: rabbit_channel:init/1 pid: <0.818.0> registered_name: [] exception exit: {{badmatch,undefined}, [{rabbit_channel,handle_deliver0,4, [{file,"rabbit_channel.erl"}, {line,2728}]}, {lists,foldl,3,[{file,"lists.erl"},{line,1594}]}, {rabbit_channel,handle_cast,2, [{file,"rabbit_channel.erl"}, {line,728}]}, {gen_server2,handle_msg,2, [{file,"gen_server2.erl"},{line,1056}]}, {proc_lib,wake_up,3, [{file,"proc_lib.erl"},{line,251}]}]} ``` This commit first gives `mc:init/3` the chance to set exchange and routing_keys annotations. If not set, `rabbit_stream_queue` will set these annotations assuming the message was originally published via the stream protocol. * Support consistent hash exchange routing for MQTT 5.0 When a consistent hash exchange is bound to the MQTT topic exchange, MQTT 5.0 messages can be routed to queues consistently based on the Correlation-Data in the PUBLISH packet. * Convert MQTT 5.0 User Property * to AMQP 0.9.1 headers * from AMQP 0.9.1 headers * to AMQP 1.0 application properties and message annotations * from AMQP 1.0 application properties and message annotations * Make use of Annotations in mc_mqtt:protocol_state/2 mc_mqtt:protocol_state/2 includes Annotations as parameter. It's cleaner to make use of these Annotations when computing the protocol state instead of relying on the caller (rabbitmq_mqtt_processor) to compute the protocol state. * Enforce AMQP 0.9.1 field name length limit The AMQP 0.9.1 spec prohibits field names longer than 128 characters. Therefore, when converting AMQP 1.0 message annotations, application properties or MQTT 5.0 User Property to AMQP 0.9.1 headers, drop any names longer than 128 characters. * Fix type specs Apply feedback from Michael Davis Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * Add mc_mqtt unit test suite Implement mc_mqtt:x_header/2 * Translate indicator that payload is UTF-8 encoded when converting between MQTT 5.0 and AMQP 1.0 * Translate single amqp-value section from AMQP 1.0 to MQTT Convert to a text representation, if possible, and indicate to MQTT client that the payload is UTF-8 encoded. This way, the MQTT client will be able to parse the payload. If conversion to text representation is not possible, encode the payload using the AMQP 1.0 type system and indiate the encoding via Content-Type message/vnd.rabbitmq.amqp. This Content-Type is not registered. Type "message" makes sense since it's a message. Vendor tree "vnd.rabbitmq.amqp" makes sense since merely subtype "amqp" is not registered. * Fix payload conversion * Translate Response Topic between MQTT and AMQP Translate MQTT 5.0 Response Topic to AMQP 1.0 reply-to address and vice versa. The Response Topic must be a UTF-8 encoded string. This commit re-uses the already defined RabbitMQ target addresses: ``` "/topic/" RK Publish to amq.topic with routing key RK "/exchange/" X "/" RK Publish to exchange X with routing key RK ``` By default, the MQTT topic exchange is configure dto be amq.topic using the 1st target address. When an operator modifies the mqtt.exchange, the 2nd target address is used. * Apply PR feedback and fix formatting Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * tidy up * Add MQTT message_containers test * consistent hash exchange: avoid amqp legacy conversion When hashing on a header value. * Avoid converting to amqp legacy when using exchange federation * Fix test flake * test and dialyzer fixes * dialyzer fix * Add MQTT protocol interoperability tests Test receiving from and sending to MQTT 5.0 and * AMQP 0.9.1 * AMQP 1.0 * STOMP * Streams * Regenerate portions of deps/rabbit/app.bzl with gazelle I'm not exactly sure how this happened, but gazell seems to have been run with an older version of the rules_erlang gazelle extension at some point. This caused generation of a structure that is no longer used. This commit updates the structure to the current pattern. * mc: refactoring * mc_amqpl: handle delivery annotations Just in case they are included. Also use iolist_to_iovec to create flat list of binaries when converting from amqp with amqp encoded payload. --------- Co-authored-by: David Ansari <david.ansari@gmx.de> Co-authored-by: Michael Davis <mcarsondavis@gmail.com> Co-authored-by: Rin Kuryloski <kuryloskip@vmware.com>
2023-08-31 18:27:13 +08:00
Options = #{correlation => Seq},
{ok, Acc, _Actions} = rabbit_queue_type:deliver([Q], Msg,
Options, Acc0),
Stream Queue This is an aggregated commit of all changes related to the initial implementation of queue types and on top of that the stream queue type. The varios commit messages have simply been included mostly un-edited below. Make rabbit_amqqueue:not_found_or_absent_dirty/1 visible For use in the stream plugin. Use bigger retention policy on max-age test Set coordinator timeout to 30s Handle coordinator unavailable error Handle operator policies as maps when checking if is applicable Add is_policy_applicable/2 to classic queues Ignore restart commands if the stream has been deleted It could happen that after termination some of the monitors are still up and trigger writer/replica restarts Policy support on stream queues Remove subscription events on stream coordinator Ensure old leaders are removed from monitors Introduce delay when retrying a failed phase Note that this ensures monitor is setup, there was a bug where no monitor was really started when re-trying the same phase Restart replicas after leader election instead of relying on old monitors Use timer for stream coordinator retries Fix stream stats for members/online Multiple fixes for replica monitoring and restart Ensure pending commands are appended at the end and re-run Ensure phase is reset with the state Remove duplicates from replica list Restart current phase on state_enter Remove unused import Ensure rabbit is running when checking for stream quorum Restart replicas Add a close/1 function to queue types So that we can get a chance of cleaning up resources if needed. Stream queues close their osiris logs at this point. fix compiler errors stream-queue: take retention into account When calculating ready messages metrics. Add osiris to the list of rabbit deps Retry restart of replicas Do not restart replicas or leaders after receiving a delete cluster command Add more logging to the stream coordinator Monitor subscribed processes on the stream coordinator Memory breakdown for stream queues Update quorum queue event formatter rabbit_msg_record fixes Refactor channel confirms Remove old unconfirmed_messages module that was designed to handle multiple queue fan in logic including all ha mirrors etc. Replaced with simpler rabbit_confirms module that handles the fan out and leaves any queue specific logic (such as confirms from mirrors) to the queue type implemention. Also this module has a dedicated test module. Which is nice. Backward compatibility with 3.8.x events Supports mixed version cluster upgrades Match specification when stream queue already exists Max age retention for stream queues Stop all replicas before starting leader election stream: disallow global qos remove IS_CLASSIC|QUORUM macros Ensure only classic queues are notified on channel down This also removes the delivering_queues map in the channel state as it should not be needed for this and just cause additional unecessary accounting. Polish AMQP 1.0/0.9.1 properties conversion Support byte in application properties, handle 1-bit representation for booleans. Use binary in header for long AMQP 1.0 ID Fix AMQP 1.0 to 0.9.1 conversion Fix test due to incorrect type Convert timestamp application properties to/from seconds AMQP 1.0 uses milliseconds for timestamp and AMQP 0.9.1 uses seconds, so conversion needed. Dialyzer fixes Handle all message-id types AMQP 1.0 is more liberal in it's allowed types of message-id and correlation-id - this adds headers to describe the type of the data in the message_id / correlation_id properties and also handles the case where the data cannot fit by again using headers. Resize stream coordinator cluster when broker configuration changes convert timestamp to and fro seconds user_id should be a binary message annotations keys need to be symbols stream-queue: default exchange and routing key As these won't be present for data written using the rabbitmq-stream plugin. Add exchange, routing key as message annotations To the AMQP 1.0 formatted data to enable roundtrip. Add osiris logging module config And update logging config test suite. Restart election when start of new leader fails The node might have just gone down so we need to try another one Only aux keeps track of phase now, as it might change if the leader election fails Stream coordinator refactor - all state is kept on the ra machine Ensure any ra cluster not a qq is not cleaned up Fixes to recovery and monitoring Add AMQP 1.0 common to dependencies Add rabbit_msg_record module To handle conversions into internal stream storage format. Use rabbitmq-common stream-queue branch Use SSH for osiris dependency Stream coordinator: delete replica Stream coordinator: add replica Stream coordinator: leader failover Stream coordinator: declare and delete Test consuming from a random offset Previous offsets should not be delivered to consumers Consume from stream replicas and multiple test fixes Use max-length-bytes and add new max-segment-size Use SSH for osiris dependency Basic cancel for stream queues Publish stream queues and settle/reject/requeue refactor Consume from stream queues Fix recovery Publish stream messages Add/delete stream replicas Use safe queue names Set retention policy for stream queues Required by the ctl command [#171207092] Stream queue delete queue fix missing callback impl Stream queue declare Queue type abstraction And use the implementing module as the value of the amqqueue record `type` field. This will allow for easy dispatch to the queue type implementation. Queue type abstraction Move queue declare into rabbit_queue_type Move queue delete into queue type implementation Queue type: dequeue/basic_get Move info inside queue type abstraction Move policy change into queue type interface Add purge to queue type Add recovery to the queue type interface Rename amqqueue quorum_nodes field To a more generic an extensible opaque queue type specific map. Fix tests and handle classic API response Fix HA queue confirm bug All mirrors need to be present as queue names. This introduces context linking allowing additional queue refs to be linked to a single "master" queue ref contining the actual queue context. Fix issue with events of deleted queues Also update queue type smoke test to use a cluster by default. correct default value of amqqueue getter Move classic queues further inside queue type interface why [TrackerId] Dialyzer fixes
2020-09-29 18:43:24 +08:00
Acc
end, QTState0, Seqs),
wait_for_confirms(sets:from_list(Seqs, [{version, 2}])),
Stream Queue This is an aggregated commit of all changes related to the initial implementation of queue types and on top of that the stream queue type. The varios commit messages have simply been included mostly un-edited below. Make rabbit_amqqueue:not_found_or_absent_dirty/1 visible For use in the stream plugin. Use bigger retention policy on max-age test Set coordinator timeout to 30s Handle coordinator unavailable error Handle operator policies as maps when checking if is applicable Add is_policy_applicable/2 to classic queues Ignore restart commands if the stream has been deleted It could happen that after termination some of the monitors are still up and trigger writer/replica restarts Policy support on stream queues Remove subscription events on stream coordinator Ensure old leaders are removed from monitors Introduce delay when retrying a failed phase Note that this ensures monitor is setup, there was a bug where no monitor was really started when re-trying the same phase Restart replicas after leader election instead of relying on old monitors Use timer for stream coordinator retries Fix stream stats for members/online Multiple fixes for replica monitoring and restart Ensure pending commands are appended at the end and re-run Ensure phase is reset with the state Remove duplicates from replica list Restart current phase on state_enter Remove unused import Ensure rabbit is running when checking for stream quorum Restart replicas Add a close/1 function to queue types So that we can get a chance of cleaning up resources if needed. Stream queues close their osiris logs at this point. fix compiler errors stream-queue: take retention into account When calculating ready messages metrics. Add osiris to the list of rabbit deps Retry restart of replicas Do not restart replicas or leaders after receiving a delete cluster command Add more logging to the stream coordinator Monitor subscribed processes on the stream coordinator Memory breakdown for stream queues Update quorum queue event formatter rabbit_msg_record fixes Refactor channel confirms Remove old unconfirmed_messages module that was designed to handle multiple queue fan in logic including all ha mirrors etc. Replaced with simpler rabbit_confirms module that handles the fan out and leaves any queue specific logic (such as confirms from mirrors) to the queue type implemention. Also this module has a dedicated test module. Which is nice. Backward compatibility with 3.8.x events Supports mixed version cluster upgrades Match specification when stream queue already exists Max age retention for stream queues Stop all replicas before starting leader election stream: disallow global qos remove IS_CLASSIC|QUORUM macros Ensure only classic queues are notified on channel down This also removes the delivering_queues map in the channel state as it should not be needed for this and just cause additional unecessary accounting. Polish AMQP 1.0/0.9.1 properties conversion Support byte in application properties, handle 1-bit representation for booleans. Use binary in header for long AMQP 1.0 ID Fix AMQP 1.0 to 0.9.1 conversion Fix test due to incorrect type Convert timestamp application properties to/from seconds AMQP 1.0 uses milliseconds for timestamp and AMQP 0.9.1 uses seconds, so conversion needed. Dialyzer fixes Handle all message-id types AMQP 1.0 is more liberal in it's allowed types of message-id and correlation-id - this adds headers to describe the type of the data in the message_id / correlation_id properties and also handles the case where the data cannot fit by again using headers. Resize stream coordinator cluster when broker configuration changes convert timestamp to and fro seconds user_id should be a binary message annotations keys need to be symbols stream-queue: default exchange and routing key As these won't be present for data written using the rabbitmq-stream plugin. Add exchange, routing key as message annotations To the AMQP 1.0 formatted data to enable roundtrip. Add osiris logging module config And update logging config test suite. Restart election when start of new leader fails The node might have just gone down so we need to try another one Only aux keeps track of phase now, as it might change if the leader election fails Stream coordinator refactor - all state is kept on the ra machine Ensure any ra cluster not a qq is not cleaned up Fixes to recovery and monitoring Add AMQP 1.0 common to dependencies Add rabbit_msg_record module To handle conversions into internal stream storage format. Use rabbitmq-common stream-queue branch Use SSH for osiris dependency Stream coordinator: delete replica Stream coordinator: add replica Stream coordinator: leader failover Stream coordinator: declare and delete Test consuming from a random offset Previous offsets should not be delivered to consumers Consume from stream replicas and multiple test fixes Use max-length-bytes and add new max-segment-size Use SSH for osiris dependency Basic cancel for stream queues Publish stream queues and settle/reject/requeue refactor Consume from stream queues Fix recovery Publish stream messages Add/delete stream replicas Use safe queue names Set retention policy for stream queues Required by the ctl command [#171207092] Stream queue delete queue fix missing callback impl Stream queue declare Queue type abstraction And use the implementing module as the value of the amqqueue record `type` field. This will allow for easy dispatch to the queue type implementation. Queue type abstraction Move queue declare into rabbit_queue_type Move queue delete into queue type implementation Queue type: dequeue/basic_get Move info inside queue type abstraction Move policy change into queue type interface Add purge to queue type Add recovery to the queue type interface Rename amqqueue quorum_nodes field To a more generic an extensible opaque queue type specific map. Fix tests and handle classic API response Fix HA queue confirm bug All mirrors need to be present as queue names. This introduces context linking allowing additional queue refs to be linked to a single "master" queue ref contining the actual queue context. Fix issue with events of deleted queues Also update queue type smoke test to use a cluster by default. correct default value of amqqueue getter Move classic queues further inside queue type interface why [TrackerId] Dialyzer fixes
2020-09-29 18:43:24 +08:00
QTState.
wait_for_confirms(Unconfirmed) ->
case sets:is_empty(Unconfirmed) of
true -> ok;
false ->
receive
{'$gen_cast', {queue_event, _QName, {confirm, Confirmed, _}}} ->
wait_for_confirms(
sets:subtract(
Unconfirmed, sets:from_list(Confirmed, [{version, 2}])))
after ?TIMEOUT ->
flush(),
exit(timeout_waiting_for_confirm)
end
end.
with_fresh_variable_queue(Fun, Mode) ->
Ref = make_ref(),
Me = self(),
%% Run in a separate process since rabbit_msg_store will send
%% bump_credit messages and we want to ignore them
spawn_link(fun() ->
QName = test_queue(),
ok = unin_empty_test_queue(QName),
VQ = variable_queue_init(test_amqqueue(QName, true), false),
S0 = variable_queue_status(VQ),
assert_props(S0, [{q1, 0}, {q2, 0},
{delta,
{delta, undefined, 0, 0, undefined}},
{q3, 0}, {q4, 0},
{len, 0}]),
VQ1 = set_queue_mode(Mode, VQ),
try
_ = rabbit_variable_queue:delete_and_terminate(
shutdown, Fun(VQ1, QName)),
Me ! Ref
catch
Type:Error:Stacktrace ->
Me ! {Ref, Type, Error, Stacktrace}
end
end),
receive
Ref -> ok;
{Ref, Type, Error, ST} -> exit({Type, Error, ST})
end,
passed.
set_queue_mode(Mode, VQ) ->
2022-09-05 19:37:33 +08:00
rabbit_variable_queue:set_queue_mode(Mode, VQ).
variable_queue_publish(IsPersistent, Count, VQ) ->
variable_queue_publish(IsPersistent, Count, fun (_N, P) -> P end, VQ).
variable_queue_publish(IsPersistent, Count, PropFun, VQ) ->
variable_queue_publish(IsPersistent, 1, Count, PropFun,
fun (_N) -> <<>> end, VQ).
variable_queue_publish(IsPersistent, Start, Count, PropFun, PayloadFun, VQ) ->
variable_queue_wait_for_shuffling_end(
lists:foldl(
fun (N, VQN) ->
Message Containers (#5077) This PR implements an approach for a "protocol (data format) agnostic core" where the format of the message isn't converted at point of reception. Currently all non AMQP 0.9.1 originating messages are converted into a AMQP 0.9.1 flavoured basic_message record before sent to a queue. If the messages are then consumed by the originating protocol they are converted back from AMQP 0.9.1. For some protocols such as MQTT 3.1 this isn't too expensive as MQTT is mostly a fairly easily mapped subset of AMQP 0.9.1 but for others such as AMQP 1.0 the conversions are awkward and in some cases lossy even if consuming from the originating protocol. This PR instead wraps all incoming messages in their originating form into a generic, extensible message container type (mc). The container module exposes an API to get common message details such as size and various properties (ttl, priority etc) directly from the source data type. Each protocol needs to implement the mc behaviour such that when a message originating form one protocol is consumed by another protocol we convert it to the target protocol at that point. The message container also contains annotations, dead letter records and other meta data we need to record during the lifetime of a message. The original protocol message is never modified unless it is consumed. This includes conversion modules to and from amqp, amqpl (AMQP 0.9.1) and mqtt. COMMIT HISTORY: * Refactor away from using the delivery{} record In many places including exchange types. This should make it easier to move towards using a message container type instead of basic_message. Add mc module and move direct replies outside of exchange Lots of changes incl classic queues Implement stream support incl amqp conversions simplify mc state record move mc.erl mc dlx stuff recent history exchange Make tracking work But doesn't take a protocol agnostic approach as we just convert everything into AMQP legacy and back. Might be good enough for now. Tracing as a whole may want a bit of a re-vamp at some point. tidy make quorum queue peek work by legacy conversion dead lettering fixes dead lettering fixes CMQ fixes rabbit_trace type fixes fixes fix Fix classic queue props test assertion fix feature flag and backwards compat Enable message_container feature flag in some SUITEs Dialyzer fixes fixes fix test fixes Various Manually update a gazelle generated file until a gazelle enhancement can be made https://github.com/rabbitmq/rules_erlang/issues/185 Add message_containers_SUITE to bazel and regen bazel files with gazelle from rules_erlang@main Simplify essential proprty access Such as durable, ttl and priority by extracting them into annotations at message container init time. Move type to remove dependenc on amqp10 stuff in mc.erl mostly because I don't know how to make bazel do the right thing add more stuff Refine routing header stuff wip Cosmetics Do not use "maybe" as type name as "maybe" is a keyword since OTP 25 which makes Erlang LS complain. * Dedup death queue names * Fix function clause crashes Fix failing tests in the MQTT shared_SUITE: A classic queue message ID can be undefined as set in https://github.com/rabbitmq/rabbitmq-server/blob/fbe79ff47b4edbc0fd95457e623d6593161ad198/deps/rabbit/src/rabbit_classic_queue_index_v2.erl#L1048 Fix failing tests in the MQTT shared_SUITE-mixed: When feature flag message_containers is disabled, the message is not an #mc{} record, but a #basic_message{} record. * Fix is_utf8_no_null crash Prior to this commit, the function crashed if invalid UTF-8 was provided, e.g.: ``` 1> rabbit_misc:is_valid_shortstr(<<"😇"/utf16>>). ** exception error: no function clause matching rabbit_misc:is_utf8_no_null(<<216,61,222,7>>) (rabbit_misc.erl, line 1481) ``` * Implement mqtt mc behaviour For now via amqp translation. This is still work in progress, but the following SUITEs pass: ``` make -C deps/rabbitmq_mqtt ct-shared t=[mqtt,v5,cluster_size_1] FULL=1 make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_1] FULL=1 ``` * Shorten mc file names Module name length matters because for each persistent message the #mc{} record is persisted to disk. ``` 1> iolist_size(term_to_iovec({mc, rabbit_mc_amqp_legacy})). 30 2> iolist_size(term_to_iovec({mc, mc_amqpl})). 17 ``` This commit renames the mc modules: ``` ag -l rabbit_mc_amqp_legacy | xargs sed -i 's/rabbit_mc_amqp_legacy/mc_amqpl/g' ag -l rabbit_mc_amqp | xargs sed -i 's/rabbit_mc_amqp/mc_amqp/g' ag -l rabbit_mqtt_mc | xargs sed -i 's/rabbit_mqtt_mc/mc_mqtt/g' ``` * mc: make deaths an annotation + fixes * Fix mc_mqtt protocol_state callback * Fix test will_delay_node_restart ``` make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_3]:will_delay_node_restart FULL=1 ``` * Bazel run gazelle * mix format rabbitmqctl.ex * Ensure ttl annotation is refelected in amqp legacy protocol state * Fix id access in message store * Fix rabbit_message_interceptor_SUITE * dializer fixes * Fix rabbit:rabbit_message_interceptor_SUITE-mixed set_annotation/3 should not result in duplicate keys * Fix MQTT shared_SUITE-mixed Up to 3.12 non-MQTT publishes were always QoS 1 regardless of delivery_mode. https://github.com/rabbitmq/rabbitmq-server/blob/75a953ce286a10aca910c098805a4f545989af38/deps/rabbitmq_mqtt/src/rabbit_mqtt_processor.erl#L2075-L2076 From now on, non-MQTT publishes are QoS 1 if durable. This makes more sense. The MQTT plugin must send a #basic_message{} to an old node that does not understand message containers. * Field content of 'v1_0.data' can be binary Fix ``` bazel test //deps/rabbitmq_mqtt:shared_SUITE-mixed \ --test_env FOCUS="-group [mqtt,v4,cluster_size_1] -case trace" \ -t- --test_sharding_strategy=disabled ``` * Remove route/2 and implement route/3 for all exchange types. This removes the route/2 callback from rabbit_exchange_type and makes route/3 mandatory instead. This is a breaking change and will require all implementations of exchange types to update their code, however this is necessary anyway for them to correctly handle the mc type. stream filtering fixes * Translate directly from MQTT to AMQP 0.9.1 * handle undecoded properties in mc_compat amqpl: put clause in right order recover death deatails from amqp data * Replace callback init_amqp with convert_from * Fix return value of lists:keyfind/3 * Translate directly from AMQP 0.9.1 to MQTT * Fix MQTT payload size MQTT payload can be a list when converted from AMQP 0.9.1 for example First conversions tests Plus some other conversion related fixes. bazel bazel translate amqp 1.0 null to undefined mc: property/2 and correlation_id/message_id return type tagged values. To ensure we can support a variety of types better. The type type tags are AMQP 1.0 flavoured. fix death recovery mc_mqtt: impl new api Add callbacks to allow protocols to compact data before storage And make readable if needing to query things repeatedly. bazel fix * more decoding * tracking mixed versions compat * mc: flip default of `durable` annotation to save some data. Assuming most messages are durable and that in memory messages suffer less from persistence overhead it makes sense for a non existent `durable` annotation to mean durable=true. * mc conversion tests and tidy up * mc make x_header unstrict again * amqpl: death record fixes * bazel * amqp -> amqpl conversion test * Fix crash in mc_amqp:size/1 Body can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0/ ct-system t=java ``` on branch native-amqp. * Fix crash in lists:flatten/1 Data can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp. * Fix crash in rabbit_writer Running test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp resulted in the following crash: ``` crasher: initial call: rabbit_writer:enter_mainloop/2 pid: <0.711.0> registered_name: [] exception error: bad argument in function size/1 called as size([<<0>>,<<"Sw">>,[<<160,2>>,<<"hi">>]]) *** argument 1: not tuple or binary in call from rabbit_binary_generator:build_content_frames/7 (rabbit_binary_generator.erl, line 89) in call from rabbit_binary_generator:build_simple_content_frames/4 (rabbit_binary_generator.erl, line 61) in call from rabbit_writer:assemble_frames/5 (rabbit_writer.erl, line 334) in call from rabbit_writer:internal_send_command_async/3 (rabbit_writer.erl, line 365) in call from rabbit_writer:handle_message/2 (rabbit_writer.erl, line 265) in call from rabbit_writer:handle_message/3 (rabbit_writer.erl, line 232) in call from rabbit_writer:mainloop1/2 (rabbit_writer.erl, line 223) ``` because #content.payload_fragments_rev is currently supposed to be a flat list of binaries instead of being an iolist. This commit fixes this crash inefficiently by calling iolist_to_binary/1. A better solution would be to allow AMQP legacy's #content.payload_fragments_rev to be an iolist. * Add accidentally deleted line back * mc: optimise mc_amqp internal format By removint the outer records for message and delivery annotations as well as application properties and footers. * mc: optimis mc_amqp map_add by using upsert * mc: refactoring and bug fixes * mc_SUITE routingheader assertions * mc remove serialize/1 callback as only used by amqp * mc_amqp: avoid returning a nested list from protocol_state * test and bug fix * move infer_type to mc_util * mc fixes and additiona assertions * Support headers exchange routing for MQTT messages When a headers exchange is bound to the MQTT topic exchange, routing will be performend based on both MQTT topic (by the topic exchange) and MQTT User Property (by the headers exchange). This combines the best worlds of both MQTT 5.0 and AMQP 0.9.1 and enables powerful routing topologies. When the User Property contains the same name multiple times, only the last name (and value) will be considered by the headers exchange. * Fix crash when sending from stream to amqpl When publishing a message via the stream protocol and consuming it via AMQP 0.9.1, the following crash occurred prior to this commit: ``` crasher: initial call: rabbit_channel:init/1 pid: <0.818.0> registered_name: [] exception exit: {{badmatch,undefined}, [{rabbit_channel,handle_deliver0,4, [{file,"rabbit_channel.erl"}, {line,2728}]}, {lists,foldl,3,[{file,"lists.erl"},{line,1594}]}, {rabbit_channel,handle_cast,2, [{file,"rabbit_channel.erl"}, {line,728}]}, {gen_server2,handle_msg,2, [{file,"gen_server2.erl"},{line,1056}]}, {proc_lib,wake_up,3, [{file,"proc_lib.erl"},{line,251}]}]} ``` This commit first gives `mc:init/3` the chance to set exchange and routing_keys annotations. If not set, `rabbit_stream_queue` will set these annotations assuming the message was originally published via the stream protocol. * Support consistent hash exchange routing for MQTT 5.0 When a consistent hash exchange is bound to the MQTT topic exchange, MQTT 5.0 messages can be routed to queues consistently based on the Correlation-Data in the PUBLISH packet. * Convert MQTT 5.0 User Property * to AMQP 0.9.1 headers * from AMQP 0.9.1 headers * to AMQP 1.0 application properties and message annotations * from AMQP 1.0 application properties and message annotations * Make use of Annotations in mc_mqtt:protocol_state/2 mc_mqtt:protocol_state/2 includes Annotations as parameter. It's cleaner to make use of these Annotations when computing the protocol state instead of relying on the caller (rabbitmq_mqtt_processor) to compute the protocol state. * Enforce AMQP 0.9.1 field name length limit The AMQP 0.9.1 spec prohibits field names longer than 128 characters. Therefore, when converting AMQP 1.0 message annotations, application properties or MQTT 5.0 User Property to AMQP 0.9.1 headers, drop any names longer than 128 characters. * Fix type specs Apply feedback from Michael Davis Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * Add mc_mqtt unit test suite Implement mc_mqtt:x_header/2 * Translate indicator that payload is UTF-8 encoded when converting between MQTT 5.0 and AMQP 1.0 * Translate single amqp-value section from AMQP 1.0 to MQTT Convert to a text representation, if possible, and indicate to MQTT client that the payload is UTF-8 encoded. This way, the MQTT client will be able to parse the payload. If conversion to text representation is not possible, encode the payload using the AMQP 1.0 type system and indiate the encoding via Content-Type message/vnd.rabbitmq.amqp. This Content-Type is not registered. Type "message" makes sense since it's a message. Vendor tree "vnd.rabbitmq.amqp" makes sense since merely subtype "amqp" is not registered. * Fix payload conversion * Translate Response Topic between MQTT and AMQP Translate MQTT 5.0 Response Topic to AMQP 1.0 reply-to address and vice versa. The Response Topic must be a UTF-8 encoded string. This commit re-uses the already defined RabbitMQ target addresses: ``` "/topic/" RK Publish to amq.topic with routing key RK "/exchange/" X "/" RK Publish to exchange X with routing key RK ``` By default, the MQTT topic exchange is configure dto be amq.topic using the 1st target address. When an operator modifies the mqtt.exchange, the 2nd target address is used. * Apply PR feedback and fix formatting Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * tidy up * Add MQTT message_containers test * consistent hash exchange: avoid amqp legacy conversion When hashing on a header value. * Avoid converting to amqp legacy when using exchange federation * Fix test flake * test and dialyzer fixes * dialyzer fix * Add MQTT protocol interoperability tests Test receiving from and sending to MQTT 5.0 and * AMQP 0.9.1 * AMQP 1.0 * STOMP * Streams * Regenerate portions of deps/rabbit/app.bzl with gazelle I'm not exactly sure how this happened, but gazell seems to have been run with an older version of the rules_erlang gazelle extension at some point. This caused generation of a structure that is no longer used. This commit updates the structure to the current pattern. * mc: refactoring * mc_amqpl: handle delivery annotations Just in case they are included. Also use iolist_to_iovec to create flat list of binaries when converting from amqp with amqp encoded payload. --------- Co-authored-by: David Ansari <david.ansari@gmx.de> Co-authored-by: Michael Davis <mcarsondavis@gmail.com> Co-authored-by: Rin Kuryloski <kuryloskip@vmware.com>
2023-08-31 18:27:13 +08:00
Msg = message(IsPersistent, PayloadFun, N),
rabbit_variable_queue:publish(
Message Containers (#5077) This PR implements an approach for a "protocol (data format) agnostic core" where the format of the message isn't converted at point of reception. Currently all non AMQP 0.9.1 originating messages are converted into a AMQP 0.9.1 flavoured basic_message record before sent to a queue. If the messages are then consumed by the originating protocol they are converted back from AMQP 0.9.1. For some protocols such as MQTT 3.1 this isn't too expensive as MQTT is mostly a fairly easily mapped subset of AMQP 0.9.1 but for others such as AMQP 1.0 the conversions are awkward and in some cases lossy even if consuming from the originating protocol. This PR instead wraps all incoming messages in their originating form into a generic, extensible message container type (mc). The container module exposes an API to get common message details such as size and various properties (ttl, priority etc) directly from the source data type. Each protocol needs to implement the mc behaviour such that when a message originating form one protocol is consumed by another protocol we convert it to the target protocol at that point. The message container also contains annotations, dead letter records and other meta data we need to record during the lifetime of a message. The original protocol message is never modified unless it is consumed. This includes conversion modules to and from amqp, amqpl (AMQP 0.9.1) and mqtt. COMMIT HISTORY: * Refactor away from using the delivery{} record In many places including exchange types. This should make it easier to move towards using a message container type instead of basic_message. Add mc module and move direct replies outside of exchange Lots of changes incl classic queues Implement stream support incl amqp conversions simplify mc state record move mc.erl mc dlx stuff recent history exchange Make tracking work But doesn't take a protocol agnostic approach as we just convert everything into AMQP legacy and back. Might be good enough for now. Tracing as a whole may want a bit of a re-vamp at some point. tidy make quorum queue peek work by legacy conversion dead lettering fixes dead lettering fixes CMQ fixes rabbit_trace type fixes fixes fix Fix classic queue props test assertion fix feature flag and backwards compat Enable message_container feature flag in some SUITEs Dialyzer fixes fixes fix test fixes Various Manually update a gazelle generated file until a gazelle enhancement can be made https://github.com/rabbitmq/rules_erlang/issues/185 Add message_containers_SUITE to bazel and regen bazel files with gazelle from rules_erlang@main Simplify essential proprty access Such as durable, ttl and priority by extracting them into annotations at message container init time. Move type to remove dependenc on amqp10 stuff in mc.erl mostly because I don't know how to make bazel do the right thing add more stuff Refine routing header stuff wip Cosmetics Do not use "maybe" as type name as "maybe" is a keyword since OTP 25 which makes Erlang LS complain. * Dedup death queue names * Fix function clause crashes Fix failing tests in the MQTT shared_SUITE: A classic queue message ID can be undefined as set in https://github.com/rabbitmq/rabbitmq-server/blob/fbe79ff47b4edbc0fd95457e623d6593161ad198/deps/rabbit/src/rabbit_classic_queue_index_v2.erl#L1048 Fix failing tests in the MQTT shared_SUITE-mixed: When feature flag message_containers is disabled, the message is not an #mc{} record, but a #basic_message{} record. * Fix is_utf8_no_null crash Prior to this commit, the function crashed if invalid UTF-8 was provided, e.g.: ``` 1> rabbit_misc:is_valid_shortstr(<<"😇"/utf16>>). ** exception error: no function clause matching rabbit_misc:is_utf8_no_null(<<216,61,222,7>>) (rabbit_misc.erl, line 1481) ``` * Implement mqtt mc behaviour For now via amqp translation. This is still work in progress, but the following SUITEs pass: ``` make -C deps/rabbitmq_mqtt ct-shared t=[mqtt,v5,cluster_size_1] FULL=1 make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_1] FULL=1 ``` * Shorten mc file names Module name length matters because for each persistent message the #mc{} record is persisted to disk. ``` 1> iolist_size(term_to_iovec({mc, rabbit_mc_amqp_legacy})). 30 2> iolist_size(term_to_iovec({mc, mc_amqpl})). 17 ``` This commit renames the mc modules: ``` ag -l rabbit_mc_amqp_legacy | xargs sed -i 's/rabbit_mc_amqp_legacy/mc_amqpl/g' ag -l rabbit_mc_amqp | xargs sed -i 's/rabbit_mc_amqp/mc_amqp/g' ag -l rabbit_mqtt_mc | xargs sed -i 's/rabbit_mqtt_mc/mc_mqtt/g' ``` * mc: make deaths an annotation + fixes * Fix mc_mqtt protocol_state callback * Fix test will_delay_node_restart ``` make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_3]:will_delay_node_restart FULL=1 ``` * Bazel run gazelle * mix format rabbitmqctl.ex * Ensure ttl annotation is refelected in amqp legacy protocol state * Fix id access in message store * Fix rabbit_message_interceptor_SUITE * dializer fixes * Fix rabbit:rabbit_message_interceptor_SUITE-mixed set_annotation/3 should not result in duplicate keys * Fix MQTT shared_SUITE-mixed Up to 3.12 non-MQTT publishes were always QoS 1 regardless of delivery_mode. https://github.com/rabbitmq/rabbitmq-server/blob/75a953ce286a10aca910c098805a4f545989af38/deps/rabbitmq_mqtt/src/rabbit_mqtt_processor.erl#L2075-L2076 From now on, non-MQTT publishes are QoS 1 if durable. This makes more sense. The MQTT plugin must send a #basic_message{} to an old node that does not understand message containers. * Field content of 'v1_0.data' can be binary Fix ``` bazel test //deps/rabbitmq_mqtt:shared_SUITE-mixed \ --test_env FOCUS="-group [mqtt,v4,cluster_size_1] -case trace" \ -t- --test_sharding_strategy=disabled ``` * Remove route/2 and implement route/3 for all exchange types. This removes the route/2 callback from rabbit_exchange_type and makes route/3 mandatory instead. This is a breaking change and will require all implementations of exchange types to update their code, however this is necessary anyway for them to correctly handle the mc type. stream filtering fixes * Translate directly from MQTT to AMQP 0.9.1 * handle undecoded properties in mc_compat amqpl: put clause in right order recover death deatails from amqp data * Replace callback init_amqp with convert_from * Fix return value of lists:keyfind/3 * Translate directly from AMQP 0.9.1 to MQTT * Fix MQTT payload size MQTT payload can be a list when converted from AMQP 0.9.1 for example First conversions tests Plus some other conversion related fixes. bazel bazel translate amqp 1.0 null to undefined mc: property/2 and correlation_id/message_id return type tagged values. To ensure we can support a variety of types better. The type type tags are AMQP 1.0 flavoured. fix death recovery mc_mqtt: impl new api Add callbacks to allow protocols to compact data before storage And make readable if needing to query things repeatedly. bazel fix * more decoding * tracking mixed versions compat * mc: flip default of `durable` annotation to save some data. Assuming most messages are durable and that in memory messages suffer less from persistence overhead it makes sense for a non existent `durable` annotation to mean durable=true. * mc conversion tests and tidy up * mc make x_header unstrict again * amqpl: death record fixes * bazel * amqp -> amqpl conversion test * Fix crash in mc_amqp:size/1 Body can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0/ ct-system t=java ``` on branch native-amqp. * Fix crash in lists:flatten/1 Data can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp. * Fix crash in rabbit_writer Running test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp resulted in the following crash: ``` crasher: initial call: rabbit_writer:enter_mainloop/2 pid: <0.711.0> registered_name: [] exception error: bad argument in function size/1 called as size([<<0>>,<<"Sw">>,[<<160,2>>,<<"hi">>]]) *** argument 1: not tuple or binary in call from rabbit_binary_generator:build_content_frames/7 (rabbit_binary_generator.erl, line 89) in call from rabbit_binary_generator:build_simple_content_frames/4 (rabbit_binary_generator.erl, line 61) in call from rabbit_writer:assemble_frames/5 (rabbit_writer.erl, line 334) in call from rabbit_writer:internal_send_command_async/3 (rabbit_writer.erl, line 365) in call from rabbit_writer:handle_message/2 (rabbit_writer.erl, line 265) in call from rabbit_writer:handle_message/3 (rabbit_writer.erl, line 232) in call from rabbit_writer:mainloop1/2 (rabbit_writer.erl, line 223) ``` because #content.payload_fragments_rev is currently supposed to be a flat list of binaries instead of being an iolist. This commit fixes this crash inefficiently by calling iolist_to_binary/1. A better solution would be to allow AMQP legacy's #content.payload_fragments_rev to be an iolist. * Add accidentally deleted line back * mc: optimise mc_amqp internal format By removint the outer records for message and delivery annotations as well as application properties and footers. * mc: optimis mc_amqp map_add by using upsert * mc: refactoring and bug fixes * mc_SUITE routingheader assertions * mc remove serialize/1 callback as only used by amqp * mc_amqp: avoid returning a nested list from protocol_state * test and bug fix * move infer_type to mc_util * mc fixes and additiona assertions * Support headers exchange routing for MQTT messages When a headers exchange is bound to the MQTT topic exchange, routing will be performend based on both MQTT topic (by the topic exchange) and MQTT User Property (by the headers exchange). This combines the best worlds of both MQTT 5.0 and AMQP 0.9.1 and enables powerful routing topologies. When the User Property contains the same name multiple times, only the last name (and value) will be considered by the headers exchange. * Fix crash when sending from stream to amqpl When publishing a message via the stream protocol and consuming it via AMQP 0.9.1, the following crash occurred prior to this commit: ``` crasher: initial call: rabbit_channel:init/1 pid: <0.818.0> registered_name: [] exception exit: {{badmatch,undefined}, [{rabbit_channel,handle_deliver0,4, [{file,"rabbit_channel.erl"}, {line,2728}]}, {lists,foldl,3,[{file,"lists.erl"},{line,1594}]}, {rabbit_channel,handle_cast,2, [{file,"rabbit_channel.erl"}, {line,728}]}, {gen_server2,handle_msg,2, [{file,"gen_server2.erl"},{line,1056}]}, {proc_lib,wake_up,3, [{file,"proc_lib.erl"},{line,251}]}]} ``` This commit first gives `mc:init/3` the chance to set exchange and routing_keys annotations. If not set, `rabbit_stream_queue` will set these annotations assuming the message was originally published via the stream protocol. * Support consistent hash exchange routing for MQTT 5.0 When a consistent hash exchange is bound to the MQTT topic exchange, MQTT 5.0 messages can be routed to queues consistently based on the Correlation-Data in the PUBLISH packet. * Convert MQTT 5.0 User Property * to AMQP 0.9.1 headers * from AMQP 0.9.1 headers * to AMQP 1.0 application properties and message annotations * from AMQP 1.0 application properties and message annotations * Make use of Annotations in mc_mqtt:protocol_state/2 mc_mqtt:protocol_state/2 includes Annotations as parameter. It's cleaner to make use of these Annotations when computing the protocol state instead of relying on the caller (rabbitmq_mqtt_processor) to compute the protocol state. * Enforce AMQP 0.9.1 field name length limit The AMQP 0.9.1 spec prohibits field names longer than 128 characters. Therefore, when converting AMQP 1.0 message annotations, application properties or MQTT 5.0 User Property to AMQP 0.9.1 headers, drop any names longer than 128 characters. * Fix type specs Apply feedback from Michael Davis Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * Add mc_mqtt unit test suite Implement mc_mqtt:x_header/2 * Translate indicator that payload is UTF-8 encoded when converting between MQTT 5.0 and AMQP 1.0 * Translate single amqp-value section from AMQP 1.0 to MQTT Convert to a text representation, if possible, and indicate to MQTT client that the payload is UTF-8 encoded. This way, the MQTT client will be able to parse the payload. If conversion to text representation is not possible, encode the payload using the AMQP 1.0 type system and indiate the encoding via Content-Type message/vnd.rabbitmq.amqp. This Content-Type is not registered. Type "message" makes sense since it's a message. Vendor tree "vnd.rabbitmq.amqp" makes sense since merely subtype "amqp" is not registered. * Fix payload conversion * Translate Response Topic between MQTT and AMQP Translate MQTT 5.0 Response Topic to AMQP 1.0 reply-to address and vice versa. The Response Topic must be a UTF-8 encoded string. This commit re-uses the already defined RabbitMQ target addresses: ``` "/topic/" RK Publish to amq.topic with routing key RK "/exchange/" X "/" RK Publish to exchange X with routing key RK ``` By default, the MQTT topic exchange is configure dto be amq.topic using the 1st target address. When an operator modifies the mqtt.exchange, the 2nd target address is used. * Apply PR feedback and fix formatting Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * tidy up * Add MQTT message_containers test * consistent hash exchange: avoid amqp legacy conversion When hashing on a header value. * Avoid converting to amqp legacy when using exchange federation * Fix test flake * test and dialyzer fixes * dialyzer fix * Add MQTT protocol interoperability tests Test receiving from and sending to MQTT 5.0 and * AMQP 0.9.1 * AMQP 1.0 * STOMP * Streams * Regenerate portions of deps/rabbit/app.bzl with gazelle I'm not exactly sure how this happened, but gazell seems to have been run with an older version of the rules_erlang gazelle extension at some point. This caused generation of a structure that is no longer used. This commit updates the structure to the current pattern. * mc: refactoring * mc_amqpl: handle delivery annotations Just in case they are included. Also use iolist_to_iovec to create flat list of binaries when converting from amqp with amqp encoded payload. --------- Co-authored-by: David Ansari <david.ansari@gmx.de> Co-authored-by: Michael Davis <mcarsondavis@gmail.com> Co-authored-by: Rin Kuryloski <kuryloskip@vmware.com>
2023-08-31 18:27:13 +08:00
Msg,
PropFun(N, #message_properties{size = 10}),
2023-10-04 18:11:54 +08:00
false, self(), VQN)
end, VQ, lists:seq(Start, Start + Count - 1))).
variable_queue_fetch(Count, IsPersistent, IsDelivered, Len, VQ) ->
lists:foldl(fun (N, {VQN, AckTagsAcc}) ->
Rem = Len - N,
Message Containers (#5077) This PR implements an approach for a "protocol (data format) agnostic core" where the format of the message isn't converted at point of reception. Currently all non AMQP 0.9.1 originating messages are converted into a AMQP 0.9.1 flavoured basic_message record before sent to a queue. If the messages are then consumed by the originating protocol they are converted back from AMQP 0.9.1. For some protocols such as MQTT 3.1 this isn't too expensive as MQTT is mostly a fairly easily mapped subset of AMQP 0.9.1 but for others such as AMQP 1.0 the conversions are awkward and in some cases lossy even if consuming from the originating protocol. This PR instead wraps all incoming messages in their originating form into a generic, extensible message container type (mc). The container module exposes an API to get common message details such as size and various properties (ttl, priority etc) directly from the source data type. Each protocol needs to implement the mc behaviour such that when a message originating form one protocol is consumed by another protocol we convert it to the target protocol at that point. The message container also contains annotations, dead letter records and other meta data we need to record during the lifetime of a message. The original protocol message is never modified unless it is consumed. This includes conversion modules to and from amqp, amqpl (AMQP 0.9.1) and mqtt. COMMIT HISTORY: * Refactor away from using the delivery{} record In many places including exchange types. This should make it easier to move towards using a message container type instead of basic_message. Add mc module and move direct replies outside of exchange Lots of changes incl classic queues Implement stream support incl amqp conversions simplify mc state record move mc.erl mc dlx stuff recent history exchange Make tracking work But doesn't take a protocol agnostic approach as we just convert everything into AMQP legacy and back. Might be good enough for now. Tracing as a whole may want a bit of a re-vamp at some point. tidy make quorum queue peek work by legacy conversion dead lettering fixes dead lettering fixes CMQ fixes rabbit_trace type fixes fixes fix Fix classic queue props test assertion fix feature flag and backwards compat Enable message_container feature flag in some SUITEs Dialyzer fixes fixes fix test fixes Various Manually update a gazelle generated file until a gazelle enhancement can be made https://github.com/rabbitmq/rules_erlang/issues/185 Add message_containers_SUITE to bazel and regen bazel files with gazelle from rules_erlang@main Simplify essential proprty access Such as durable, ttl and priority by extracting them into annotations at message container init time. Move type to remove dependenc on amqp10 stuff in mc.erl mostly because I don't know how to make bazel do the right thing add more stuff Refine routing header stuff wip Cosmetics Do not use "maybe" as type name as "maybe" is a keyword since OTP 25 which makes Erlang LS complain. * Dedup death queue names * Fix function clause crashes Fix failing tests in the MQTT shared_SUITE: A classic queue message ID can be undefined as set in https://github.com/rabbitmq/rabbitmq-server/blob/fbe79ff47b4edbc0fd95457e623d6593161ad198/deps/rabbit/src/rabbit_classic_queue_index_v2.erl#L1048 Fix failing tests in the MQTT shared_SUITE-mixed: When feature flag message_containers is disabled, the message is not an #mc{} record, but a #basic_message{} record. * Fix is_utf8_no_null crash Prior to this commit, the function crashed if invalid UTF-8 was provided, e.g.: ``` 1> rabbit_misc:is_valid_shortstr(<<"😇"/utf16>>). ** exception error: no function clause matching rabbit_misc:is_utf8_no_null(<<216,61,222,7>>) (rabbit_misc.erl, line 1481) ``` * Implement mqtt mc behaviour For now via amqp translation. This is still work in progress, but the following SUITEs pass: ``` make -C deps/rabbitmq_mqtt ct-shared t=[mqtt,v5,cluster_size_1] FULL=1 make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_1] FULL=1 ``` * Shorten mc file names Module name length matters because for each persistent message the #mc{} record is persisted to disk. ``` 1> iolist_size(term_to_iovec({mc, rabbit_mc_amqp_legacy})). 30 2> iolist_size(term_to_iovec({mc, mc_amqpl})). 17 ``` This commit renames the mc modules: ``` ag -l rabbit_mc_amqp_legacy | xargs sed -i 's/rabbit_mc_amqp_legacy/mc_amqpl/g' ag -l rabbit_mc_amqp | xargs sed -i 's/rabbit_mc_amqp/mc_amqp/g' ag -l rabbit_mqtt_mc | xargs sed -i 's/rabbit_mqtt_mc/mc_mqtt/g' ``` * mc: make deaths an annotation + fixes * Fix mc_mqtt protocol_state callback * Fix test will_delay_node_restart ``` make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_3]:will_delay_node_restart FULL=1 ``` * Bazel run gazelle * mix format rabbitmqctl.ex * Ensure ttl annotation is refelected in amqp legacy protocol state * Fix id access in message store * Fix rabbit_message_interceptor_SUITE * dializer fixes * Fix rabbit:rabbit_message_interceptor_SUITE-mixed set_annotation/3 should not result in duplicate keys * Fix MQTT shared_SUITE-mixed Up to 3.12 non-MQTT publishes were always QoS 1 regardless of delivery_mode. https://github.com/rabbitmq/rabbitmq-server/blob/75a953ce286a10aca910c098805a4f545989af38/deps/rabbitmq_mqtt/src/rabbit_mqtt_processor.erl#L2075-L2076 From now on, non-MQTT publishes are QoS 1 if durable. This makes more sense. The MQTT plugin must send a #basic_message{} to an old node that does not understand message containers. * Field content of 'v1_0.data' can be binary Fix ``` bazel test //deps/rabbitmq_mqtt:shared_SUITE-mixed \ --test_env FOCUS="-group [mqtt,v4,cluster_size_1] -case trace" \ -t- --test_sharding_strategy=disabled ``` * Remove route/2 and implement route/3 for all exchange types. This removes the route/2 callback from rabbit_exchange_type and makes route/3 mandatory instead. This is a breaking change and will require all implementations of exchange types to update their code, however this is necessary anyway for them to correctly handle the mc type. stream filtering fixes * Translate directly from MQTT to AMQP 0.9.1 * handle undecoded properties in mc_compat amqpl: put clause in right order recover death deatails from amqp data * Replace callback init_amqp with convert_from * Fix return value of lists:keyfind/3 * Translate directly from AMQP 0.9.1 to MQTT * Fix MQTT payload size MQTT payload can be a list when converted from AMQP 0.9.1 for example First conversions tests Plus some other conversion related fixes. bazel bazel translate amqp 1.0 null to undefined mc: property/2 and correlation_id/message_id return type tagged values. To ensure we can support a variety of types better. The type type tags are AMQP 1.0 flavoured. fix death recovery mc_mqtt: impl new api Add callbacks to allow protocols to compact data before storage And make readable if needing to query things repeatedly. bazel fix * more decoding * tracking mixed versions compat * mc: flip default of `durable` annotation to save some data. Assuming most messages are durable and that in memory messages suffer less from persistence overhead it makes sense for a non existent `durable` annotation to mean durable=true. * mc conversion tests and tidy up * mc make x_header unstrict again * amqpl: death record fixes * bazel * amqp -> amqpl conversion test * Fix crash in mc_amqp:size/1 Body can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0/ ct-system t=java ``` on branch native-amqp. * Fix crash in lists:flatten/1 Data can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp. * Fix crash in rabbit_writer Running test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp resulted in the following crash: ``` crasher: initial call: rabbit_writer:enter_mainloop/2 pid: <0.711.0> registered_name: [] exception error: bad argument in function size/1 called as size([<<0>>,<<"Sw">>,[<<160,2>>,<<"hi">>]]) *** argument 1: not tuple or binary in call from rabbit_binary_generator:build_content_frames/7 (rabbit_binary_generator.erl, line 89) in call from rabbit_binary_generator:build_simple_content_frames/4 (rabbit_binary_generator.erl, line 61) in call from rabbit_writer:assemble_frames/5 (rabbit_writer.erl, line 334) in call from rabbit_writer:internal_send_command_async/3 (rabbit_writer.erl, line 365) in call from rabbit_writer:handle_message/2 (rabbit_writer.erl, line 265) in call from rabbit_writer:handle_message/3 (rabbit_writer.erl, line 232) in call from rabbit_writer:mainloop1/2 (rabbit_writer.erl, line 223) ``` because #content.payload_fragments_rev is currently supposed to be a flat list of binaries instead of being an iolist. This commit fixes this crash inefficiently by calling iolist_to_binary/1. A better solution would be to allow AMQP legacy's #content.payload_fragments_rev to be an iolist. * Add accidentally deleted line back * mc: optimise mc_amqp internal format By removint the outer records for message and delivery annotations as well as application properties and footers. * mc: optimis mc_amqp map_add by using upsert * mc: refactoring and bug fixes * mc_SUITE routingheader assertions * mc remove serialize/1 callback as only used by amqp * mc_amqp: avoid returning a nested list from protocol_state * test and bug fix * move infer_type to mc_util * mc fixes and additiona assertions * Support headers exchange routing for MQTT messages When a headers exchange is bound to the MQTT topic exchange, routing will be performend based on both MQTT topic (by the topic exchange) and MQTT User Property (by the headers exchange). This combines the best worlds of both MQTT 5.0 and AMQP 0.9.1 and enables powerful routing topologies. When the User Property contains the same name multiple times, only the last name (and value) will be considered by the headers exchange. * Fix crash when sending from stream to amqpl When publishing a message via the stream protocol and consuming it via AMQP 0.9.1, the following crash occurred prior to this commit: ``` crasher: initial call: rabbit_channel:init/1 pid: <0.818.0> registered_name: [] exception exit: {{badmatch,undefined}, [{rabbit_channel,handle_deliver0,4, [{file,"rabbit_channel.erl"}, {line,2728}]}, {lists,foldl,3,[{file,"lists.erl"},{line,1594}]}, {rabbit_channel,handle_cast,2, [{file,"rabbit_channel.erl"}, {line,728}]}, {gen_server2,handle_msg,2, [{file,"gen_server2.erl"},{line,1056}]}, {proc_lib,wake_up,3, [{file,"proc_lib.erl"},{line,251}]}]} ``` This commit first gives `mc:init/3` the chance to set exchange and routing_keys annotations. If not set, `rabbit_stream_queue` will set these annotations assuming the message was originally published via the stream protocol. * Support consistent hash exchange routing for MQTT 5.0 When a consistent hash exchange is bound to the MQTT topic exchange, MQTT 5.0 messages can be routed to queues consistently based on the Correlation-Data in the PUBLISH packet. * Convert MQTT 5.0 User Property * to AMQP 0.9.1 headers * from AMQP 0.9.1 headers * to AMQP 1.0 application properties and message annotations * from AMQP 1.0 application properties and message annotations * Make use of Annotations in mc_mqtt:protocol_state/2 mc_mqtt:protocol_state/2 includes Annotations as parameter. It's cleaner to make use of these Annotations when computing the protocol state instead of relying on the caller (rabbitmq_mqtt_processor) to compute the protocol state. * Enforce AMQP 0.9.1 field name length limit The AMQP 0.9.1 spec prohibits field names longer than 128 characters. Therefore, when converting AMQP 1.0 message annotations, application properties or MQTT 5.0 User Property to AMQP 0.9.1 headers, drop any names longer than 128 characters. * Fix type specs Apply feedback from Michael Davis Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * Add mc_mqtt unit test suite Implement mc_mqtt:x_header/2 * Translate indicator that payload is UTF-8 encoded when converting between MQTT 5.0 and AMQP 1.0 * Translate single amqp-value section from AMQP 1.0 to MQTT Convert to a text representation, if possible, and indicate to MQTT client that the payload is UTF-8 encoded. This way, the MQTT client will be able to parse the payload. If conversion to text representation is not possible, encode the payload using the AMQP 1.0 type system and indiate the encoding via Content-Type message/vnd.rabbitmq.amqp. This Content-Type is not registered. Type "message" makes sense since it's a message. Vendor tree "vnd.rabbitmq.amqp" makes sense since merely subtype "amqp" is not registered. * Fix payload conversion * Translate Response Topic between MQTT and AMQP Translate MQTT 5.0 Response Topic to AMQP 1.0 reply-to address and vice versa. The Response Topic must be a UTF-8 encoded string. This commit re-uses the already defined RabbitMQ target addresses: ``` "/topic/" RK Publish to amq.topic with routing key RK "/exchange/" X "/" RK Publish to exchange X with routing key RK ``` By default, the MQTT topic exchange is configure dto be amq.topic using the 1st target address. When an operator modifies the mqtt.exchange, the 2nd target address is used. * Apply PR feedback and fix formatting Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * tidy up * Add MQTT message_containers test * consistent hash exchange: avoid amqp legacy conversion When hashing on a header value. * Avoid converting to amqp legacy when using exchange federation * Fix test flake * test and dialyzer fixes * dialyzer fix * Add MQTT protocol interoperability tests Test receiving from and sending to MQTT 5.0 and * AMQP 0.9.1 * AMQP 1.0 * STOMP * Streams * Regenerate portions of deps/rabbit/app.bzl with gazelle I'm not exactly sure how this happened, but gazell seems to have been run with an older version of the rules_erlang gazelle extension at some point. This caused generation of a structure that is no longer used. This commit updates the structure to the current pattern. * mc: refactoring * mc_amqpl: handle delivery annotations Just in case they are included. Also use iolist_to_iovec to create flat list of binaries when converting from amqp with amqp encoded payload. --------- Co-authored-by: David Ansari <david.ansari@gmx.de> Co-authored-by: Michael Davis <mcarsondavis@gmail.com> Co-authored-by: Rin Kuryloski <kuryloskip@vmware.com>
2023-08-31 18:27:13 +08:00
{{Msg, IsDelivered, AckTagN}, VQM} =
rabbit_variable_queue:fetch(true, VQN),
Message Containers (#5077) This PR implements an approach for a "protocol (data format) agnostic core" where the format of the message isn't converted at point of reception. Currently all non AMQP 0.9.1 originating messages are converted into a AMQP 0.9.1 flavoured basic_message record before sent to a queue. If the messages are then consumed by the originating protocol they are converted back from AMQP 0.9.1. For some protocols such as MQTT 3.1 this isn't too expensive as MQTT is mostly a fairly easily mapped subset of AMQP 0.9.1 but for others such as AMQP 1.0 the conversions are awkward and in some cases lossy even if consuming from the originating protocol. This PR instead wraps all incoming messages in their originating form into a generic, extensible message container type (mc). The container module exposes an API to get common message details such as size and various properties (ttl, priority etc) directly from the source data type. Each protocol needs to implement the mc behaviour such that when a message originating form one protocol is consumed by another protocol we convert it to the target protocol at that point. The message container also contains annotations, dead letter records and other meta data we need to record during the lifetime of a message. The original protocol message is never modified unless it is consumed. This includes conversion modules to and from amqp, amqpl (AMQP 0.9.1) and mqtt. COMMIT HISTORY: * Refactor away from using the delivery{} record In many places including exchange types. This should make it easier to move towards using a message container type instead of basic_message. Add mc module and move direct replies outside of exchange Lots of changes incl classic queues Implement stream support incl amqp conversions simplify mc state record move mc.erl mc dlx stuff recent history exchange Make tracking work But doesn't take a protocol agnostic approach as we just convert everything into AMQP legacy and back. Might be good enough for now. Tracing as a whole may want a bit of a re-vamp at some point. tidy make quorum queue peek work by legacy conversion dead lettering fixes dead lettering fixes CMQ fixes rabbit_trace type fixes fixes fix Fix classic queue props test assertion fix feature flag and backwards compat Enable message_container feature flag in some SUITEs Dialyzer fixes fixes fix test fixes Various Manually update a gazelle generated file until a gazelle enhancement can be made https://github.com/rabbitmq/rules_erlang/issues/185 Add message_containers_SUITE to bazel and regen bazel files with gazelle from rules_erlang@main Simplify essential proprty access Such as durable, ttl and priority by extracting them into annotations at message container init time. Move type to remove dependenc on amqp10 stuff in mc.erl mostly because I don't know how to make bazel do the right thing add more stuff Refine routing header stuff wip Cosmetics Do not use "maybe" as type name as "maybe" is a keyword since OTP 25 which makes Erlang LS complain. * Dedup death queue names * Fix function clause crashes Fix failing tests in the MQTT shared_SUITE: A classic queue message ID can be undefined as set in https://github.com/rabbitmq/rabbitmq-server/blob/fbe79ff47b4edbc0fd95457e623d6593161ad198/deps/rabbit/src/rabbit_classic_queue_index_v2.erl#L1048 Fix failing tests in the MQTT shared_SUITE-mixed: When feature flag message_containers is disabled, the message is not an #mc{} record, but a #basic_message{} record. * Fix is_utf8_no_null crash Prior to this commit, the function crashed if invalid UTF-8 was provided, e.g.: ``` 1> rabbit_misc:is_valid_shortstr(<<"😇"/utf16>>). ** exception error: no function clause matching rabbit_misc:is_utf8_no_null(<<216,61,222,7>>) (rabbit_misc.erl, line 1481) ``` * Implement mqtt mc behaviour For now via amqp translation. This is still work in progress, but the following SUITEs pass: ``` make -C deps/rabbitmq_mqtt ct-shared t=[mqtt,v5,cluster_size_1] FULL=1 make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_1] FULL=1 ``` * Shorten mc file names Module name length matters because for each persistent message the #mc{} record is persisted to disk. ``` 1> iolist_size(term_to_iovec({mc, rabbit_mc_amqp_legacy})). 30 2> iolist_size(term_to_iovec({mc, mc_amqpl})). 17 ``` This commit renames the mc modules: ``` ag -l rabbit_mc_amqp_legacy | xargs sed -i 's/rabbit_mc_amqp_legacy/mc_amqpl/g' ag -l rabbit_mc_amqp | xargs sed -i 's/rabbit_mc_amqp/mc_amqp/g' ag -l rabbit_mqtt_mc | xargs sed -i 's/rabbit_mqtt_mc/mc_mqtt/g' ``` * mc: make deaths an annotation + fixes * Fix mc_mqtt protocol_state callback * Fix test will_delay_node_restart ``` make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_3]:will_delay_node_restart FULL=1 ``` * Bazel run gazelle * mix format rabbitmqctl.ex * Ensure ttl annotation is refelected in amqp legacy protocol state * Fix id access in message store * Fix rabbit_message_interceptor_SUITE * dializer fixes * Fix rabbit:rabbit_message_interceptor_SUITE-mixed set_annotation/3 should not result in duplicate keys * Fix MQTT shared_SUITE-mixed Up to 3.12 non-MQTT publishes were always QoS 1 regardless of delivery_mode. https://github.com/rabbitmq/rabbitmq-server/blob/75a953ce286a10aca910c098805a4f545989af38/deps/rabbitmq_mqtt/src/rabbit_mqtt_processor.erl#L2075-L2076 From now on, non-MQTT publishes are QoS 1 if durable. This makes more sense. The MQTT plugin must send a #basic_message{} to an old node that does not understand message containers. * Field content of 'v1_0.data' can be binary Fix ``` bazel test //deps/rabbitmq_mqtt:shared_SUITE-mixed \ --test_env FOCUS="-group [mqtt,v4,cluster_size_1] -case trace" \ -t- --test_sharding_strategy=disabled ``` * Remove route/2 and implement route/3 for all exchange types. This removes the route/2 callback from rabbit_exchange_type and makes route/3 mandatory instead. This is a breaking change and will require all implementations of exchange types to update their code, however this is necessary anyway for them to correctly handle the mc type. stream filtering fixes * Translate directly from MQTT to AMQP 0.9.1 * handle undecoded properties in mc_compat amqpl: put clause in right order recover death deatails from amqp data * Replace callback init_amqp with convert_from * Fix return value of lists:keyfind/3 * Translate directly from AMQP 0.9.1 to MQTT * Fix MQTT payload size MQTT payload can be a list when converted from AMQP 0.9.1 for example First conversions tests Plus some other conversion related fixes. bazel bazel translate amqp 1.0 null to undefined mc: property/2 and correlation_id/message_id return type tagged values. To ensure we can support a variety of types better. The type type tags are AMQP 1.0 flavoured. fix death recovery mc_mqtt: impl new api Add callbacks to allow protocols to compact data before storage And make readable if needing to query things repeatedly. bazel fix * more decoding * tracking mixed versions compat * mc: flip default of `durable` annotation to save some data. Assuming most messages are durable and that in memory messages suffer less from persistence overhead it makes sense for a non existent `durable` annotation to mean durable=true. * mc conversion tests and tidy up * mc make x_header unstrict again * amqpl: death record fixes * bazel * amqp -> amqpl conversion test * Fix crash in mc_amqp:size/1 Body can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0/ ct-system t=java ``` on branch native-amqp. * Fix crash in lists:flatten/1 Data can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp. * Fix crash in rabbit_writer Running test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp resulted in the following crash: ``` crasher: initial call: rabbit_writer:enter_mainloop/2 pid: <0.711.0> registered_name: [] exception error: bad argument in function size/1 called as size([<<0>>,<<"Sw">>,[<<160,2>>,<<"hi">>]]) *** argument 1: not tuple or binary in call from rabbit_binary_generator:build_content_frames/7 (rabbit_binary_generator.erl, line 89) in call from rabbit_binary_generator:build_simple_content_frames/4 (rabbit_binary_generator.erl, line 61) in call from rabbit_writer:assemble_frames/5 (rabbit_writer.erl, line 334) in call from rabbit_writer:internal_send_command_async/3 (rabbit_writer.erl, line 365) in call from rabbit_writer:handle_message/2 (rabbit_writer.erl, line 265) in call from rabbit_writer:handle_message/3 (rabbit_writer.erl, line 232) in call from rabbit_writer:mainloop1/2 (rabbit_writer.erl, line 223) ``` because #content.payload_fragments_rev is currently supposed to be a flat list of binaries instead of being an iolist. This commit fixes this crash inefficiently by calling iolist_to_binary/1. A better solution would be to allow AMQP legacy's #content.payload_fragments_rev to be an iolist. * Add accidentally deleted line back * mc: optimise mc_amqp internal format By removint the outer records for message and delivery annotations as well as application properties and footers. * mc: optimis mc_amqp map_add by using upsert * mc: refactoring and bug fixes * mc_SUITE routingheader assertions * mc remove serialize/1 callback as only used by amqp * mc_amqp: avoid returning a nested list from protocol_state * test and bug fix * move infer_type to mc_util * mc fixes and additiona assertions * Support headers exchange routing for MQTT messages When a headers exchange is bound to the MQTT topic exchange, routing will be performend based on both MQTT topic (by the topic exchange) and MQTT User Property (by the headers exchange). This combines the best worlds of both MQTT 5.0 and AMQP 0.9.1 and enables powerful routing topologies. When the User Property contains the same name multiple times, only the last name (and value) will be considered by the headers exchange. * Fix crash when sending from stream to amqpl When publishing a message via the stream protocol and consuming it via AMQP 0.9.1, the following crash occurred prior to this commit: ``` crasher: initial call: rabbit_channel:init/1 pid: <0.818.0> registered_name: [] exception exit: {{badmatch,undefined}, [{rabbit_channel,handle_deliver0,4, [{file,"rabbit_channel.erl"}, {line,2728}]}, {lists,foldl,3,[{file,"lists.erl"},{line,1594}]}, {rabbit_channel,handle_cast,2, [{file,"rabbit_channel.erl"}, {line,728}]}, {gen_server2,handle_msg,2, [{file,"gen_server2.erl"},{line,1056}]}, {proc_lib,wake_up,3, [{file,"proc_lib.erl"},{line,251}]}]} ``` This commit first gives `mc:init/3` the chance to set exchange and routing_keys annotations. If not set, `rabbit_stream_queue` will set these annotations assuming the message was originally published via the stream protocol. * Support consistent hash exchange routing for MQTT 5.0 When a consistent hash exchange is bound to the MQTT topic exchange, MQTT 5.0 messages can be routed to queues consistently based on the Correlation-Data in the PUBLISH packet. * Convert MQTT 5.0 User Property * to AMQP 0.9.1 headers * from AMQP 0.9.1 headers * to AMQP 1.0 application properties and message annotations * from AMQP 1.0 application properties and message annotations * Make use of Annotations in mc_mqtt:protocol_state/2 mc_mqtt:protocol_state/2 includes Annotations as parameter. It's cleaner to make use of these Annotations when computing the protocol state instead of relying on the caller (rabbitmq_mqtt_processor) to compute the protocol state. * Enforce AMQP 0.9.1 field name length limit The AMQP 0.9.1 spec prohibits field names longer than 128 characters. Therefore, when converting AMQP 1.0 message annotations, application properties or MQTT 5.0 User Property to AMQP 0.9.1 headers, drop any names longer than 128 characters. * Fix type specs Apply feedback from Michael Davis Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * Add mc_mqtt unit test suite Implement mc_mqtt:x_header/2 * Translate indicator that payload is UTF-8 encoded when converting between MQTT 5.0 and AMQP 1.0 * Translate single amqp-value section from AMQP 1.0 to MQTT Convert to a text representation, if possible, and indicate to MQTT client that the payload is UTF-8 encoded. This way, the MQTT client will be able to parse the payload. If conversion to text representation is not possible, encode the payload using the AMQP 1.0 type system and indiate the encoding via Content-Type message/vnd.rabbitmq.amqp. This Content-Type is not registered. Type "message" makes sense since it's a message. Vendor tree "vnd.rabbitmq.amqp" makes sense since merely subtype "amqp" is not registered. * Fix payload conversion * Translate Response Topic between MQTT and AMQP Translate MQTT 5.0 Response Topic to AMQP 1.0 reply-to address and vice versa. The Response Topic must be a UTF-8 encoded string. This commit re-uses the already defined RabbitMQ target addresses: ``` "/topic/" RK Publish to amq.topic with routing key RK "/exchange/" X "/" RK Publish to exchange X with routing key RK ``` By default, the MQTT topic exchange is configure dto be amq.topic using the 1st target address. When an operator modifies the mqtt.exchange, the 2nd target address is used. * Apply PR feedback and fix formatting Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * tidy up * Add MQTT message_containers test * consistent hash exchange: avoid amqp legacy conversion When hashing on a header value. * Avoid converting to amqp legacy when using exchange federation * Fix test flake * test and dialyzer fixes * dialyzer fix * Add MQTT protocol interoperability tests Test receiving from and sending to MQTT 5.0 and * AMQP 0.9.1 * AMQP 1.0 * STOMP * Streams * Regenerate portions of deps/rabbit/app.bzl with gazelle I'm not exactly sure how this happened, but gazell seems to have been run with an older version of the rules_erlang gazelle extension at some point. This caused generation of a structure that is no longer used. This commit updates the structure to the current pattern. * mc: refactoring * mc_amqpl: handle delivery annotations Just in case they are included. Also use iolist_to_iovec to create flat list of binaries when converting from amqp with amqp encoded payload. --------- Co-authored-by: David Ansari <david.ansari@gmx.de> Co-authored-by: Michael Davis <mcarsondavis@gmail.com> Co-authored-by: Rin Kuryloski <kuryloskip@vmware.com>
2023-08-31 18:27:13 +08:00
IsPersistent = mc:is_persistent(Msg),
Rem = rabbit_variable_queue:len(VQM),
{VQM, [AckTagN | AckTagsAcc]}
end, {VQ, []}, lists:seq(1, Count)).
test_amqqueue(QName, Durable) ->
Make #amqqueue{} a private record + add a feature flag The #amqqueue{} record is now isolated in the `amqqueue` module: all accesses and modifications to the record are made through this helper. The type `rabbit_types:amqqueue()` is now `amqqueue:amqqueue()`. `amqqueue` knows about the new #amqqueue{} record introduced with quorum queues. If the actual record is the old one (as defined in RabbitMQ 3.7.x), it calls `amqqueue_v1` which knows about the old definition. When it needs to produce an #amqqueue{} record, it verifies if it is allowed to, using the new Feature flags subsystem and the `quorum_queue` feature flag specifically, and proceeds. If an old format #amqqueue{} is required, the creation is deferred to `amqqueue_v1`. The new modules come with a couple headers: they provide macros to replace pattern matching with `when` conditions. The `amqqueue_v*.hrl` headers are generated using the `macros()` functions in `amqqueue` and `amqqueue_v1` modules. Some operations now depend on the state of the `quorum_queue` feature flag. In particular, creating a quorum queue is denied if the feature flag is disabled. In the process, the following modules were moved from rabbitmq-common to rabbitmq-server: * rabbit_backing_queue * rabbit_queue_master_locator Likewise, the few functions were copied from `rabbit_misc` to `rabbit_amqqueue` because they depend on `amqqueue`. They don't really need to live in rabbitmq-common and they prevent further progress with the feature flags: feature flags need informations about the cluster. The `quorum_queue` feature flag comes with a migration function which is responsible of converting the #amqqueue{} records in Mnesia tables `rabbit_queue` and `rabbit_durable_queue`. Therefore, the upgrade function which did this is removed in this commit. [#159298729]
2018-10-11 18:12:39 +08:00
rabbit_amqqueue:pseudo_queue(QName, self(), Durable).
assert_prop(List, Prop, Value) ->
case proplists:get_value(Prop, List)of
Value -> ok;
_ -> {exit, Prop, exp, Value, List}
end.
assert_props(List, PropVals) ->
Res = [assert_prop(List, Prop, Value) || {Prop, Value} <- PropVals],
case lists:usort(Res) of
[ok] -> ok;
Error -> error(Error -- [ok])
end.
publish_fetch_and_ack(0, _Len, VQ0) ->
VQ0;
publish_fetch_and_ack(N, Len, VQ0) ->
VQ1 = variable_queue_publish(false, 1, VQ0),
{{_Msg, false, AckTag}, VQ2} = rabbit_variable_queue:fetch(true, VQ1),
Len = rabbit_variable_queue:len(VQ2),
{_Guids, VQ3} = rabbit_variable_queue:ack([AckTag], VQ2),
publish_fetch_and_ack(N-1, Len, VQ3).
variable_queue_status(VQ) ->
Keys = rabbit_backing_queue:info_keys() -- [backing_queue_status],
[{K, rabbit_variable_queue:info(K, VQ)} || K <- Keys] ++
rabbit_variable_queue:info(backing_queue_status, VQ).
variable_queue_wait_for_shuffling_end(VQ) ->
case credit_flow:blocked() of
false -> VQ;
Stream Queue This is an aggregated commit of all changes related to the initial implementation of queue types and on top of that the stream queue type. The varios commit messages have simply been included mostly un-edited below. Make rabbit_amqqueue:not_found_or_absent_dirty/1 visible For use in the stream plugin. Use bigger retention policy on max-age test Set coordinator timeout to 30s Handle coordinator unavailable error Handle operator policies as maps when checking if is applicable Add is_policy_applicable/2 to classic queues Ignore restart commands if the stream has been deleted It could happen that after termination some of the monitors are still up and trigger writer/replica restarts Policy support on stream queues Remove subscription events on stream coordinator Ensure old leaders are removed from monitors Introduce delay when retrying a failed phase Note that this ensures monitor is setup, there was a bug where no monitor was really started when re-trying the same phase Restart replicas after leader election instead of relying on old monitors Use timer for stream coordinator retries Fix stream stats for members/online Multiple fixes for replica monitoring and restart Ensure pending commands are appended at the end and re-run Ensure phase is reset with the state Remove duplicates from replica list Restart current phase on state_enter Remove unused import Ensure rabbit is running when checking for stream quorum Restart replicas Add a close/1 function to queue types So that we can get a chance of cleaning up resources if needed. Stream queues close their osiris logs at this point. fix compiler errors stream-queue: take retention into account When calculating ready messages metrics. Add osiris to the list of rabbit deps Retry restart of replicas Do not restart replicas or leaders after receiving a delete cluster command Add more logging to the stream coordinator Monitor subscribed processes on the stream coordinator Memory breakdown for stream queues Update quorum queue event formatter rabbit_msg_record fixes Refactor channel confirms Remove old unconfirmed_messages module that was designed to handle multiple queue fan in logic including all ha mirrors etc. Replaced with simpler rabbit_confirms module that handles the fan out and leaves any queue specific logic (such as confirms from mirrors) to the queue type implemention. Also this module has a dedicated test module. Which is nice. Backward compatibility with 3.8.x events Supports mixed version cluster upgrades Match specification when stream queue already exists Max age retention for stream queues Stop all replicas before starting leader election stream: disallow global qos remove IS_CLASSIC|QUORUM macros Ensure only classic queues are notified on channel down This also removes the delivering_queues map in the channel state as it should not be needed for this and just cause additional unecessary accounting. Polish AMQP 1.0/0.9.1 properties conversion Support byte in application properties, handle 1-bit representation for booleans. Use binary in header for long AMQP 1.0 ID Fix AMQP 1.0 to 0.9.1 conversion Fix test due to incorrect type Convert timestamp application properties to/from seconds AMQP 1.0 uses milliseconds for timestamp and AMQP 0.9.1 uses seconds, so conversion needed. Dialyzer fixes Handle all message-id types AMQP 1.0 is more liberal in it's allowed types of message-id and correlation-id - this adds headers to describe the type of the data in the message_id / correlation_id properties and also handles the case where the data cannot fit by again using headers. Resize stream coordinator cluster when broker configuration changes convert timestamp to and fro seconds user_id should be a binary message annotations keys need to be symbols stream-queue: default exchange and routing key As these won't be present for data written using the rabbitmq-stream plugin. Add exchange, routing key as message annotations To the AMQP 1.0 formatted data to enable roundtrip. Add osiris logging module config And update logging config test suite. Restart election when start of new leader fails The node might have just gone down so we need to try another one Only aux keeps track of phase now, as it might change if the leader election fails Stream coordinator refactor - all state is kept on the ra machine Ensure any ra cluster not a qq is not cleaned up Fixes to recovery and monitoring Add AMQP 1.0 common to dependencies Add rabbit_msg_record module To handle conversions into internal stream storage format. Use rabbitmq-common stream-queue branch Use SSH for osiris dependency Stream coordinator: delete replica Stream coordinator: add replica Stream coordinator: leader failover Stream coordinator: declare and delete Test consuming from a random offset Previous offsets should not be delivered to consumers Consume from stream replicas and multiple test fixes Use max-length-bytes and add new max-segment-size Use SSH for osiris dependency Basic cancel for stream queues Publish stream queues and settle/reject/requeue refactor Consume from stream queues Fix recovery Publish stream messages Add/delete stream replicas Use safe queue names Set retention policy for stream queues Required by the ctl command [#171207092] Stream queue delete queue fix missing callback impl Stream queue declare Queue type abstraction And use the implementing module as the value of the amqqueue record `type` field. This will allow for easy dispatch to the queue type implementation. Queue type abstraction Move queue declare into rabbit_queue_type Move queue delete into queue type implementation Queue type: dequeue/basic_get Move info inside queue type abstraction Move policy change into queue type interface Add purge to queue type Add recovery to the queue type interface Rename amqqueue quorum_nodes field To a more generic an extensible opaque queue type specific map. Fix tests and handle classic API response Fix HA queue confirm bug All mirrors need to be present as queue names. This introduces context linking allowing additional queue refs to be linked to a single "master" queue ref contining the actual queue context. Fix issue with events of deleted queues Also update queue type smoke test to use a cluster by default. correct default value of amqqueue getter Move classic queues further inside queue type interface why [TrackerId] Dialyzer fixes
2020-09-29 18:43:24 +08:00
true ->
receive
{bump_credit, Msg} ->
credit_flow:handle_bump_msg(Msg),
variable_queue_wait_for_shuffling_end(
rabbit_variable_queue:resume(VQ))
end
end.
msg2int(#basic_message{content = #content{ payload_fragments_rev = P}}) ->
Message Containers (#5077) This PR implements an approach for a "protocol (data format) agnostic core" where the format of the message isn't converted at point of reception. Currently all non AMQP 0.9.1 originating messages are converted into a AMQP 0.9.1 flavoured basic_message record before sent to a queue. If the messages are then consumed by the originating protocol they are converted back from AMQP 0.9.1. For some protocols such as MQTT 3.1 this isn't too expensive as MQTT is mostly a fairly easily mapped subset of AMQP 0.9.1 but for others such as AMQP 1.0 the conversions are awkward and in some cases lossy even if consuming from the originating protocol. This PR instead wraps all incoming messages in their originating form into a generic, extensible message container type (mc). The container module exposes an API to get common message details such as size and various properties (ttl, priority etc) directly from the source data type. Each protocol needs to implement the mc behaviour such that when a message originating form one protocol is consumed by another protocol we convert it to the target protocol at that point. The message container also contains annotations, dead letter records and other meta data we need to record during the lifetime of a message. The original protocol message is never modified unless it is consumed. This includes conversion modules to and from amqp, amqpl (AMQP 0.9.1) and mqtt. COMMIT HISTORY: * Refactor away from using the delivery{} record In many places including exchange types. This should make it easier to move towards using a message container type instead of basic_message. Add mc module and move direct replies outside of exchange Lots of changes incl classic queues Implement stream support incl amqp conversions simplify mc state record move mc.erl mc dlx stuff recent history exchange Make tracking work But doesn't take a protocol agnostic approach as we just convert everything into AMQP legacy and back. Might be good enough for now. Tracing as a whole may want a bit of a re-vamp at some point. tidy make quorum queue peek work by legacy conversion dead lettering fixes dead lettering fixes CMQ fixes rabbit_trace type fixes fixes fix Fix classic queue props test assertion fix feature flag and backwards compat Enable message_container feature flag in some SUITEs Dialyzer fixes fixes fix test fixes Various Manually update a gazelle generated file until a gazelle enhancement can be made https://github.com/rabbitmq/rules_erlang/issues/185 Add message_containers_SUITE to bazel and regen bazel files with gazelle from rules_erlang@main Simplify essential proprty access Such as durable, ttl and priority by extracting them into annotations at message container init time. Move type to remove dependenc on amqp10 stuff in mc.erl mostly because I don't know how to make bazel do the right thing add more stuff Refine routing header stuff wip Cosmetics Do not use "maybe" as type name as "maybe" is a keyword since OTP 25 which makes Erlang LS complain. * Dedup death queue names * Fix function clause crashes Fix failing tests in the MQTT shared_SUITE: A classic queue message ID can be undefined as set in https://github.com/rabbitmq/rabbitmq-server/blob/fbe79ff47b4edbc0fd95457e623d6593161ad198/deps/rabbit/src/rabbit_classic_queue_index_v2.erl#L1048 Fix failing tests in the MQTT shared_SUITE-mixed: When feature flag message_containers is disabled, the message is not an #mc{} record, but a #basic_message{} record. * Fix is_utf8_no_null crash Prior to this commit, the function crashed if invalid UTF-8 was provided, e.g.: ``` 1> rabbit_misc:is_valid_shortstr(<<"😇"/utf16>>). ** exception error: no function clause matching rabbit_misc:is_utf8_no_null(<<216,61,222,7>>) (rabbit_misc.erl, line 1481) ``` * Implement mqtt mc behaviour For now via amqp translation. This is still work in progress, but the following SUITEs pass: ``` make -C deps/rabbitmq_mqtt ct-shared t=[mqtt,v5,cluster_size_1] FULL=1 make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_1] FULL=1 ``` * Shorten mc file names Module name length matters because for each persistent message the #mc{} record is persisted to disk. ``` 1> iolist_size(term_to_iovec({mc, rabbit_mc_amqp_legacy})). 30 2> iolist_size(term_to_iovec({mc, mc_amqpl})). 17 ``` This commit renames the mc modules: ``` ag -l rabbit_mc_amqp_legacy | xargs sed -i 's/rabbit_mc_amqp_legacy/mc_amqpl/g' ag -l rabbit_mc_amqp | xargs sed -i 's/rabbit_mc_amqp/mc_amqp/g' ag -l rabbit_mqtt_mc | xargs sed -i 's/rabbit_mqtt_mc/mc_mqtt/g' ``` * mc: make deaths an annotation + fixes * Fix mc_mqtt protocol_state callback * Fix test will_delay_node_restart ``` make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_3]:will_delay_node_restart FULL=1 ``` * Bazel run gazelle * mix format rabbitmqctl.ex * Ensure ttl annotation is refelected in amqp legacy protocol state * Fix id access in message store * Fix rabbit_message_interceptor_SUITE * dializer fixes * Fix rabbit:rabbit_message_interceptor_SUITE-mixed set_annotation/3 should not result in duplicate keys * Fix MQTT shared_SUITE-mixed Up to 3.12 non-MQTT publishes were always QoS 1 regardless of delivery_mode. https://github.com/rabbitmq/rabbitmq-server/blob/75a953ce286a10aca910c098805a4f545989af38/deps/rabbitmq_mqtt/src/rabbit_mqtt_processor.erl#L2075-L2076 From now on, non-MQTT publishes are QoS 1 if durable. This makes more sense. The MQTT plugin must send a #basic_message{} to an old node that does not understand message containers. * Field content of 'v1_0.data' can be binary Fix ``` bazel test //deps/rabbitmq_mqtt:shared_SUITE-mixed \ --test_env FOCUS="-group [mqtt,v4,cluster_size_1] -case trace" \ -t- --test_sharding_strategy=disabled ``` * Remove route/2 and implement route/3 for all exchange types. This removes the route/2 callback from rabbit_exchange_type and makes route/3 mandatory instead. This is a breaking change and will require all implementations of exchange types to update their code, however this is necessary anyway for them to correctly handle the mc type. stream filtering fixes * Translate directly from MQTT to AMQP 0.9.1 * handle undecoded properties in mc_compat amqpl: put clause in right order recover death deatails from amqp data * Replace callback init_amqp with convert_from * Fix return value of lists:keyfind/3 * Translate directly from AMQP 0.9.1 to MQTT * Fix MQTT payload size MQTT payload can be a list when converted from AMQP 0.9.1 for example First conversions tests Plus some other conversion related fixes. bazel bazel translate amqp 1.0 null to undefined mc: property/2 and correlation_id/message_id return type tagged values. To ensure we can support a variety of types better. The type type tags are AMQP 1.0 flavoured. fix death recovery mc_mqtt: impl new api Add callbacks to allow protocols to compact data before storage And make readable if needing to query things repeatedly. bazel fix * more decoding * tracking mixed versions compat * mc: flip default of `durable` annotation to save some data. Assuming most messages are durable and that in memory messages suffer less from persistence overhead it makes sense for a non existent `durable` annotation to mean durable=true. * mc conversion tests and tidy up * mc make x_header unstrict again * amqpl: death record fixes * bazel * amqp -> amqpl conversion test * Fix crash in mc_amqp:size/1 Body can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0/ ct-system t=java ``` on branch native-amqp. * Fix crash in lists:flatten/1 Data can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp. * Fix crash in rabbit_writer Running test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp resulted in the following crash: ``` crasher: initial call: rabbit_writer:enter_mainloop/2 pid: <0.711.0> registered_name: [] exception error: bad argument in function size/1 called as size([<<0>>,<<"Sw">>,[<<160,2>>,<<"hi">>]]) *** argument 1: not tuple or binary in call from rabbit_binary_generator:build_content_frames/7 (rabbit_binary_generator.erl, line 89) in call from rabbit_binary_generator:build_simple_content_frames/4 (rabbit_binary_generator.erl, line 61) in call from rabbit_writer:assemble_frames/5 (rabbit_writer.erl, line 334) in call from rabbit_writer:internal_send_command_async/3 (rabbit_writer.erl, line 365) in call from rabbit_writer:handle_message/2 (rabbit_writer.erl, line 265) in call from rabbit_writer:handle_message/3 (rabbit_writer.erl, line 232) in call from rabbit_writer:mainloop1/2 (rabbit_writer.erl, line 223) ``` because #content.payload_fragments_rev is currently supposed to be a flat list of binaries instead of being an iolist. This commit fixes this crash inefficiently by calling iolist_to_binary/1. A better solution would be to allow AMQP legacy's #content.payload_fragments_rev to be an iolist. * Add accidentally deleted line back * mc: optimise mc_amqp internal format By removint the outer records for message and delivery annotations as well as application properties and footers. * mc: optimis mc_amqp map_add by using upsert * mc: refactoring and bug fixes * mc_SUITE routingheader assertions * mc remove serialize/1 callback as only used by amqp * mc_amqp: avoid returning a nested list from protocol_state * test and bug fix * move infer_type to mc_util * mc fixes and additiona assertions * Support headers exchange routing for MQTT messages When a headers exchange is bound to the MQTT topic exchange, routing will be performend based on both MQTT topic (by the topic exchange) and MQTT User Property (by the headers exchange). This combines the best worlds of both MQTT 5.0 and AMQP 0.9.1 and enables powerful routing topologies. When the User Property contains the same name multiple times, only the last name (and value) will be considered by the headers exchange. * Fix crash when sending from stream to amqpl When publishing a message via the stream protocol and consuming it via AMQP 0.9.1, the following crash occurred prior to this commit: ``` crasher: initial call: rabbit_channel:init/1 pid: <0.818.0> registered_name: [] exception exit: {{badmatch,undefined}, [{rabbit_channel,handle_deliver0,4, [{file,"rabbit_channel.erl"}, {line,2728}]}, {lists,foldl,3,[{file,"lists.erl"},{line,1594}]}, {rabbit_channel,handle_cast,2, [{file,"rabbit_channel.erl"}, {line,728}]}, {gen_server2,handle_msg,2, [{file,"gen_server2.erl"},{line,1056}]}, {proc_lib,wake_up,3, [{file,"proc_lib.erl"},{line,251}]}]} ``` This commit first gives `mc:init/3` the chance to set exchange and routing_keys annotations. If not set, `rabbit_stream_queue` will set these annotations assuming the message was originally published via the stream protocol. * Support consistent hash exchange routing for MQTT 5.0 When a consistent hash exchange is bound to the MQTT topic exchange, MQTT 5.0 messages can be routed to queues consistently based on the Correlation-Data in the PUBLISH packet. * Convert MQTT 5.0 User Property * to AMQP 0.9.1 headers * from AMQP 0.9.1 headers * to AMQP 1.0 application properties and message annotations * from AMQP 1.0 application properties and message annotations * Make use of Annotations in mc_mqtt:protocol_state/2 mc_mqtt:protocol_state/2 includes Annotations as parameter. It's cleaner to make use of these Annotations when computing the protocol state instead of relying on the caller (rabbitmq_mqtt_processor) to compute the protocol state. * Enforce AMQP 0.9.1 field name length limit The AMQP 0.9.1 spec prohibits field names longer than 128 characters. Therefore, when converting AMQP 1.0 message annotations, application properties or MQTT 5.0 User Property to AMQP 0.9.1 headers, drop any names longer than 128 characters. * Fix type specs Apply feedback from Michael Davis Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * Add mc_mqtt unit test suite Implement mc_mqtt:x_header/2 * Translate indicator that payload is UTF-8 encoded when converting between MQTT 5.0 and AMQP 1.0 * Translate single amqp-value section from AMQP 1.0 to MQTT Convert to a text representation, if possible, and indicate to MQTT client that the payload is UTF-8 encoded. This way, the MQTT client will be able to parse the payload. If conversion to text representation is not possible, encode the payload using the AMQP 1.0 type system and indiate the encoding via Content-Type message/vnd.rabbitmq.amqp. This Content-Type is not registered. Type "message" makes sense since it's a message. Vendor tree "vnd.rabbitmq.amqp" makes sense since merely subtype "amqp" is not registered. * Fix payload conversion * Translate Response Topic between MQTT and AMQP Translate MQTT 5.0 Response Topic to AMQP 1.0 reply-to address and vice versa. The Response Topic must be a UTF-8 encoded string. This commit re-uses the already defined RabbitMQ target addresses: ``` "/topic/" RK Publish to amq.topic with routing key RK "/exchange/" X "/" RK Publish to exchange X with routing key RK ``` By default, the MQTT topic exchange is configure dto be amq.topic using the 1st target address. When an operator modifies the mqtt.exchange, the 2nd target address is used. * Apply PR feedback and fix formatting Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * tidy up * Add MQTT message_containers test * consistent hash exchange: avoid amqp legacy conversion When hashing on a header value. * Avoid converting to amqp legacy when using exchange federation * Fix test flake * test and dialyzer fixes * dialyzer fix * Add MQTT protocol interoperability tests Test receiving from and sending to MQTT 5.0 and * AMQP 0.9.1 * AMQP 1.0 * STOMP * Streams * Regenerate portions of deps/rabbit/app.bzl with gazelle I'm not exactly sure how this happened, but gazell seems to have been run with an older version of the rules_erlang gazelle extension at some point. This caused generation of a structure that is no longer used. This commit updates the structure to the current pattern. * mc: refactoring * mc_amqpl: handle delivery annotations Just in case they are included. Also use iolist_to_iovec to create flat list of binaries when converting from amqp with amqp encoded payload. --------- Co-authored-by: David Ansari <david.ansari@gmx.de> Co-authored-by: Michael Davis <mcarsondavis@gmail.com> Co-authored-by: Rin Kuryloski <kuryloskip@vmware.com>
2023-08-31 18:27:13 +08:00
binary_to_term(list_to_binary(lists:reverse(P)));
msg2int(Msg) ->
#content{payload_fragments_rev = P} = mc:protocol_state(Msg),
binary_to_term(list_to_binary(lists:reverse(P))).
ack_subset(AckSeqs, Interval, Rem) ->
lists:filter(fun ({_Ack, N}) -> (N + Rem) rem Interval == 0 end, AckSeqs).
requeue_one_by_one(Acks, VQ) ->
lists:foldl(fun (AckTag, VQN) ->
{_MsgId, VQM} = rabbit_variable_queue:requeue(
[AckTag], VQN),
VQM
end, VQ, Acks).
%% Create a vq with messages in q1, delta, and q3, and holes (in the
%% form of pending acks) in the latter two.
variable_queue_with_holes(VQ0) ->
Interval = 2048, %% should match vq:IO_BATCH_SIZE
2021-10-11 17:49:11 +08:00
IndexMod = index_mod(),
Count = IndexMod:next_segment_boundary(0)*2 + 2 * Interval,
Seq = lists:seq(1, Count),
Make backing_queue_SUITE fast on macOS This very small patch requires extended explanations. The patch swaps two lines in a rabbit_variable_queue setup: one which sets the memory hint to 0 which results in reduce_memory_usage to always flush to disk and fsync; and another which publishes a lot of messages to the queue that will after that point be manipulated further to get the queue in the exact right state for the relevant tests. The problem with calling reduce_memory_usage after every single message has been published is not writing to disk (v2 tests do not suffer from performance issues in that regard) but rather that rabbit_queue_index will always flush its journal (containing the one message), which results in opening the segment file, appending to it, and closing it. The file handling is done by file_handle_cache which, in this case, will always fsync the data before closing the file. And that's this one fsync per message that makes the relevant tests very slow. By swapping the lines, meaning we publish all messages first and then set the memory hint to 0, we end up with a single reduce_memory_usage call that results in an fsync, at the end. (There may be other fsyncs as part of normal operations.) And still get the same result because all messages will have been flushed to disk, only this time in far fewer operations. This doesn't seem to have been causing problems on CI which already runs the tests very fast but should help macOS and possibly other development environments.
2022-03-18 19:47:22 +08:00
VQ1 = variable_queue_publish(
false, 1, Count,
Make backing_queue_SUITE fast on macOS This very small patch requires extended explanations. The patch swaps two lines in a rabbit_variable_queue setup: one which sets the memory hint to 0 which results in reduce_memory_usage to always flush to disk and fsync; and another which publishes a lot of messages to the queue that will after that point be manipulated further to get the queue in the exact right state for the relevant tests. The problem with calling reduce_memory_usage after every single message has been published is not writing to disk (v2 tests do not suffer from performance issues in that regard) but rather that rabbit_queue_index will always flush its journal (containing the one message), which results in opening the segment file, appending to it, and closing it. The file handling is done by file_handle_cache which, in this case, will always fsync the data before closing the file. And that's this one fsync per message that makes the relevant tests very slow. By swapping the lines, meaning we publish all messages first and then set the memory hint to 0, we end up with a single reduce_memory_usage call that results in an fsync, at the end. (There may be other fsyncs as part of normal operations.) And still get the same result because all messages will have been flushed to disk, only this time in far fewer operations. This doesn't seem to have been causing problems on CI which already runs the tests very fast but should help macOS and possibly other development environments.
2022-03-18 19:47:22 +08:00
fun (_, P) -> P end, fun erlang:term_to_binary/1, VQ0),
{VQ3, AcksR} = variable_queue_fetch(Count, false, false, Count, VQ1),
Acks = lists:reverse(AcksR),
AckSeqs = lists:zip(Acks, Seq),
[{Subset1, _Seq1}, {Subset2, _Seq2}, {Subset3, Seq3}] =
[lists:unzip(ack_subset(AckSeqs, Interval, I)) || I <- [0, 1, 2]],
%% we requeue in three phases in order to exercise requeuing logic
%% in various vq states
{_MsgIds, VQ4} = rabbit_variable_queue:requeue(
Acks -- (Subset1 ++ Subset2 ++ Subset3), VQ3),
VQ5 = requeue_one_by_one(Subset1, VQ4),
%% by now we have some messages (and holes) in delta
VQ6 = requeue_one_by_one(Subset2, VQ5),
%% add the q1 tail
VQ8 = variable_queue_publish(
true, Count + 1, Interval,
fun (_, P) -> P end, fun erlang:term_to_binary/1, VQ6),
%% assertions
vq_with_holes_assertions(VQ8),
Depth = Count + Interval,
Depth = rabbit_variable_queue:depth(VQ8),
Len = Depth - length(Subset3),
Len = rabbit_variable_queue:len(VQ8),
Stream Queue This is an aggregated commit of all changes related to the initial implementation of queue types and on top of that the stream queue type. The varios commit messages have simply been included mostly un-edited below. Make rabbit_amqqueue:not_found_or_absent_dirty/1 visible For use in the stream plugin. Use bigger retention policy on max-age test Set coordinator timeout to 30s Handle coordinator unavailable error Handle operator policies as maps when checking if is applicable Add is_policy_applicable/2 to classic queues Ignore restart commands if the stream has been deleted It could happen that after termination some of the monitors are still up and trigger writer/replica restarts Policy support on stream queues Remove subscription events on stream coordinator Ensure old leaders are removed from monitors Introduce delay when retrying a failed phase Note that this ensures monitor is setup, there was a bug where no monitor was really started when re-trying the same phase Restart replicas after leader election instead of relying on old monitors Use timer for stream coordinator retries Fix stream stats for members/online Multiple fixes for replica monitoring and restart Ensure pending commands are appended at the end and re-run Ensure phase is reset with the state Remove duplicates from replica list Restart current phase on state_enter Remove unused import Ensure rabbit is running when checking for stream quorum Restart replicas Add a close/1 function to queue types So that we can get a chance of cleaning up resources if needed. Stream queues close their osiris logs at this point. fix compiler errors stream-queue: take retention into account When calculating ready messages metrics. Add osiris to the list of rabbit deps Retry restart of replicas Do not restart replicas or leaders after receiving a delete cluster command Add more logging to the stream coordinator Monitor subscribed processes on the stream coordinator Memory breakdown for stream queues Update quorum queue event formatter rabbit_msg_record fixes Refactor channel confirms Remove old unconfirmed_messages module that was designed to handle multiple queue fan in logic including all ha mirrors etc. Replaced with simpler rabbit_confirms module that handles the fan out and leaves any queue specific logic (such as confirms from mirrors) to the queue type implemention. Also this module has a dedicated test module. Which is nice. Backward compatibility with 3.8.x events Supports mixed version cluster upgrades Match specification when stream queue already exists Max age retention for stream queues Stop all replicas before starting leader election stream: disallow global qos remove IS_CLASSIC|QUORUM macros Ensure only classic queues are notified on channel down This also removes the delivering_queues map in the channel state as it should not be needed for this and just cause additional unecessary accounting. Polish AMQP 1.0/0.9.1 properties conversion Support byte in application properties, handle 1-bit representation for booleans. Use binary in header for long AMQP 1.0 ID Fix AMQP 1.0 to 0.9.1 conversion Fix test due to incorrect type Convert timestamp application properties to/from seconds AMQP 1.0 uses milliseconds for timestamp and AMQP 0.9.1 uses seconds, so conversion needed. Dialyzer fixes Handle all message-id types AMQP 1.0 is more liberal in it's allowed types of message-id and correlation-id - this adds headers to describe the type of the data in the message_id / correlation_id properties and also handles the case where the data cannot fit by again using headers. Resize stream coordinator cluster when broker configuration changes convert timestamp to and fro seconds user_id should be a binary message annotations keys need to be symbols stream-queue: default exchange and routing key As these won't be present for data written using the rabbitmq-stream plugin. Add exchange, routing key as message annotations To the AMQP 1.0 formatted data to enable roundtrip. Add osiris logging module config And update logging config test suite. Restart election when start of new leader fails The node might have just gone down so we need to try another one Only aux keeps track of phase now, as it might change if the leader election fails Stream coordinator refactor - all state is kept on the ra machine Ensure any ra cluster not a qq is not cleaned up Fixes to recovery and monitoring Add AMQP 1.0 common to dependencies Add rabbit_msg_record module To handle conversions into internal stream storage format. Use rabbitmq-common stream-queue branch Use SSH for osiris dependency Stream coordinator: delete replica Stream coordinator: add replica Stream coordinator: leader failover Stream coordinator: declare and delete Test consuming from a random offset Previous offsets should not be delivered to consumers Consume from stream replicas and multiple test fixes Use max-length-bytes and add new max-segment-size Use SSH for osiris dependency Basic cancel for stream queues Publish stream queues and settle/reject/requeue refactor Consume from stream queues Fix recovery Publish stream messages Add/delete stream replicas Use safe queue names Set retention policy for stream queues Required by the ctl command [#171207092] Stream queue delete queue fix missing callback impl Stream queue declare Queue type abstraction And use the implementing module as the value of the amqqueue record `type` field. This will allow for easy dispatch to the queue type implementation. Queue type abstraction Move queue declare into rabbit_queue_type Move queue delete into queue type implementation Queue type: dequeue/basic_get Move info inside queue type abstraction Move policy change into queue type interface Add purge to queue type Add recovery to the queue type interface Rename amqqueue quorum_nodes field To a more generic an extensible opaque queue type specific map. Fix tests and handle classic API response Fix HA queue confirm bug All mirrors need to be present as queue names. This introduces context linking allowing additional queue refs to be linked to a single "master" queue ref contining the actual queue context. Fix issue with events of deleted queues Also update queue type smoke test to use a cluster by default. correct default value of amqqueue getter Move classic queues further inside queue type interface why [TrackerId] Dialyzer fixes
2020-09-29 18:43:24 +08:00
{Seq3, Seq -- Seq3, lists:seq(Count + 1, Count + Interval), VQ8}.
vq_with_holes_assertions(VQ) ->
[false =
case V of
{delta, _, 0, _, _} -> true;
0 -> true;
_ -> false
end || {K, V} <- variable_queue_status(VQ),
lists:member(K, [delta, q3])].
check_variable_queue_status(VQ0, Props) ->
VQ1 = variable_queue_wait_for_shuffling_end(VQ0),
S = variable_queue_status(VQ1),
assert_props(S, Props),
VQ1.
Stream Queue This is an aggregated commit of all changes related to the initial implementation of queue types and on top of that the stream queue type. The varios commit messages have simply been included mostly un-edited below. Make rabbit_amqqueue:not_found_or_absent_dirty/1 visible For use in the stream plugin. Use bigger retention policy on max-age test Set coordinator timeout to 30s Handle coordinator unavailable error Handle operator policies as maps when checking if is applicable Add is_policy_applicable/2 to classic queues Ignore restart commands if the stream has been deleted It could happen that after termination some of the monitors are still up and trigger writer/replica restarts Policy support on stream queues Remove subscription events on stream coordinator Ensure old leaders are removed from monitors Introduce delay when retrying a failed phase Note that this ensures monitor is setup, there was a bug where no monitor was really started when re-trying the same phase Restart replicas after leader election instead of relying on old monitors Use timer for stream coordinator retries Fix stream stats for members/online Multiple fixes for replica monitoring and restart Ensure pending commands are appended at the end and re-run Ensure phase is reset with the state Remove duplicates from replica list Restart current phase on state_enter Remove unused import Ensure rabbit is running when checking for stream quorum Restart replicas Add a close/1 function to queue types So that we can get a chance of cleaning up resources if needed. Stream queues close their osiris logs at this point. fix compiler errors stream-queue: take retention into account When calculating ready messages metrics. Add osiris to the list of rabbit deps Retry restart of replicas Do not restart replicas or leaders after receiving a delete cluster command Add more logging to the stream coordinator Monitor subscribed processes on the stream coordinator Memory breakdown for stream queues Update quorum queue event formatter rabbit_msg_record fixes Refactor channel confirms Remove old unconfirmed_messages module that was designed to handle multiple queue fan in logic including all ha mirrors etc. Replaced with simpler rabbit_confirms module that handles the fan out and leaves any queue specific logic (such as confirms from mirrors) to the queue type implemention. Also this module has a dedicated test module. Which is nice. Backward compatibility with 3.8.x events Supports mixed version cluster upgrades Match specification when stream queue already exists Max age retention for stream queues Stop all replicas before starting leader election stream: disallow global qos remove IS_CLASSIC|QUORUM macros Ensure only classic queues are notified on channel down This also removes the delivering_queues map in the channel state as it should not be needed for this and just cause additional unecessary accounting. Polish AMQP 1.0/0.9.1 properties conversion Support byte in application properties, handle 1-bit representation for booleans. Use binary in header for long AMQP 1.0 ID Fix AMQP 1.0 to 0.9.1 conversion Fix test due to incorrect type Convert timestamp application properties to/from seconds AMQP 1.0 uses milliseconds for timestamp and AMQP 0.9.1 uses seconds, so conversion needed. Dialyzer fixes Handle all message-id types AMQP 1.0 is more liberal in it's allowed types of message-id and correlation-id - this adds headers to describe the type of the data in the message_id / correlation_id properties and also handles the case where the data cannot fit by again using headers. Resize stream coordinator cluster when broker configuration changes convert timestamp to and fro seconds user_id should be a binary message annotations keys need to be symbols stream-queue: default exchange and routing key As these won't be present for data written using the rabbitmq-stream plugin. Add exchange, routing key as message annotations To the AMQP 1.0 formatted data to enable roundtrip. Add osiris logging module config And update logging config test suite. Restart election when start of new leader fails The node might have just gone down so we need to try another one Only aux keeps track of phase now, as it might change if the leader election fails Stream coordinator refactor - all state is kept on the ra machine Ensure any ra cluster not a qq is not cleaned up Fixes to recovery and monitoring Add AMQP 1.0 common to dependencies Add rabbit_msg_record module To handle conversions into internal stream storage format. Use rabbitmq-common stream-queue branch Use SSH for osiris dependency Stream coordinator: delete replica Stream coordinator: add replica Stream coordinator: leader failover Stream coordinator: declare and delete Test consuming from a random offset Previous offsets should not be delivered to consumers Consume from stream replicas and multiple test fixes Use max-length-bytes and add new max-segment-size Use SSH for osiris dependency Basic cancel for stream queues Publish stream queues and settle/reject/requeue refactor Consume from stream queues Fix recovery Publish stream messages Add/delete stream replicas Use safe queue names Set retention policy for stream queues Required by the ctl command [#171207092] Stream queue delete queue fix missing callback impl Stream queue declare Queue type abstraction And use the implementing module as the value of the amqqueue record `type` field. This will allow for easy dispatch to the queue type implementation. Queue type abstraction Move queue declare into rabbit_queue_type Move queue delete into queue type implementation Queue type: dequeue/basic_get Move info inside queue type abstraction Move policy change into queue type interface Add purge to queue type Add recovery to the queue type interface Rename amqqueue quorum_nodes field To a more generic an extensible opaque queue type specific map. Fix tests and handle classic API response Fix HA queue confirm bug All mirrors need to be present as queue names. This introduces context linking allowing additional queue refs to be linked to a single "master" queue ref contining the actual queue context. Fix issue with events of deleted queues Also update queue type smoke test to use a cluster by default. correct default value of amqqueue getter Move classic queues further inside queue type interface why [TrackerId] Dialyzer fixes
2020-09-29 18:43:24 +08:00
flush() ->
receive
Any ->
ct:pal("flush ~tp", [Any]),
Stream Queue This is an aggregated commit of all changes related to the initial implementation of queue types and on top of that the stream queue type. The varios commit messages have simply been included mostly un-edited below. Make rabbit_amqqueue:not_found_or_absent_dirty/1 visible For use in the stream plugin. Use bigger retention policy on max-age test Set coordinator timeout to 30s Handle coordinator unavailable error Handle operator policies as maps when checking if is applicable Add is_policy_applicable/2 to classic queues Ignore restart commands if the stream has been deleted It could happen that after termination some of the monitors are still up and trigger writer/replica restarts Policy support on stream queues Remove subscription events on stream coordinator Ensure old leaders are removed from monitors Introduce delay when retrying a failed phase Note that this ensures monitor is setup, there was a bug where no monitor was really started when re-trying the same phase Restart replicas after leader election instead of relying on old monitors Use timer for stream coordinator retries Fix stream stats for members/online Multiple fixes for replica monitoring and restart Ensure pending commands are appended at the end and re-run Ensure phase is reset with the state Remove duplicates from replica list Restart current phase on state_enter Remove unused import Ensure rabbit is running when checking for stream quorum Restart replicas Add a close/1 function to queue types So that we can get a chance of cleaning up resources if needed. Stream queues close their osiris logs at this point. fix compiler errors stream-queue: take retention into account When calculating ready messages metrics. Add osiris to the list of rabbit deps Retry restart of replicas Do not restart replicas or leaders after receiving a delete cluster command Add more logging to the stream coordinator Monitor subscribed processes on the stream coordinator Memory breakdown for stream queues Update quorum queue event formatter rabbit_msg_record fixes Refactor channel confirms Remove old unconfirmed_messages module that was designed to handle multiple queue fan in logic including all ha mirrors etc. Replaced with simpler rabbit_confirms module that handles the fan out and leaves any queue specific logic (such as confirms from mirrors) to the queue type implemention. Also this module has a dedicated test module. Which is nice. Backward compatibility with 3.8.x events Supports mixed version cluster upgrades Match specification when stream queue already exists Max age retention for stream queues Stop all replicas before starting leader election stream: disallow global qos remove IS_CLASSIC|QUORUM macros Ensure only classic queues are notified on channel down This also removes the delivering_queues map in the channel state as it should not be needed for this and just cause additional unecessary accounting. Polish AMQP 1.0/0.9.1 properties conversion Support byte in application properties, handle 1-bit representation for booleans. Use binary in header for long AMQP 1.0 ID Fix AMQP 1.0 to 0.9.1 conversion Fix test due to incorrect type Convert timestamp application properties to/from seconds AMQP 1.0 uses milliseconds for timestamp and AMQP 0.9.1 uses seconds, so conversion needed. Dialyzer fixes Handle all message-id types AMQP 1.0 is more liberal in it's allowed types of message-id and correlation-id - this adds headers to describe the type of the data in the message_id / correlation_id properties and also handles the case where the data cannot fit by again using headers. Resize stream coordinator cluster when broker configuration changes convert timestamp to and fro seconds user_id should be a binary message annotations keys need to be symbols stream-queue: default exchange and routing key As these won't be present for data written using the rabbitmq-stream plugin. Add exchange, routing key as message annotations To the AMQP 1.0 formatted data to enable roundtrip. Add osiris logging module config And update logging config test suite. Restart election when start of new leader fails The node might have just gone down so we need to try another one Only aux keeps track of phase now, as it might change if the leader election fails Stream coordinator refactor - all state is kept on the ra machine Ensure any ra cluster not a qq is not cleaned up Fixes to recovery and monitoring Add AMQP 1.0 common to dependencies Add rabbit_msg_record module To handle conversions into internal stream storage format. Use rabbitmq-common stream-queue branch Use SSH for osiris dependency Stream coordinator: delete replica Stream coordinator: add replica Stream coordinator: leader failover Stream coordinator: declare and delete Test consuming from a random offset Previous offsets should not be delivered to consumers Consume from stream replicas and multiple test fixes Use max-length-bytes and add new max-segment-size Use SSH for osiris dependency Basic cancel for stream queues Publish stream queues and settle/reject/requeue refactor Consume from stream queues Fix recovery Publish stream messages Add/delete stream replicas Use safe queue names Set retention policy for stream queues Required by the ctl command [#171207092] Stream queue delete queue fix missing callback impl Stream queue declare Queue type abstraction And use the implementing module as the value of the amqqueue record `type` field. This will allow for easy dispatch to the queue type implementation. Queue type abstraction Move queue declare into rabbit_queue_type Move queue delete into queue type implementation Queue type: dequeue/basic_get Move info inside queue type abstraction Move policy change into queue type interface Add purge to queue type Add recovery to the queue type interface Rename amqqueue quorum_nodes field To a more generic an extensible opaque queue type specific map. Fix tests and handle classic API response Fix HA queue confirm bug All mirrors need to be present as queue names. This introduces context linking allowing additional queue refs to be linked to a single "master" queue ref contining the actual queue context. Fix issue with events of deleted queues Also update queue type smoke test to use a cluster by default. correct default value of amqqueue getter Move classic queues further inside queue type interface why [TrackerId] Dialyzer fixes
2020-09-29 18:43:24 +08:00
flush()
after 0 ->
ok
end.
Message Containers (#5077) This PR implements an approach for a "protocol (data format) agnostic core" where the format of the message isn't converted at point of reception. Currently all non AMQP 0.9.1 originating messages are converted into a AMQP 0.9.1 flavoured basic_message record before sent to a queue. If the messages are then consumed by the originating protocol they are converted back from AMQP 0.9.1. For some protocols such as MQTT 3.1 this isn't too expensive as MQTT is mostly a fairly easily mapped subset of AMQP 0.9.1 but for others such as AMQP 1.0 the conversions are awkward and in some cases lossy even if consuming from the originating protocol. This PR instead wraps all incoming messages in their originating form into a generic, extensible message container type (mc). The container module exposes an API to get common message details such as size and various properties (ttl, priority etc) directly from the source data type. Each protocol needs to implement the mc behaviour such that when a message originating form one protocol is consumed by another protocol we convert it to the target protocol at that point. The message container also contains annotations, dead letter records and other meta data we need to record during the lifetime of a message. The original protocol message is never modified unless it is consumed. This includes conversion modules to and from amqp, amqpl (AMQP 0.9.1) and mqtt. COMMIT HISTORY: * Refactor away from using the delivery{} record In many places including exchange types. This should make it easier to move towards using a message container type instead of basic_message. Add mc module and move direct replies outside of exchange Lots of changes incl classic queues Implement stream support incl amqp conversions simplify mc state record move mc.erl mc dlx stuff recent history exchange Make tracking work But doesn't take a protocol agnostic approach as we just convert everything into AMQP legacy and back. Might be good enough for now. Tracing as a whole may want a bit of a re-vamp at some point. tidy make quorum queue peek work by legacy conversion dead lettering fixes dead lettering fixes CMQ fixes rabbit_trace type fixes fixes fix Fix classic queue props test assertion fix feature flag and backwards compat Enable message_container feature flag in some SUITEs Dialyzer fixes fixes fix test fixes Various Manually update a gazelle generated file until a gazelle enhancement can be made https://github.com/rabbitmq/rules_erlang/issues/185 Add message_containers_SUITE to bazel and regen bazel files with gazelle from rules_erlang@main Simplify essential proprty access Such as durable, ttl and priority by extracting them into annotations at message container init time. Move type to remove dependenc on amqp10 stuff in mc.erl mostly because I don't know how to make bazel do the right thing add more stuff Refine routing header stuff wip Cosmetics Do not use "maybe" as type name as "maybe" is a keyword since OTP 25 which makes Erlang LS complain. * Dedup death queue names * Fix function clause crashes Fix failing tests in the MQTT shared_SUITE: A classic queue message ID can be undefined as set in https://github.com/rabbitmq/rabbitmq-server/blob/fbe79ff47b4edbc0fd95457e623d6593161ad198/deps/rabbit/src/rabbit_classic_queue_index_v2.erl#L1048 Fix failing tests in the MQTT shared_SUITE-mixed: When feature flag message_containers is disabled, the message is not an #mc{} record, but a #basic_message{} record. * Fix is_utf8_no_null crash Prior to this commit, the function crashed if invalid UTF-8 was provided, e.g.: ``` 1> rabbit_misc:is_valid_shortstr(<<"😇"/utf16>>). ** exception error: no function clause matching rabbit_misc:is_utf8_no_null(<<216,61,222,7>>) (rabbit_misc.erl, line 1481) ``` * Implement mqtt mc behaviour For now via amqp translation. This is still work in progress, but the following SUITEs pass: ``` make -C deps/rabbitmq_mqtt ct-shared t=[mqtt,v5,cluster_size_1] FULL=1 make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_1] FULL=1 ``` * Shorten mc file names Module name length matters because for each persistent message the #mc{} record is persisted to disk. ``` 1> iolist_size(term_to_iovec({mc, rabbit_mc_amqp_legacy})). 30 2> iolist_size(term_to_iovec({mc, mc_amqpl})). 17 ``` This commit renames the mc modules: ``` ag -l rabbit_mc_amqp_legacy | xargs sed -i 's/rabbit_mc_amqp_legacy/mc_amqpl/g' ag -l rabbit_mc_amqp | xargs sed -i 's/rabbit_mc_amqp/mc_amqp/g' ag -l rabbit_mqtt_mc | xargs sed -i 's/rabbit_mqtt_mc/mc_mqtt/g' ``` * mc: make deaths an annotation + fixes * Fix mc_mqtt protocol_state callback * Fix test will_delay_node_restart ``` make -C deps/rabbitmq_mqtt ct-v5 t=[mqtt,cluster_size_3]:will_delay_node_restart FULL=1 ``` * Bazel run gazelle * mix format rabbitmqctl.ex * Ensure ttl annotation is refelected in amqp legacy protocol state * Fix id access in message store * Fix rabbit_message_interceptor_SUITE * dializer fixes * Fix rabbit:rabbit_message_interceptor_SUITE-mixed set_annotation/3 should not result in duplicate keys * Fix MQTT shared_SUITE-mixed Up to 3.12 non-MQTT publishes were always QoS 1 regardless of delivery_mode. https://github.com/rabbitmq/rabbitmq-server/blob/75a953ce286a10aca910c098805a4f545989af38/deps/rabbitmq_mqtt/src/rabbit_mqtt_processor.erl#L2075-L2076 From now on, non-MQTT publishes are QoS 1 if durable. This makes more sense. The MQTT plugin must send a #basic_message{} to an old node that does not understand message containers. * Field content of 'v1_0.data' can be binary Fix ``` bazel test //deps/rabbitmq_mqtt:shared_SUITE-mixed \ --test_env FOCUS="-group [mqtt,v4,cluster_size_1] -case trace" \ -t- --test_sharding_strategy=disabled ``` * Remove route/2 and implement route/3 for all exchange types. This removes the route/2 callback from rabbit_exchange_type and makes route/3 mandatory instead. This is a breaking change and will require all implementations of exchange types to update their code, however this is necessary anyway for them to correctly handle the mc type. stream filtering fixes * Translate directly from MQTT to AMQP 0.9.1 * handle undecoded properties in mc_compat amqpl: put clause in right order recover death deatails from amqp data * Replace callback init_amqp with convert_from * Fix return value of lists:keyfind/3 * Translate directly from AMQP 0.9.1 to MQTT * Fix MQTT payload size MQTT payload can be a list when converted from AMQP 0.9.1 for example First conversions tests Plus some other conversion related fixes. bazel bazel translate amqp 1.0 null to undefined mc: property/2 and correlation_id/message_id return type tagged values. To ensure we can support a variety of types better. The type type tags are AMQP 1.0 flavoured. fix death recovery mc_mqtt: impl new api Add callbacks to allow protocols to compact data before storage And make readable if needing to query things repeatedly. bazel fix * more decoding * tracking mixed versions compat * mc: flip default of `durable` annotation to save some data. Assuming most messages are durable and that in memory messages suffer less from persistence overhead it makes sense for a non existent `durable` annotation to mean durable=true. * mc conversion tests and tidy up * mc make x_header unstrict again * amqpl: death record fixes * bazel * amqp -> amqpl conversion test * Fix crash in mc_amqp:size/1 Body can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0/ ct-system t=java ``` on branch native-amqp. * Fix crash in lists:flatten/1 Data can be a single amqp-value section (instead of being a list) as shown by test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp. * Fix crash in rabbit_writer Running test ``` make -C deps/rabbitmq_amqp1_0 ct-system t=dotnet:roundtrip_to_amqp_091 ``` on branch native-amqp resulted in the following crash: ``` crasher: initial call: rabbit_writer:enter_mainloop/2 pid: <0.711.0> registered_name: [] exception error: bad argument in function size/1 called as size([<<0>>,<<"Sw">>,[<<160,2>>,<<"hi">>]]) *** argument 1: not tuple or binary in call from rabbit_binary_generator:build_content_frames/7 (rabbit_binary_generator.erl, line 89) in call from rabbit_binary_generator:build_simple_content_frames/4 (rabbit_binary_generator.erl, line 61) in call from rabbit_writer:assemble_frames/5 (rabbit_writer.erl, line 334) in call from rabbit_writer:internal_send_command_async/3 (rabbit_writer.erl, line 365) in call from rabbit_writer:handle_message/2 (rabbit_writer.erl, line 265) in call from rabbit_writer:handle_message/3 (rabbit_writer.erl, line 232) in call from rabbit_writer:mainloop1/2 (rabbit_writer.erl, line 223) ``` because #content.payload_fragments_rev is currently supposed to be a flat list of binaries instead of being an iolist. This commit fixes this crash inefficiently by calling iolist_to_binary/1. A better solution would be to allow AMQP legacy's #content.payload_fragments_rev to be an iolist. * Add accidentally deleted line back * mc: optimise mc_amqp internal format By removint the outer records for message and delivery annotations as well as application properties and footers. * mc: optimis mc_amqp map_add by using upsert * mc: refactoring and bug fixes * mc_SUITE routingheader assertions * mc remove serialize/1 callback as only used by amqp * mc_amqp: avoid returning a nested list from protocol_state * test and bug fix * move infer_type to mc_util * mc fixes and additiona assertions * Support headers exchange routing for MQTT messages When a headers exchange is bound to the MQTT topic exchange, routing will be performend based on both MQTT topic (by the topic exchange) and MQTT User Property (by the headers exchange). This combines the best worlds of both MQTT 5.0 and AMQP 0.9.1 and enables powerful routing topologies. When the User Property contains the same name multiple times, only the last name (and value) will be considered by the headers exchange. * Fix crash when sending from stream to amqpl When publishing a message via the stream protocol and consuming it via AMQP 0.9.1, the following crash occurred prior to this commit: ``` crasher: initial call: rabbit_channel:init/1 pid: <0.818.0> registered_name: [] exception exit: {{badmatch,undefined}, [{rabbit_channel,handle_deliver0,4, [{file,"rabbit_channel.erl"}, {line,2728}]}, {lists,foldl,3,[{file,"lists.erl"},{line,1594}]}, {rabbit_channel,handle_cast,2, [{file,"rabbit_channel.erl"}, {line,728}]}, {gen_server2,handle_msg,2, [{file,"gen_server2.erl"},{line,1056}]}, {proc_lib,wake_up,3, [{file,"proc_lib.erl"},{line,251}]}]} ``` This commit first gives `mc:init/3` the chance to set exchange and routing_keys annotations. If not set, `rabbit_stream_queue` will set these annotations assuming the message was originally published via the stream protocol. * Support consistent hash exchange routing for MQTT 5.0 When a consistent hash exchange is bound to the MQTT topic exchange, MQTT 5.0 messages can be routed to queues consistently based on the Correlation-Data in the PUBLISH packet. * Convert MQTT 5.0 User Property * to AMQP 0.9.1 headers * from AMQP 0.9.1 headers * to AMQP 1.0 application properties and message annotations * from AMQP 1.0 application properties and message annotations * Make use of Annotations in mc_mqtt:protocol_state/2 mc_mqtt:protocol_state/2 includes Annotations as parameter. It's cleaner to make use of these Annotations when computing the protocol state instead of relying on the caller (rabbitmq_mqtt_processor) to compute the protocol state. * Enforce AMQP 0.9.1 field name length limit The AMQP 0.9.1 spec prohibits field names longer than 128 characters. Therefore, when converting AMQP 1.0 message annotations, application properties or MQTT 5.0 User Property to AMQP 0.9.1 headers, drop any names longer than 128 characters. * Fix type specs Apply feedback from Michael Davis Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * Add mc_mqtt unit test suite Implement mc_mqtt:x_header/2 * Translate indicator that payload is UTF-8 encoded when converting between MQTT 5.0 and AMQP 1.0 * Translate single amqp-value section from AMQP 1.0 to MQTT Convert to a text representation, if possible, and indicate to MQTT client that the payload is UTF-8 encoded. This way, the MQTT client will be able to parse the payload. If conversion to text representation is not possible, encode the payload using the AMQP 1.0 type system and indiate the encoding via Content-Type message/vnd.rabbitmq.amqp. This Content-Type is not registered. Type "message" makes sense since it's a message. Vendor tree "vnd.rabbitmq.amqp" makes sense since merely subtype "amqp" is not registered. * Fix payload conversion * Translate Response Topic between MQTT and AMQP Translate MQTT 5.0 Response Topic to AMQP 1.0 reply-to address and vice versa. The Response Topic must be a UTF-8 encoded string. This commit re-uses the already defined RabbitMQ target addresses: ``` "/topic/" RK Publish to amq.topic with routing key RK "/exchange/" X "/" RK Publish to exchange X with routing key RK ``` By default, the MQTT topic exchange is configure dto be amq.topic using the 1st target address. When an operator modifies the mqtt.exchange, the 2nd target address is used. * Apply PR feedback and fix formatting Co-authored-by: Michael Davis <mcarsondavis@gmail.com> * tidy up * Add MQTT message_containers test * consistent hash exchange: avoid amqp legacy conversion When hashing on a header value. * Avoid converting to amqp legacy when using exchange federation * Fix test flake * test and dialyzer fixes * dialyzer fix * Add MQTT protocol interoperability tests Test receiving from and sending to MQTT 5.0 and * AMQP 0.9.1 * AMQP 1.0 * STOMP * Streams * Regenerate portions of deps/rabbit/app.bzl with gazelle I'm not exactly sure how this happened, but gazell seems to have been run with an older version of the rules_erlang gazelle extension at some point. This caused generation of a structure that is no longer used. This commit updates the structure to the current pattern. * mc: refactoring * mc_amqpl: handle delivery annotations Just in case they are included. Also use iolist_to_iovec to create flat list of binaries when converting from amqp with amqp encoded payload. --------- Co-authored-by: David Ansari <david.ansari@gmx.de> Co-authored-by: Michael Davis <mcarsondavis@gmail.com> Co-authored-by: Rin Kuryloski <kuryloskip@vmware.com>
2023-08-31 18:27:13 +08:00
message(IsPersistent, PayloadFun, N) ->
#basic_message{content = Content,
exchange_name = Ex,
id = Id} =
rabbit_basic:message(rabbit_misc:r(<<>>, exchange, <<>>),
<<>>, #'P_basic'{delivery_mode = case IsPersistent of
true -> 2;
false -> 1
end},
PayloadFun(N)),
{ok, Msg} = mc_amqpl:message(Ex, <<>>, Content, #{id => Id}),
Msg.