2020-07-14 08:51:47 +08:00
|
|
|
%% 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/.
|
2020-05-12 22:42:00 +08:00
|
|
|
%%
|
2024-01-20 08:53:28 +08:00
|
|
|
%% Copyright (c) 2007-2025 Broadcom. All Rights Reserved. The term “Broadcom” refers to Broadcom Inc. and/or its subsidiaries. All rights reserved.
|
2020-05-12 22:42:00 +08:00
|
|
|
%%
|
|
|
|
|
|
|
|
|
|
-module(rabbit_maintenance).
|
|
|
|
|
|
2021-03-29 17:01:43 +08:00
|
|
|
-include_lib("rabbit_common/include/rabbit.hrl").
|
2025-07-11 18:37:10 +08:00
|
|
|
-include_lib("kernel/include/logger.hrl").
|
2020-07-09 13:26:52 +08:00
|
|
|
|
2024-01-30 01:04:35 +08:00
|
|
|
%% FIXME: Ra consistent queries are currently fragile in the sense that the
|
|
|
|
|
%% query function may run on a remote node and the function reference or MFA
|
|
|
|
|
%% may not be valid on that node. That's why consistent queries in this module
|
|
|
|
|
%% are in fact local queries when Khepri is enabled.
|
|
|
|
|
%%
|
|
|
|
|
%% See `rabbit_db_maintenance:get_consistent_in_khepri/1'.
|
|
|
|
|
|
2020-07-09 13:26:52 +08:00
|
|
|
-export([
|
2020-07-14 08:41:25 +08:00
|
|
|
is_enabled/0,
|
2020-07-09 13:26:52 +08:00
|
|
|
drain/0,
|
|
|
|
|
revive/0,
|
|
|
|
|
mark_as_being_drained/0,
|
|
|
|
|
unmark_as_being_drained/0,
|
|
|
|
|
is_being_drained_local_read/1,
|
|
|
|
|
is_being_drained_consistent_read/1,
|
|
|
|
|
status_local_read/1,
|
|
|
|
|
status_consistent_read/1,
|
|
|
|
|
filter_out_drained_nodes_local_read/1,
|
|
|
|
|
filter_out_drained_nodes_consistent_read/1,
|
|
|
|
|
suspend_all_client_listeners/0,
|
|
|
|
|
resume_all_client_listeners/0,
|
|
|
|
|
close_all_client_connections/0,
|
|
|
|
|
primary_replica_transfer_candidate_nodes/0,
|
2021-01-26 19:47:15 +08:00
|
|
|
random_primary_replica_transfer_candidate_node/2,
|
Allow to use Khepri database to store metadata instead of Mnesia
[Why]
Mnesia is a very powerful and convenient tool for Erlang applications:
it is a persistent disc-based database, it handles replication accross
multiple Erlang nodes and it is available out-of-the-box from the
Erlang/OTP distribution. RabbitMQ relies on Mnesia to manage all its
metadata:
* virtual hosts' properties
* intenal users
* queue, exchange and binding declarations (not queues data)
* runtime parameters and policies
* ...
Unfortunately Mnesia makes it difficult to handle network partition and,
as a consequence, the merge conflicts between Erlang nodes once the
network partition is resolved. RabbitMQ provides several partition
handling strategies but they are not bullet-proof. Users still hit
situations where it is a pain to repair a cluster following a network
partition.
[How]
@kjnilsson created Ra [1], a Raft consensus library that RabbitMQ
already uses successfully to implement quorum queues and streams for
instance. Those queues do not suffer from network partitions.
We created Khepri [2], a new persistent and replicated database engine
based on Ra and we want to use it in place of Mnesia in RabbitMQ to
solve the problems with network partitions.
This patch integrates Khepri as an experimental feature. When enabled,
RabbitMQ will store all its metadata in Khepri instead of Mnesia.
This change comes with behavior changes. While Khepri remains disabled,
you should see no changes to the behavior of RabbitMQ. If there are
changes, it is a bug. After Khepri is enabled, there are significant
changes of behavior that you should be aware of.
Because it is based on the Raft consensus algorithm, when there is a
network partition, only the cluster members that are in the partition
with at least `(Number of nodes in the cluster ÷ 2) + 1` number of nodes
can "make progress". In other words, only those nodes may write to the
Khepri database and read from the database and expect a consistent
result.
For instance in a cluster of 5 RabbitMQ nodes:
* If there are two partitions, one with 3 nodes, one with 2 nodes, only
the group of 3 nodes will be able to write to the database.
* If there are three partitions, two with 2 nodes, one with 1 node, none
of the group can write to the database.
Because the Khepri database will be used for all kind of metadata, it
means that RabbitMQ nodes that can't write to the database will be
unable to perform some operations. A list of operations and what to
expect is documented in the associated pull request and the RabbitMQ
website.
This requirement from Raft also affects the startup of RabbitMQ nodes in
a cluster. Indeed, at least a quorum number of nodes must be started at
once to allow nodes to become ready.
To enable Khepri, you need to enable the `khepri_db` feature flag:
rabbitmqctl enable_feature_flag khepri_db
When the `khepri_db` feature flag is enabled, the migration code
performs the following two tasks:
1. It synchronizes the Khepri cluster membership from the Mnesia
cluster. It uses `mnesia_to_khepri:sync_cluster_membership/1` from
the `khepri_mnesia_migration` application [3].
2. It copies data from relevant Mnesia tables to Khepri, doing some
conversion if necessary on the way. Again, it uses
`mnesia_to_khepri:copy_tables/4` from `khepri_mnesia_migration` to do
it.
This can be performed on a running standalone RabbitMQ node or cluster.
Data will be migrated from Mnesia to Khepri without any service
interruption. Note that during the migration, the performance may
decrease and the memory footprint may go up.
Because this feature flag is considered experimental, it is not enabled
by default even on a brand new RabbitMQ deployment.
More about the implementation details below:
In the past months, all accesses to Mnesia were isolated in a collection
of `rabbit_db*` modules. This is where the integration of Khepri mostly
takes place: we use a function called `rabbit_khepri:handle_fallback/1`
which selects the database and perform the query or the transaction.
Here is an example from `rabbit_db_vhost`:
* Up until RabbitMQ 3.12.x:
get(VHostName) when is_binary(VHostName) ->
get_in_mnesia(VHostName).
* Starting with RabbitMQ 3.13.0:
get(VHostName) when is_binary(VHostName) ->
rabbit_khepri:handle_fallback(
#{mnesia => fun() -> get_in_mnesia(VHostName) end,
khepri => fun() -> get_in_khepri(VHostName) end}).
This `rabbit_khepri:handle_fallback/1` function relies on two things:
1. the fact that the `khepri_db` feature flag is enabled, in which case
it always executes the Khepri-based variant.
4. the ability or not to read and write to Mnesia tables otherwise.
Before the feature flag is enabled, or during the migration, the
function will try to execute the Mnesia-based variant. If it succeeds,
then it returns the result. If it fails because one or more Mnesia
tables can't be used, it restarts from scratch: it means the feature
flag is being enabled and depending on the outcome, either the
Mnesia-based variant will succeed (the feature flag couldn't be enabled)
or the feature flag will be marked as enabled and it will call the
Khepri-based variant. The meat of this function really lives in the
`khepri_mnesia_migration` application [3] and
`rabbit_khepri:handle_fallback/1` is a wrapper on top of it that knows
about the feature flag.
However, some calls to the database do not depend on the existence of
Mnesia tables, such as functions where we need to learn about the
members of a cluster. For those, we can't rely on exceptions from
Mnesia. Therefore, we just look at the state of the feature flag to
determine which database to use. There are two situations though:
* Sometimes, we need the feature flag state query to block because the
function interested in it can't return a valid answer during the
migration. Here is an example:
case rabbit_khepri:is_enabled(RemoteNode) of
true -> can_join_using_khepri(RemoteNode);
false -> can_join_using_mnesia(RemoteNode)
end
* Sometimes, we need the feature flag state query to NOT block (for
instance because it would cause a deadlock). Here is an example:
case rabbit_khepri:get_feature_state() of
enabled -> members_using_khepri();
_ -> members_using_mnesia()
end
Direct accesses to Mnesia still exists. They are limited to code that is
specific to Mnesia such as classic queue mirroring or network partitions
handling strategies.
Now, to discover the Mnesia tables to migrate and how to migrate them,
we use an Erlang module attribute called
`rabbit_mnesia_tables_to_khepri_db` which indicates a list of Mnesia
tables and an associated converter module. Here is an example in the
`rabbitmq_recent_history_exchange` plugin:
-rabbit_mnesia_tables_to_khepri_db(
[{?RH_TABLE, rabbit_db_rh_exchange_m2k_converter}]).
The converter module — `rabbit_db_rh_exchange_m2k_converter` in this
example — is is fact a "sub" converter module called but
`rabbit_db_m2k_converter`. See the documentation of a `mnesia_to_khepri`
converter module to learn more about these modules.
[1] https://github.com/rabbitmq/ra
[2] https://github.com/rabbitmq/khepri
[3] https://github.com/rabbitmq/khepri_mnesia_migration
See #7206.
Co-authored-by: Jean-Sébastien Pédron <jean-sebastien@rabbitmq.com>
Co-authored-by: Diana Parra Corbacho <dparracorbac@vmware.com>
Co-authored-by: Michael Davis <mcarsondavis@gmail.com>
2023-01-05 20:57:50 +08:00
|
|
|
table_definitions/0
|
2020-07-14 08:41:25 +08:00
|
|
|
]).
|
2020-07-09 13:26:52 +08:00
|
|
|
|
|
|
|
|
-define(DEFAULT_STATUS, regular).
|
|
|
|
|
-define(DRAINING_STATUS, draining).
|
|
|
|
|
|
|
|
|
|
-type maintenance_status() :: ?DEFAULT_STATUS | ?DRAINING_STATUS.
|
|
|
|
|
|
|
|
|
|
-export_type([
|
|
|
|
|
maintenance_status/0
|
|
|
|
|
]).
|
|
|
|
|
|
2022-05-08 01:33:09 +08:00
|
|
|
%%
|
|
|
|
|
%% Boot
|
|
|
|
|
%%
|
|
|
|
|
|
Allow to use Khepri database to store metadata instead of Mnesia
[Why]
Mnesia is a very powerful and convenient tool for Erlang applications:
it is a persistent disc-based database, it handles replication accross
multiple Erlang nodes and it is available out-of-the-box from the
Erlang/OTP distribution. RabbitMQ relies on Mnesia to manage all its
metadata:
* virtual hosts' properties
* intenal users
* queue, exchange and binding declarations (not queues data)
* runtime parameters and policies
* ...
Unfortunately Mnesia makes it difficult to handle network partition and,
as a consequence, the merge conflicts between Erlang nodes once the
network partition is resolved. RabbitMQ provides several partition
handling strategies but they are not bullet-proof. Users still hit
situations where it is a pain to repair a cluster following a network
partition.
[How]
@kjnilsson created Ra [1], a Raft consensus library that RabbitMQ
already uses successfully to implement quorum queues and streams for
instance. Those queues do not suffer from network partitions.
We created Khepri [2], a new persistent and replicated database engine
based on Ra and we want to use it in place of Mnesia in RabbitMQ to
solve the problems with network partitions.
This patch integrates Khepri as an experimental feature. When enabled,
RabbitMQ will store all its metadata in Khepri instead of Mnesia.
This change comes with behavior changes. While Khepri remains disabled,
you should see no changes to the behavior of RabbitMQ. If there are
changes, it is a bug. After Khepri is enabled, there are significant
changes of behavior that you should be aware of.
Because it is based on the Raft consensus algorithm, when there is a
network partition, only the cluster members that are in the partition
with at least `(Number of nodes in the cluster ÷ 2) + 1` number of nodes
can "make progress". In other words, only those nodes may write to the
Khepri database and read from the database and expect a consistent
result.
For instance in a cluster of 5 RabbitMQ nodes:
* If there are two partitions, one with 3 nodes, one with 2 nodes, only
the group of 3 nodes will be able to write to the database.
* If there are three partitions, two with 2 nodes, one with 1 node, none
of the group can write to the database.
Because the Khepri database will be used for all kind of metadata, it
means that RabbitMQ nodes that can't write to the database will be
unable to perform some operations. A list of operations and what to
expect is documented in the associated pull request and the RabbitMQ
website.
This requirement from Raft also affects the startup of RabbitMQ nodes in
a cluster. Indeed, at least a quorum number of nodes must be started at
once to allow nodes to become ready.
To enable Khepri, you need to enable the `khepri_db` feature flag:
rabbitmqctl enable_feature_flag khepri_db
When the `khepri_db` feature flag is enabled, the migration code
performs the following two tasks:
1. It synchronizes the Khepri cluster membership from the Mnesia
cluster. It uses `mnesia_to_khepri:sync_cluster_membership/1` from
the `khepri_mnesia_migration` application [3].
2. It copies data from relevant Mnesia tables to Khepri, doing some
conversion if necessary on the way. Again, it uses
`mnesia_to_khepri:copy_tables/4` from `khepri_mnesia_migration` to do
it.
This can be performed on a running standalone RabbitMQ node or cluster.
Data will be migrated from Mnesia to Khepri without any service
interruption. Note that during the migration, the performance may
decrease and the memory footprint may go up.
Because this feature flag is considered experimental, it is not enabled
by default even on a brand new RabbitMQ deployment.
More about the implementation details below:
In the past months, all accesses to Mnesia were isolated in a collection
of `rabbit_db*` modules. This is where the integration of Khepri mostly
takes place: we use a function called `rabbit_khepri:handle_fallback/1`
which selects the database and perform the query or the transaction.
Here is an example from `rabbit_db_vhost`:
* Up until RabbitMQ 3.12.x:
get(VHostName) when is_binary(VHostName) ->
get_in_mnesia(VHostName).
* Starting with RabbitMQ 3.13.0:
get(VHostName) when is_binary(VHostName) ->
rabbit_khepri:handle_fallback(
#{mnesia => fun() -> get_in_mnesia(VHostName) end,
khepri => fun() -> get_in_khepri(VHostName) end}).
This `rabbit_khepri:handle_fallback/1` function relies on two things:
1. the fact that the `khepri_db` feature flag is enabled, in which case
it always executes the Khepri-based variant.
4. the ability or not to read and write to Mnesia tables otherwise.
Before the feature flag is enabled, or during the migration, the
function will try to execute the Mnesia-based variant. If it succeeds,
then it returns the result. If it fails because one or more Mnesia
tables can't be used, it restarts from scratch: it means the feature
flag is being enabled and depending on the outcome, either the
Mnesia-based variant will succeed (the feature flag couldn't be enabled)
or the feature flag will be marked as enabled and it will call the
Khepri-based variant. The meat of this function really lives in the
`khepri_mnesia_migration` application [3] and
`rabbit_khepri:handle_fallback/1` is a wrapper on top of it that knows
about the feature flag.
However, some calls to the database do not depend on the existence of
Mnesia tables, such as functions where we need to learn about the
members of a cluster. For those, we can't rely on exceptions from
Mnesia. Therefore, we just look at the state of the feature flag to
determine which database to use. There are two situations though:
* Sometimes, we need the feature flag state query to block because the
function interested in it can't return a valid answer during the
migration. Here is an example:
case rabbit_khepri:is_enabled(RemoteNode) of
true -> can_join_using_khepri(RemoteNode);
false -> can_join_using_mnesia(RemoteNode)
end
* Sometimes, we need the feature flag state query to NOT block (for
instance because it would cause a deadlock). Here is an example:
case rabbit_khepri:get_feature_state() of
enabled -> members_using_khepri();
_ -> members_using_mnesia()
end
Direct accesses to Mnesia still exists. They are limited to code that is
specific to Mnesia such as classic queue mirroring or network partitions
handling strategies.
Now, to discover the Mnesia tables to migrate and how to migrate them,
we use an Erlang module attribute called
`rabbit_mnesia_tables_to_khepri_db` which indicates a list of Mnesia
tables and an associated converter module. Here is an example in the
`rabbitmq_recent_history_exchange` plugin:
-rabbit_mnesia_tables_to_khepri_db(
[{?RH_TABLE, rabbit_db_rh_exchange_m2k_converter}]).
The converter module — `rabbit_db_rh_exchange_m2k_converter` in this
example — is is fact a "sub" converter module called but
`rabbit_db_m2k_converter`. See the documentation of a `mnesia_to_khepri`
converter module to learn more about these modules.
[1] https://github.com/rabbitmq/ra
[2] https://github.com/rabbitmq/khepri
[3] https://github.com/rabbitmq/khepri_mnesia_migration
See #7206.
Co-authored-by: Jean-Sébastien Pédron <jean-sebastien@rabbitmq.com>
Co-authored-by: Diana Parra Corbacho <dparracorbac@vmware.com>
Co-authored-by: Michael Davis <mcarsondavis@gmail.com>
2023-01-05 20:57:50 +08:00
|
|
|
table_definitions() ->
|
|
|
|
|
rabbit_db_maintenance:table_definitions().
|
2022-05-08 01:33:09 +08:00
|
|
|
|
2020-05-12 22:42:00 +08:00
|
|
|
%%
|
|
|
|
|
%% API
|
|
|
|
|
%%
|
|
|
|
|
|
2020-07-14 08:41:25 +08:00
|
|
|
-spec is_enabled() -> boolean().
|
|
|
|
|
is_enabled() ->
|
2022-07-15 23:23:18 +08:00
|
|
|
true.
|
2020-07-14 08:41:25 +08:00
|
|
|
|
2020-06-25 22:59:08 +08:00
|
|
|
-spec drain() -> ok.
|
2020-06-16 20:03:25 +08:00
|
|
|
drain() ->
|
2025-07-11 18:37:10 +08:00
|
|
|
?LOG_WARNING("This node is being put into maintenance (drain) mode"),
|
2020-06-16 20:03:25 +08:00
|
|
|
mark_as_being_drained(),
|
2025-07-11 18:37:10 +08:00
|
|
|
?LOG_INFO("Marked this node as undergoing maintenance"),
|
2023-01-24 05:26:14 +08:00
|
|
|
_ = suspend_all_client_listeners(),
|
2025-07-11 18:37:10 +08:00
|
|
|
?LOG_WARNING("Suspended all listeners and will no longer accept client connections"),
|
2020-06-16 20:03:25 +08:00
|
|
|
{ok, NConnections} = close_all_client_connections(),
|
2025-07-11 18:37:10 +08:00
|
|
|
?LOG_WARNING("Closed ~b local client connections", [NConnections]),
|
2020-07-09 00:10:55 +08:00
|
|
|
%% allow plugins to react e.g. by closing their protocol connections
|
|
|
|
|
rabbit_event:notify(maintenance_connections_closed, #{
|
|
|
|
|
reason => <<"node is being put into maintenance">>
|
|
|
|
|
}),
|
2020-06-16 20:03:25 +08:00
|
|
|
|
|
|
|
|
TransferCandidates = primary_replica_transfer_candidate_nodes(),
|
2021-01-27 23:52:17 +08:00
|
|
|
%% Note: only QQ leadership is transferred because it is a reasonably quick thing to do a lot of queues
|
|
|
|
|
%% in the cluster, unlike with CMQs.
|
2025-02-03 22:59:34 +08:00
|
|
|
rabbit_queue_type:drain(TransferCandidates),
|
2022-12-02 21:55:06 +08:00
|
|
|
|
Allow to use Khepri database to store metadata instead of Mnesia
[Why]
Mnesia is a very powerful and convenient tool for Erlang applications:
it is a persistent disc-based database, it handles replication accross
multiple Erlang nodes and it is available out-of-the-box from the
Erlang/OTP distribution. RabbitMQ relies on Mnesia to manage all its
metadata:
* virtual hosts' properties
* intenal users
* queue, exchange and binding declarations (not queues data)
* runtime parameters and policies
* ...
Unfortunately Mnesia makes it difficult to handle network partition and,
as a consequence, the merge conflicts between Erlang nodes once the
network partition is resolved. RabbitMQ provides several partition
handling strategies but they are not bullet-proof. Users still hit
situations where it is a pain to repair a cluster following a network
partition.
[How]
@kjnilsson created Ra [1], a Raft consensus library that RabbitMQ
already uses successfully to implement quorum queues and streams for
instance. Those queues do not suffer from network partitions.
We created Khepri [2], a new persistent and replicated database engine
based on Ra and we want to use it in place of Mnesia in RabbitMQ to
solve the problems with network partitions.
This patch integrates Khepri as an experimental feature. When enabled,
RabbitMQ will store all its metadata in Khepri instead of Mnesia.
This change comes with behavior changes. While Khepri remains disabled,
you should see no changes to the behavior of RabbitMQ. If there are
changes, it is a bug. After Khepri is enabled, there are significant
changes of behavior that you should be aware of.
Because it is based on the Raft consensus algorithm, when there is a
network partition, only the cluster members that are in the partition
with at least `(Number of nodes in the cluster ÷ 2) + 1` number of nodes
can "make progress". In other words, only those nodes may write to the
Khepri database and read from the database and expect a consistent
result.
For instance in a cluster of 5 RabbitMQ nodes:
* If there are two partitions, one with 3 nodes, one with 2 nodes, only
the group of 3 nodes will be able to write to the database.
* If there are three partitions, two with 2 nodes, one with 1 node, none
of the group can write to the database.
Because the Khepri database will be used for all kind of metadata, it
means that RabbitMQ nodes that can't write to the database will be
unable to perform some operations. A list of operations and what to
expect is documented in the associated pull request and the RabbitMQ
website.
This requirement from Raft also affects the startup of RabbitMQ nodes in
a cluster. Indeed, at least a quorum number of nodes must be started at
once to allow nodes to become ready.
To enable Khepri, you need to enable the `khepri_db` feature flag:
rabbitmqctl enable_feature_flag khepri_db
When the `khepri_db` feature flag is enabled, the migration code
performs the following two tasks:
1. It synchronizes the Khepri cluster membership from the Mnesia
cluster. It uses `mnesia_to_khepri:sync_cluster_membership/1` from
the `khepri_mnesia_migration` application [3].
2. It copies data from relevant Mnesia tables to Khepri, doing some
conversion if necessary on the way. Again, it uses
`mnesia_to_khepri:copy_tables/4` from `khepri_mnesia_migration` to do
it.
This can be performed on a running standalone RabbitMQ node or cluster.
Data will be migrated from Mnesia to Khepri without any service
interruption. Note that during the migration, the performance may
decrease and the memory footprint may go up.
Because this feature flag is considered experimental, it is not enabled
by default even on a brand new RabbitMQ deployment.
More about the implementation details below:
In the past months, all accesses to Mnesia were isolated in a collection
of `rabbit_db*` modules. This is where the integration of Khepri mostly
takes place: we use a function called `rabbit_khepri:handle_fallback/1`
which selects the database and perform the query or the transaction.
Here is an example from `rabbit_db_vhost`:
* Up until RabbitMQ 3.12.x:
get(VHostName) when is_binary(VHostName) ->
get_in_mnesia(VHostName).
* Starting with RabbitMQ 3.13.0:
get(VHostName) when is_binary(VHostName) ->
rabbit_khepri:handle_fallback(
#{mnesia => fun() -> get_in_mnesia(VHostName) end,
khepri => fun() -> get_in_khepri(VHostName) end}).
This `rabbit_khepri:handle_fallback/1` function relies on two things:
1. the fact that the `khepri_db` feature flag is enabled, in which case
it always executes the Khepri-based variant.
4. the ability or not to read and write to Mnesia tables otherwise.
Before the feature flag is enabled, or during the migration, the
function will try to execute the Mnesia-based variant. If it succeeds,
then it returns the result. If it fails because one or more Mnesia
tables can't be used, it restarts from scratch: it means the feature
flag is being enabled and depending on the outcome, either the
Mnesia-based variant will succeed (the feature flag couldn't be enabled)
or the feature flag will be marked as enabled and it will call the
Khepri-based variant. The meat of this function really lives in the
`khepri_mnesia_migration` application [3] and
`rabbit_khepri:handle_fallback/1` is a wrapper on top of it that knows
about the feature flag.
However, some calls to the database do not depend on the existence of
Mnesia tables, such as functions where we need to learn about the
members of a cluster. For those, we can't rely on exceptions from
Mnesia. Therefore, we just look at the state of the feature flag to
determine which database to use. There are two situations though:
* Sometimes, we need the feature flag state query to block because the
function interested in it can't return a valid answer during the
migration. Here is an example:
case rabbit_khepri:is_enabled(RemoteNode) of
true -> can_join_using_khepri(RemoteNode);
false -> can_join_using_mnesia(RemoteNode)
end
* Sometimes, we need the feature flag state query to NOT block (for
instance because it would cause a deadlock). Here is an example:
case rabbit_khepri:get_feature_state() of
enabled -> members_using_khepri();
_ -> members_using_mnesia()
end
Direct accesses to Mnesia still exists. They are limited to code that is
specific to Mnesia such as classic queue mirroring or network partitions
handling strategies.
Now, to discover the Mnesia tables to migrate and how to migrate them,
we use an Erlang module attribute called
`rabbit_mnesia_tables_to_khepri_db` which indicates a list of Mnesia
tables and an associated converter module. Here is an example in the
`rabbitmq_recent_history_exchange` plugin:
-rabbit_mnesia_tables_to_khepri_db(
[{?RH_TABLE, rabbit_db_rh_exchange_m2k_converter}]).
The converter module — `rabbit_db_rh_exchange_m2k_converter` in this
example — is is fact a "sub" converter module called but
`rabbit_db_m2k_converter`. See the documentation of a `mnesia_to_khepri`
converter module to learn more about these modules.
[1] https://github.com/rabbitmq/ra
[2] https://github.com/rabbitmq/khepri
[3] https://github.com/rabbitmq/khepri_mnesia_migration
See #7206.
Co-authored-by: Jean-Sébastien Pédron <jean-sebastien@rabbitmq.com>
Co-authored-by: Diana Parra Corbacho <dparracorbac@vmware.com>
Co-authored-by: Michael Davis <mcarsondavis@gmail.com>
2023-01-05 20:57:50 +08:00
|
|
|
transfer_leadership_of_metadata_store(TransferCandidates),
|
|
|
|
|
|
2020-07-09 00:10:55 +08:00
|
|
|
%% allow plugins to react
|
|
|
|
|
rabbit_event:notify(maintenance_draining, #{
|
|
|
|
|
reason => <<"node is being put into maintenance">>
|
|
|
|
|
}),
|
2025-07-11 18:37:10 +08:00
|
|
|
?LOG_INFO("Node is ready to be shut down for maintenance or upgrade"),
|
2020-06-16 20:03:25 +08:00
|
|
|
|
|
|
|
|
ok.
|
|
|
|
|
|
2020-06-25 22:59:08 +08:00
|
|
|
-spec revive() -> ok.
|
2020-06-16 20:03:25 +08:00
|
|
|
revive() ->
|
2025-07-11 18:37:10 +08:00
|
|
|
?LOG_INFO("This node is being revived from maintenance (drain) mode"),
|
2025-02-03 22:59:34 +08:00
|
|
|
rabbit_queue_type:revive(),
|
2025-07-11 18:37:10 +08:00
|
|
|
?LOG_INFO("Resumed all listeners and will accept client connections again"),
|
2023-01-24 05:26:14 +08:00
|
|
|
_ = resume_all_client_listeners(),
|
2025-07-11 18:37:10 +08:00
|
|
|
?LOG_INFO("Resumed all listeners and will accept client connections again"),
|
2020-06-16 20:03:25 +08:00
|
|
|
unmark_as_being_drained(),
|
2025-07-11 18:37:10 +08:00
|
|
|
?LOG_INFO("Marked this node as back from maintenance and ready to serve clients"),
|
2020-06-16 20:03:25 +08:00
|
|
|
|
2020-07-09 00:10:55 +08:00
|
|
|
%% allow plugins to react
|
|
|
|
|
rabbit_event:notify(maintenance_revived, #{}),
|
2020-06-16 20:03:25 +08:00
|
|
|
|
2020-07-14 08:41:25 +08:00
|
|
|
ok.
|
2023-01-02 12:17:36 +08:00
|
|
|
|
2020-06-08 18:37:54 +08:00
|
|
|
-spec mark_as_being_drained() -> boolean().
|
|
|
|
|
mark_as_being_drained() ->
|
2025-07-11 18:37:10 +08:00
|
|
|
?LOG_DEBUG("Marking the node as undergoing maintenance"),
|
2023-01-16 18:57:58 +08:00
|
|
|
rabbit_db_maintenance:set(?DRAINING_STATUS).
|
2023-01-02 12:17:36 +08:00
|
|
|
|
2020-06-08 18:37:54 +08:00
|
|
|
-spec unmark_as_being_drained() -> boolean().
|
|
|
|
|
unmark_as_being_drained() ->
|
2025-07-11 18:37:10 +08:00
|
|
|
?LOG_DEBUG("Unmarking the node as undergoing maintenance"),
|
2023-01-16 18:57:58 +08:00
|
|
|
rabbit_db_maintenance:set(?DEFAULT_STATUS).
|
2023-01-02 12:17:36 +08:00
|
|
|
|
2020-06-08 18:37:54 +08:00
|
|
|
-spec is_being_drained_local_read(node()) -> boolean().
|
|
|
|
|
is_being_drained_local_read(Node) ->
|
2020-07-09 13:26:52 +08:00
|
|
|
Status = status_local_read(Node),
|
|
|
|
|
Status =:= ?DRAINING_STATUS.
|
2020-05-30 00:21:25 +08:00
|
|
|
|
2020-06-08 18:37:54 +08:00
|
|
|
-spec is_being_drained_consistent_read(node()) -> boolean().
|
|
|
|
|
is_being_drained_consistent_read(Node) ->
|
2020-07-09 13:26:52 +08:00
|
|
|
Status = status_consistent_read(Node),
|
|
|
|
|
Status =:= ?DRAINING_STATUS.
|
|
|
|
|
|
|
|
|
|
-spec status_local_read(node()) -> maintenance_status().
|
|
|
|
|
status_local_read(Node) ->
|
2023-01-16 20:35:17 +08:00
|
|
|
case rabbit_db_maintenance:get(Node) of
|
|
|
|
|
undefined ->
|
|
|
|
|
?DEFAULT_STATUS;
|
|
|
|
|
Status ->
|
|
|
|
|
Status
|
|
|
|
|
end.
|
2023-01-02 12:17:36 +08:00
|
|
|
|
2020-07-09 13:26:52 +08:00
|
|
|
-spec status_consistent_read(node()) -> maintenance_status().
|
|
|
|
|
status_consistent_read(Node) ->
|
2023-01-16 20:35:17 +08:00
|
|
|
case rabbit_db_maintenance:get_consistent(Node) of
|
|
|
|
|
undefined ->
|
|
|
|
|
?DEFAULT_STATUS;
|
|
|
|
|
Status ->
|
|
|
|
|
Status
|
|
|
|
|
end.
|
2023-01-02 12:17:36 +08:00
|
|
|
|
2020-06-16 20:03:25 +08:00
|
|
|
-spec filter_out_drained_nodes_local_read([node()]) -> [node()].
|
|
|
|
|
filter_out_drained_nodes_local_read(Nodes) ->
|
2020-06-16 04:12:51 +08:00
|
|
|
lists:filter(fun(N) -> not is_being_drained_local_read(N) end, Nodes).
|
2023-01-02 12:17:36 +08:00
|
|
|
|
2020-06-16 20:03:25 +08:00
|
|
|
-spec filter_out_drained_nodes_consistent_read([node()]) -> [node()].
|
|
|
|
|
filter_out_drained_nodes_consistent_read(Nodes) ->
|
|
|
|
|
lists:filter(fun(N) -> not is_being_drained_consistent_read(N) end, Nodes).
|
2023-01-02 12:17:36 +08:00
|
|
|
|
2020-05-19 16:24:13 +08:00
|
|
|
-spec suspend_all_client_listeners() -> rabbit_types:ok_or_error(any()).
|
2020-05-19 08:50:17 +08:00
|
|
|
%% Pauses all listeners on the current node except for
|
|
|
|
|
%% Erlang distribution (clustering and CLI tools).
|
2021-02-01 20:05:07 +08:00
|
|
|
%% A resumed listener will not accept any new client connections
|
2020-05-19 08:50:17 +08:00
|
|
|
%% but previously established connections won't be interrupted.
|
|
|
|
|
suspend_all_client_listeners() ->
|
|
|
|
|
Listeners = rabbit_networking:node_client_listeners(node()),
|
2025-07-11 18:37:10 +08:00
|
|
|
?LOG_INFO("Asked to suspend ~b client connection listeners. "
|
2020-05-19 16:24:13 +08:00
|
|
|
"No new client connections will be accepted until these listeners are resumed!", [length(Listeners)]),
|
|
|
|
|
Results = lists:foldl(local_listener_fold_fun(fun ranch:suspend_listener/1), [], Listeners),
|
|
|
|
|
lists:foldl(fun ok_or_first_error/2, ok, Results).
|
2020-05-12 22:42:00 +08:00
|
|
|
|
2020-05-19 16:24:13 +08:00
|
|
|
-spec resume_all_client_listeners() -> rabbit_types:ok_or_error(any()).
|
2020-05-19 08:50:17 +08:00
|
|
|
%% Resumes all listeners on the current node except for
|
|
|
|
|
%% Erlang distribution (clustering and CLI tools).
|
|
|
|
|
%% A resumed listener will accept new client connections.
|
|
|
|
|
resume_all_client_listeners() ->
|
|
|
|
|
Listeners = rabbit_networking:node_client_listeners(node()),
|
2025-07-11 18:37:10 +08:00
|
|
|
?LOG_INFO("Asked to resume ~b client connection listeners. "
|
2020-05-19 16:24:13 +08:00
|
|
|
"New client connections will be accepted from now on", [length(Listeners)]),
|
|
|
|
|
Results = lists:foldl(local_listener_fold_fun(fun ranch:resume_listener/1), [], Listeners),
|
|
|
|
|
lists:foldl(fun ok_or_first_error/2, ok, Results).
|
2020-05-12 22:42:00 +08:00
|
|
|
|
2020-06-16 20:03:25 +08:00
|
|
|
-spec close_all_client_connections() -> {'ok', non_neg_integer()}.
|
2020-05-12 22:42:00 +08:00
|
|
|
close_all_client_connections() ->
|
2020-06-16 20:03:25 +08:00
|
|
|
Pids = rabbit_networking:local_connections(),
|
|
|
|
|
rabbit_networking:close_connections(Pids, "Node was put into maintenance mode"),
|
|
|
|
|
{ok, length(Pids)}.
|
2020-05-19 16:24:13 +08:00
|
|
|
|
Allow to use Khepri database to store metadata instead of Mnesia
[Why]
Mnesia is a very powerful and convenient tool for Erlang applications:
it is a persistent disc-based database, it handles replication accross
multiple Erlang nodes and it is available out-of-the-box from the
Erlang/OTP distribution. RabbitMQ relies on Mnesia to manage all its
metadata:
* virtual hosts' properties
* intenal users
* queue, exchange and binding declarations (not queues data)
* runtime parameters and policies
* ...
Unfortunately Mnesia makes it difficult to handle network partition and,
as a consequence, the merge conflicts between Erlang nodes once the
network partition is resolved. RabbitMQ provides several partition
handling strategies but they are not bullet-proof. Users still hit
situations where it is a pain to repair a cluster following a network
partition.
[How]
@kjnilsson created Ra [1], a Raft consensus library that RabbitMQ
already uses successfully to implement quorum queues and streams for
instance. Those queues do not suffer from network partitions.
We created Khepri [2], a new persistent and replicated database engine
based on Ra and we want to use it in place of Mnesia in RabbitMQ to
solve the problems with network partitions.
This patch integrates Khepri as an experimental feature. When enabled,
RabbitMQ will store all its metadata in Khepri instead of Mnesia.
This change comes with behavior changes. While Khepri remains disabled,
you should see no changes to the behavior of RabbitMQ. If there are
changes, it is a bug. After Khepri is enabled, there are significant
changes of behavior that you should be aware of.
Because it is based on the Raft consensus algorithm, when there is a
network partition, only the cluster members that are in the partition
with at least `(Number of nodes in the cluster ÷ 2) + 1` number of nodes
can "make progress". In other words, only those nodes may write to the
Khepri database and read from the database and expect a consistent
result.
For instance in a cluster of 5 RabbitMQ nodes:
* If there are two partitions, one with 3 nodes, one with 2 nodes, only
the group of 3 nodes will be able to write to the database.
* If there are three partitions, two with 2 nodes, one with 1 node, none
of the group can write to the database.
Because the Khepri database will be used for all kind of metadata, it
means that RabbitMQ nodes that can't write to the database will be
unable to perform some operations. A list of operations and what to
expect is documented in the associated pull request and the RabbitMQ
website.
This requirement from Raft also affects the startup of RabbitMQ nodes in
a cluster. Indeed, at least a quorum number of nodes must be started at
once to allow nodes to become ready.
To enable Khepri, you need to enable the `khepri_db` feature flag:
rabbitmqctl enable_feature_flag khepri_db
When the `khepri_db` feature flag is enabled, the migration code
performs the following two tasks:
1. It synchronizes the Khepri cluster membership from the Mnesia
cluster. It uses `mnesia_to_khepri:sync_cluster_membership/1` from
the `khepri_mnesia_migration` application [3].
2. It copies data from relevant Mnesia tables to Khepri, doing some
conversion if necessary on the way. Again, it uses
`mnesia_to_khepri:copy_tables/4` from `khepri_mnesia_migration` to do
it.
This can be performed on a running standalone RabbitMQ node or cluster.
Data will be migrated from Mnesia to Khepri without any service
interruption. Note that during the migration, the performance may
decrease and the memory footprint may go up.
Because this feature flag is considered experimental, it is not enabled
by default even on a brand new RabbitMQ deployment.
More about the implementation details below:
In the past months, all accesses to Mnesia were isolated in a collection
of `rabbit_db*` modules. This is where the integration of Khepri mostly
takes place: we use a function called `rabbit_khepri:handle_fallback/1`
which selects the database and perform the query or the transaction.
Here is an example from `rabbit_db_vhost`:
* Up until RabbitMQ 3.12.x:
get(VHostName) when is_binary(VHostName) ->
get_in_mnesia(VHostName).
* Starting with RabbitMQ 3.13.0:
get(VHostName) when is_binary(VHostName) ->
rabbit_khepri:handle_fallback(
#{mnesia => fun() -> get_in_mnesia(VHostName) end,
khepri => fun() -> get_in_khepri(VHostName) end}).
This `rabbit_khepri:handle_fallback/1` function relies on two things:
1. the fact that the `khepri_db` feature flag is enabled, in which case
it always executes the Khepri-based variant.
4. the ability or not to read and write to Mnesia tables otherwise.
Before the feature flag is enabled, or during the migration, the
function will try to execute the Mnesia-based variant. If it succeeds,
then it returns the result. If it fails because one or more Mnesia
tables can't be used, it restarts from scratch: it means the feature
flag is being enabled and depending on the outcome, either the
Mnesia-based variant will succeed (the feature flag couldn't be enabled)
or the feature flag will be marked as enabled and it will call the
Khepri-based variant. The meat of this function really lives in the
`khepri_mnesia_migration` application [3] and
`rabbit_khepri:handle_fallback/1` is a wrapper on top of it that knows
about the feature flag.
However, some calls to the database do not depend on the existence of
Mnesia tables, such as functions where we need to learn about the
members of a cluster. For those, we can't rely on exceptions from
Mnesia. Therefore, we just look at the state of the feature flag to
determine which database to use. There are two situations though:
* Sometimes, we need the feature flag state query to block because the
function interested in it can't return a valid answer during the
migration. Here is an example:
case rabbit_khepri:is_enabled(RemoteNode) of
true -> can_join_using_khepri(RemoteNode);
false -> can_join_using_mnesia(RemoteNode)
end
* Sometimes, we need the feature flag state query to NOT block (for
instance because it would cause a deadlock). Here is an example:
case rabbit_khepri:get_feature_state() of
enabled -> members_using_khepri();
_ -> members_using_mnesia()
end
Direct accesses to Mnesia still exists. They are limited to code that is
specific to Mnesia such as classic queue mirroring or network partitions
handling strategies.
Now, to discover the Mnesia tables to migrate and how to migrate them,
we use an Erlang module attribute called
`rabbit_mnesia_tables_to_khepri_db` which indicates a list of Mnesia
tables and an associated converter module. Here is an example in the
`rabbitmq_recent_history_exchange` plugin:
-rabbit_mnesia_tables_to_khepri_db(
[{?RH_TABLE, rabbit_db_rh_exchange_m2k_converter}]).
The converter module — `rabbit_db_rh_exchange_m2k_converter` in this
example — is is fact a "sub" converter module called but
`rabbit_db_m2k_converter`. See the documentation of a `mnesia_to_khepri`
converter module to learn more about these modules.
[1] https://github.com/rabbitmq/ra
[2] https://github.com/rabbitmq/khepri
[3] https://github.com/rabbitmq/khepri_mnesia_migration
See #7206.
Co-authored-by: Jean-Sébastien Pédron <jean-sebastien@rabbitmq.com>
Co-authored-by: Diana Parra Corbacho <dparracorbac@vmware.com>
Co-authored-by: Michael Davis <mcarsondavis@gmail.com>
2023-01-05 20:57:50 +08:00
|
|
|
transfer_leadership_of_metadata_store(TransferCandidates) ->
|
2025-07-11 18:37:10 +08:00
|
|
|
?LOG_INFO("Will transfer leadership of metadata store with current leader on this node",
|
Allow to use Khepri database to store metadata instead of Mnesia
[Why]
Mnesia is a very powerful and convenient tool for Erlang applications:
it is a persistent disc-based database, it handles replication accross
multiple Erlang nodes and it is available out-of-the-box from the
Erlang/OTP distribution. RabbitMQ relies on Mnesia to manage all its
metadata:
* virtual hosts' properties
* intenal users
* queue, exchange and binding declarations (not queues data)
* runtime parameters and policies
* ...
Unfortunately Mnesia makes it difficult to handle network partition and,
as a consequence, the merge conflicts between Erlang nodes once the
network partition is resolved. RabbitMQ provides several partition
handling strategies but they are not bullet-proof. Users still hit
situations where it is a pain to repair a cluster following a network
partition.
[How]
@kjnilsson created Ra [1], a Raft consensus library that RabbitMQ
already uses successfully to implement quorum queues and streams for
instance. Those queues do not suffer from network partitions.
We created Khepri [2], a new persistent and replicated database engine
based on Ra and we want to use it in place of Mnesia in RabbitMQ to
solve the problems with network partitions.
This patch integrates Khepri as an experimental feature. When enabled,
RabbitMQ will store all its metadata in Khepri instead of Mnesia.
This change comes with behavior changes. While Khepri remains disabled,
you should see no changes to the behavior of RabbitMQ. If there are
changes, it is a bug. After Khepri is enabled, there are significant
changes of behavior that you should be aware of.
Because it is based on the Raft consensus algorithm, when there is a
network partition, only the cluster members that are in the partition
with at least `(Number of nodes in the cluster ÷ 2) + 1` number of nodes
can "make progress". In other words, only those nodes may write to the
Khepri database and read from the database and expect a consistent
result.
For instance in a cluster of 5 RabbitMQ nodes:
* If there are two partitions, one with 3 nodes, one with 2 nodes, only
the group of 3 nodes will be able to write to the database.
* If there are three partitions, two with 2 nodes, one with 1 node, none
of the group can write to the database.
Because the Khepri database will be used for all kind of metadata, it
means that RabbitMQ nodes that can't write to the database will be
unable to perform some operations. A list of operations and what to
expect is documented in the associated pull request and the RabbitMQ
website.
This requirement from Raft also affects the startup of RabbitMQ nodes in
a cluster. Indeed, at least a quorum number of nodes must be started at
once to allow nodes to become ready.
To enable Khepri, you need to enable the `khepri_db` feature flag:
rabbitmqctl enable_feature_flag khepri_db
When the `khepri_db` feature flag is enabled, the migration code
performs the following two tasks:
1. It synchronizes the Khepri cluster membership from the Mnesia
cluster. It uses `mnesia_to_khepri:sync_cluster_membership/1` from
the `khepri_mnesia_migration` application [3].
2. It copies data from relevant Mnesia tables to Khepri, doing some
conversion if necessary on the way. Again, it uses
`mnesia_to_khepri:copy_tables/4` from `khepri_mnesia_migration` to do
it.
This can be performed on a running standalone RabbitMQ node or cluster.
Data will be migrated from Mnesia to Khepri without any service
interruption. Note that during the migration, the performance may
decrease and the memory footprint may go up.
Because this feature flag is considered experimental, it is not enabled
by default even on a brand new RabbitMQ deployment.
More about the implementation details below:
In the past months, all accesses to Mnesia were isolated in a collection
of `rabbit_db*` modules. This is where the integration of Khepri mostly
takes place: we use a function called `rabbit_khepri:handle_fallback/1`
which selects the database and perform the query or the transaction.
Here is an example from `rabbit_db_vhost`:
* Up until RabbitMQ 3.12.x:
get(VHostName) when is_binary(VHostName) ->
get_in_mnesia(VHostName).
* Starting with RabbitMQ 3.13.0:
get(VHostName) when is_binary(VHostName) ->
rabbit_khepri:handle_fallback(
#{mnesia => fun() -> get_in_mnesia(VHostName) end,
khepri => fun() -> get_in_khepri(VHostName) end}).
This `rabbit_khepri:handle_fallback/1` function relies on two things:
1. the fact that the `khepri_db` feature flag is enabled, in which case
it always executes the Khepri-based variant.
4. the ability or not to read and write to Mnesia tables otherwise.
Before the feature flag is enabled, or during the migration, the
function will try to execute the Mnesia-based variant. If it succeeds,
then it returns the result. If it fails because one or more Mnesia
tables can't be used, it restarts from scratch: it means the feature
flag is being enabled and depending on the outcome, either the
Mnesia-based variant will succeed (the feature flag couldn't be enabled)
or the feature flag will be marked as enabled and it will call the
Khepri-based variant. The meat of this function really lives in the
`khepri_mnesia_migration` application [3] and
`rabbit_khepri:handle_fallback/1` is a wrapper on top of it that knows
about the feature flag.
However, some calls to the database do not depend on the existence of
Mnesia tables, such as functions where we need to learn about the
members of a cluster. For those, we can't rely on exceptions from
Mnesia. Therefore, we just look at the state of the feature flag to
determine which database to use. There are two situations though:
* Sometimes, we need the feature flag state query to block because the
function interested in it can't return a valid answer during the
migration. Here is an example:
case rabbit_khepri:is_enabled(RemoteNode) of
true -> can_join_using_khepri(RemoteNode);
false -> can_join_using_mnesia(RemoteNode)
end
* Sometimes, we need the feature flag state query to NOT block (for
instance because it would cause a deadlock). Here is an example:
case rabbit_khepri:get_feature_state() of
enabled -> members_using_khepri();
_ -> members_using_mnesia()
end
Direct accesses to Mnesia still exists. They are limited to code that is
specific to Mnesia such as classic queue mirroring or network partitions
handling strategies.
Now, to discover the Mnesia tables to migrate and how to migrate them,
we use an Erlang module attribute called
`rabbit_mnesia_tables_to_khepri_db` which indicates a list of Mnesia
tables and an associated converter module. Here is an example in the
`rabbitmq_recent_history_exchange` plugin:
-rabbit_mnesia_tables_to_khepri_db(
[{?RH_TABLE, rabbit_db_rh_exchange_m2k_converter}]).
The converter module — `rabbit_db_rh_exchange_m2k_converter` in this
example — is is fact a "sub" converter module called but
`rabbit_db_m2k_converter`. See the documentation of a `mnesia_to_khepri`
converter module to learn more about these modules.
[1] https://github.com/rabbitmq/ra
[2] https://github.com/rabbitmq/khepri
[3] https://github.com/rabbitmq/khepri_mnesia_migration
See #7206.
Co-authored-by: Jean-Sébastien Pédron <jean-sebastien@rabbitmq.com>
Co-authored-by: Diana Parra Corbacho <dparracorbac@vmware.com>
Co-authored-by: Michael Davis <mcarsondavis@gmail.com>
2023-01-05 20:57:50 +08:00
|
|
|
[]),
|
|
|
|
|
case rabbit_khepri:transfer_leadership(TransferCandidates) of
|
|
|
|
|
{ok, Node} when Node == node(); Node == undefined ->
|
2025-07-11 18:37:10 +08:00
|
|
|
?LOG_INFO("Skipping leadership transfer of metadata store: current leader is not on this node");
|
Allow to use Khepri database to store metadata instead of Mnesia
[Why]
Mnesia is a very powerful and convenient tool for Erlang applications:
it is a persistent disc-based database, it handles replication accross
multiple Erlang nodes and it is available out-of-the-box from the
Erlang/OTP distribution. RabbitMQ relies on Mnesia to manage all its
metadata:
* virtual hosts' properties
* intenal users
* queue, exchange and binding declarations (not queues data)
* runtime parameters and policies
* ...
Unfortunately Mnesia makes it difficult to handle network partition and,
as a consequence, the merge conflicts between Erlang nodes once the
network partition is resolved. RabbitMQ provides several partition
handling strategies but they are not bullet-proof. Users still hit
situations where it is a pain to repair a cluster following a network
partition.
[How]
@kjnilsson created Ra [1], a Raft consensus library that RabbitMQ
already uses successfully to implement quorum queues and streams for
instance. Those queues do not suffer from network partitions.
We created Khepri [2], a new persistent and replicated database engine
based on Ra and we want to use it in place of Mnesia in RabbitMQ to
solve the problems with network partitions.
This patch integrates Khepri as an experimental feature. When enabled,
RabbitMQ will store all its metadata in Khepri instead of Mnesia.
This change comes with behavior changes. While Khepri remains disabled,
you should see no changes to the behavior of RabbitMQ. If there are
changes, it is a bug. After Khepri is enabled, there are significant
changes of behavior that you should be aware of.
Because it is based on the Raft consensus algorithm, when there is a
network partition, only the cluster members that are in the partition
with at least `(Number of nodes in the cluster ÷ 2) + 1` number of nodes
can "make progress". In other words, only those nodes may write to the
Khepri database and read from the database and expect a consistent
result.
For instance in a cluster of 5 RabbitMQ nodes:
* If there are two partitions, one with 3 nodes, one with 2 nodes, only
the group of 3 nodes will be able to write to the database.
* If there are three partitions, two with 2 nodes, one with 1 node, none
of the group can write to the database.
Because the Khepri database will be used for all kind of metadata, it
means that RabbitMQ nodes that can't write to the database will be
unable to perform some operations. A list of operations and what to
expect is documented in the associated pull request and the RabbitMQ
website.
This requirement from Raft also affects the startup of RabbitMQ nodes in
a cluster. Indeed, at least a quorum number of nodes must be started at
once to allow nodes to become ready.
To enable Khepri, you need to enable the `khepri_db` feature flag:
rabbitmqctl enable_feature_flag khepri_db
When the `khepri_db` feature flag is enabled, the migration code
performs the following two tasks:
1. It synchronizes the Khepri cluster membership from the Mnesia
cluster. It uses `mnesia_to_khepri:sync_cluster_membership/1` from
the `khepri_mnesia_migration` application [3].
2. It copies data from relevant Mnesia tables to Khepri, doing some
conversion if necessary on the way. Again, it uses
`mnesia_to_khepri:copy_tables/4` from `khepri_mnesia_migration` to do
it.
This can be performed on a running standalone RabbitMQ node or cluster.
Data will be migrated from Mnesia to Khepri without any service
interruption. Note that during the migration, the performance may
decrease and the memory footprint may go up.
Because this feature flag is considered experimental, it is not enabled
by default even on a brand new RabbitMQ deployment.
More about the implementation details below:
In the past months, all accesses to Mnesia were isolated in a collection
of `rabbit_db*` modules. This is where the integration of Khepri mostly
takes place: we use a function called `rabbit_khepri:handle_fallback/1`
which selects the database and perform the query or the transaction.
Here is an example from `rabbit_db_vhost`:
* Up until RabbitMQ 3.12.x:
get(VHostName) when is_binary(VHostName) ->
get_in_mnesia(VHostName).
* Starting with RabbitMQ 3.13.0:
get(VHostName) when is_binary(VHostName) ->
rabbit_khepri:handle_fallback(
#{mnesia => fun() -> get_in_mnesia(VHostName) end,
khepri => fun() -> get_in_khepri(VHostName) end}).
This `rabbit_khepri:handle_fallback/1` function relies on two things:
1. the fact that the `khepri_db` feature flag is enabled, in which case
it always executes the Khepri-based variant.
4. the ability or not to read and write to Mnesia tables otherwise.
Before the feature flag is enabled, or during the migration, the
function will try to execute the Mnesia-based variant. If it succeeds,
then it returns the result. If it fails because one or more Mnesia
tables can't be used, it restarts from scratch: it means the feature
flag is being enabled and depending on the outcome, either the
Mnesia-based variant will succeed (the feature flag couldn't be enabled)
or the feature flag will be marked as enabled and it will call the
Khepri-based variant. The meat of this function really lives in the
`khepri_mnesia_migration` application [3] and
`rabbit_khepri:handle_fallback/1` is a wrapper on top of it that knows
about the feature flag.
However, some calls to the database do not depend on the existence of
Mnesia tables, such as functions where we need to learn about the
members of a cluster. For those, we can't rely on exceptions from
Mnesia. Therefore, we just look at the state of the feature flag to
determine which database to use. There are two situations though:
* Sometimes, we need the feature flag state query to block because the
function interested in it can't return a valid answer during the
migration. Here is an example:
case rabbit_khepri:is_enabled(RemoteNode) of
true -> can_join_using_khepri(RemoteNode);
false -> can_join_using_mnesia(RemoteNode)
end
* Sometimes, we need the feature flag state query to NOT block (for
instance because it would cause a deadlock). Here is an example:
case rabbit_khepri:get_feature_state() of
enabled -> members_using_khepri();
_ -> members_using_mnesia()
end
Direct accesses to Mnesia still exists. They are limited to code that is
specific to Mnesia such as classic queue mirroring or network partitions
handling strategies.
Now, to discover the Mnesia tables to migrate and how to migrate them,
we use an Erlang module attribute called
`rabbit_mnesia_tables_to_khepri_db` which indicates a list of Mnesia
tables and an associated converter module. Here is an example in the
`rabbitmq_recent_history_exchange` plugin:
-rabbit_mnesia_tables_to_khepri_db(
[{?RH_TABLE, rabbit_db_rh_exchange_m2k_converter}]).
The converter module — `rabbit_db_rh_exchange_m2k_converter` in this
example — is is fact a "sub" converter module called but
`rabbit_db_m2k_converter`. See the documentation of a `mnesia_to_khepri`
converter module to learn more about these modules.
[1] https://github.com/rabbitmq/ra
[2] https://github.com/rabbitmq/khepri
[3] https://github.com/rabbitmq/khepri_mnesia_migration
See #7206.
Co-authored-by: Jean-Sébastien Pédron <jean-sebastien@rabbitmq.com>
Co-authored-by: Diana Parra Corbacho <dparracorbac@vmware.com>
Co-authored-by: Michael Davis <mcarsondavis@gmail.com>
2023-01-05 20:57:50 +08:00
|
|
|
{ok, Node} ->
|
2025-07-11 18:37:10 +08:00
|
|
|
?LOG_INFO("Leadership transfer for metadata store on this node has been done. The new leader is ~p", [Node]);
|
Allow to use Khepri database to store metadata instead of Mnesia
[Why]
Mnesia is a very powerful and convenient tool for Erlang applications:
it is a persistent disc-based database, it handles replication accross
multiple Erlang nodes and it is available out-of-the-box from the
Erlang/OTP distribution. RabbitMQ relies on Mnesia to manage all its
metadata:
* virtual hosts' properties
* intenal users
* queue, exchange and binding declarations (not queues data)
* runtime parameters and policies
* ...
Unfortunately Mnesia makes it difficult to handle network partition and,
as a consequence, the merge conflicts between Erlang nodes once the
network partition is resolved. RabbitMQ provides several partition
handling strategies but they are not bullet-proof. Users still hit
situations where it is a pain to repair a cluster following a network
partition.
[How]
@kjnilsson created Ra [1], a Raft consensus library that RabbitMQ
already uses successfully to implement quorum queues and streams for
instance. Those queues do not suffer from network partitions.
We created Khepri [2], a new persistent and replicated database engine
based on Ra and we want to use it in place of Mnesia in RabbitMQ to
solve the problems with network partitions.
This patch integrates Khepri as an experimental feature. When enabled,
RabbitMQ will store all its metadata in Khepri instead of Mnesia.
This change comes with behavior changes. While Khepri remains disabled,
you should see no changes to the behavior of RabbitMQ. If there are
changes, it is a bug. After Khepri is enabled, there are significant
changes of behavior that you should be aware of.
Because it is based on the Raft consensus algorithm, when there is a
network partition, only the cluster members that are in the partition
with at least `(Number of nodes in the cluster ÷ 2) + 1` number of nodes
can "make progress". In other words, only those nodes may write to the
Khepri database and read from the database and expect a consistent
result.
For instance in a cluster of 5 RabbitMQ nodes:
* If there are two partitions, one with 3 nodes, one with 2 nodes, only
the group of 3 nodes will be able to write to the database.
* If there are three partitions, two with 2 nodes, one with 1 node, none
of the group can write to the database.
Because the Khepri database will be used for all kind of metadata, it
means that RabbitMQ nodes that can't write to the database will be
unable to perform some operations. A list of operations and what to
expect is documented in the associated pull request and the RabbitMQ
website.
This requirement from Raft also affects the startup of RabbitMQ nodes in
a cluster. Indeed, at least a quorum number of nodes must be started at
once to allow nodes to become ready.
To enable Khepri, you need to enable the `khepri_db` feature flag:
rabbitmqctl enable_feature_flag khepri_db
When the `khepri_db` feature flag is enabled, the migration code
performs the following two tasks:
1. It synchronizes the Khepri cluster membership from the Mnesia
cluster. It uses `mnesia_to_khepri:sync_cluster_membership/1` from
the `khepri_mnesia_migration` application [3].
2. It copies data from relevant Mnesia tables to Khepri, doing some
conversion if necessary on the way. Again, it uses
`mnesia_to_khepri:copy_tables/4` from `khepri_mnesia_migration` to do
it.
This can be performed on a running standalone RabbitMQ node or cluster.
Data will be migrated from Mnesia to Khepri without any service
interruption. Note that during the migration, the performance may
decrease and the memory footprint may go up.
Because this feature flag is considered experimental, it is not enabled
by default even on a brand new RabbitMQ deployment.
More about the implementation details below:
In the past months, all accesses to Mnesia were isolated in a collection
of `rabbit_db*` modules. This is where the integration of Khepri mostly
takes place: we use a function called `rabbit_khepri:handle_fallback/1`
which selects the database and perform the query or the transaction.
Here is an example from `rabbit_db_vhost`:
* Up until RabbitMQ 3.12.x:
get(VHostName) when is_binary(VHostName) ->
get_in_mnesia(VHostName).
* Starting with RabbitMQ 3.13.0:
get(VHostName) when is_binary(VHostName) ->
rabbit_khepri:handle_fallback(
#{mnesia => fun() -> get_in_mnesia(VHostName) end,
khepri => fun() -> get_in_khepri(VHostName) end}).
This `rabbit_khepri:handle_fallback/1` function relies on two things:
1. the fact that the `khepri_db` feature flag is enabled, in which case
it always executes the Khepri-based variant.
4. the ability or not to read and write to Mnesia tables otherwise.
Before the feature flag is enabled, or during the migration, the
function will try to execute the Mnesia-based variant. If it succeeds,
then it returns the result. If it fails because one or more Mnesia
tables can't be used, it restarts from scratch: it means the feature
flag is being enabled and depending on the outcome, either the
Mnesia-based variant will succeed (the feature flag couldn't be enabled)
or the feature flag will be marked as enabled and it will call the
Khepri-based variant. The meat of this function really lives in the
`khepri_mnesia_migration` application [3] and
`rabbit_khepri:handle_fallback/1` is a wrapper on top of it that knows
about the feature flag.
However, some calls to the database do not depend on the existence of
Mnesia tables, such as functions where we need to learn about the
members of a cluster. For those, we can't rely on exceptions from
Mnesia. Therefore, we just look at the state of the feature flag to
determine which database to use. There are two situations though:
* Sometimes, we need the feature flag state query to block because the
function interested in it can't return a valid answer during the
migration. Here is an example:
case rabbit_khepri:is_enabled(RemoteNode) of
true -> can_join_using_khepri(RemoteNode);
false -> can_join_using_mnesia(RemoteNode)
end
* Sometimes, we need the feature flag state query to NOT block (for
instance because it would cause a deadlock). Here is an example:
case rabbit_khepri:get_feature_state() of
enabled -> members_using_khepri();
_ -> members_using_mnesia()
end
Direct accesses to Mnesia still exists. They are limited to code that is
specific to Mnesia such as classic queue mirroring or network partitions
handling strategies.
Now, to discover the Mnesia tables to migrate and how to migrate them,
we use an Erlang module attribute called
`rabbit_mnesia_tables_to_khepri_db` which indicates a list of Mnesia
tables and an associated converter module. Here is an example in the
`rabbitmq_recent_history_exchange` plugin:
-rabbit_mnesia_tables_to_khepri_db(
[{?RH_TABLE, rabbit_db_rh_exchange_m2k_converter}]).
The converter module — `rabbit_db_rh_exchange_m2k_converter` in this
example — is is fact a "sub" converter module called but
`rabbit_db_m2k_converter`. See the documentation of a `mnesia_to_khepri`
converter module to learn more about these modules.
[1] https://github.com/rabbitmq/ra
[2] https://github.com/rabbitmq/khepri
[3] https://github.com/rabbitmq/khepri_mnesia_migration
See #7206.
Co-authored-by: Jean-Sébastien Pédron <jean-sebastien@rabbitmq.com>
Co-authored-by: Diana Parra Corbacho <dparracorbac@vmware.com>
Co-authored-by: Michael Davis <mcarsondavis@gmail.com>
2023-01-05 20:57:50 +08:00
|
|
|
Error ->
|
2025-07-11 18:37:10 +08:00
|
|
|
?LOG_WARNING("Skipping leadership transfer of metadata store: ~p", [Error])
|
Allow to use Khepri database to store metadata instead of Mnesia
[Why]
Mnesia is a very powerful and convenient tool for Erlang applications:
it is a persistent disc-based database, it handles replication accross
multiple Erlang nodes and it is available out-of-the-box from the
Erlang/OTP distribution. RabbitMQ relies on Mnesia to manage all its
metadata:
* virtual hosts' properties
* intenal users
* queue, exchange and binding declarations (not queues data)
* runtime parameters and policies
* ...
Unfortunately Mnesia makes it difficult to handle network partition and,
as a consequence, the merge conflicts between Erlang nodes once the
network partition is resolved. RabbitMQ provides several partition
handling strategies but they are not bullet-proof. Users still hit
situations where it is a pain to repair a cluster following a network
partition.
[How]
@kjnilsson created Ra [1], a Raft consensus library that RabbitMQ
already uses successfully to implement quorum queues and streams for
instance. Those queues do not suffer from network partitions.
We created Khepri [2], a new persistent and replicated database engine
based on Ra and we want to use it in place of Mnesia in RabbitMQ to
solve the problems with network partitions.
This patch integrates Khepri as an experimental feature. When enabled,
RabbitMQ will store all its metadata in Khepri instead of Mnesia.
This change comes with behavior changes. While Khepri remains disabled,
you should see no changes to the behavior of RabbitMQ. If there are
changes, it is a bug. After Khepri is enabled, there are significant
changes of behavior that you should be aware of.
Because it is based on the Raft consensus algorithm, when there is a
network partition, only the cluster members that are in the partition
with at least `(Number of nodes in the cluster ÷ 2) + 1` number of nodes
can "make progress". In other words, only those nodes may write to the
Khepri database and read from the database and expect a consistent
result.
For instance in a cluster of 5 RabbitMQ nodes:
* If there are two partitions, one with 3 nodes, one with 2 nodes, only
the group of 3 nodes will be able to write to the database.
* If there are three partitions, two with 2 nodes, one with 1 node, none
of the group can write to the database.
Because the Khepri database will be used for all kind of metadata, it
means that RabbitMQ nodes that can't write to the database will be
unable to perform some operations. A list of operations and what to
expect is documented in the associated pull request and the RabbitMQ
website.
This requirement from Raft also affects the startup of RabbitMQ nodes in
a cluster. Indeed, at least a quorum number of nodes must be started at
once to allow nodes to become ready.
To enable Khepri, you need to enable the `khepri_db` feature flag:
rabbitmqctl enable_feature_flag khepri_db
When the `khepri_db` feature flag is enabled, the migration code
performs the following two tasks:
1. It synchronizes the Khepri cluster membership from the Mnesia
cluster. It uses `mnesia_to_khepri:sync_cluster_membership/1` from
the `khepri_mnesia_migration` application [3].
2. It copies data from relevant Mnesia tables to Khepri, doing some
conversion if necessary on the way. Again, it uses
`mnesia_to_khepri:copy_tables/4` from `khepri_mnesia_migration` to do
it.
This can be performed on a running standalone RabbitMQ node or cluster.
Data will be migrated from Mnesia to Khepri without any service
interruption. Note that during the migration, the performance may
decrease and the memory footprint may go up.
Because this feature flag is considered experimental, it is not enabled
by default even on a brand new RabbitMQ deployment.
More about the implementation details below:
In the past months, all accesses to Mnesia were isolated in a collection
of `rabbit_db*` modules. This is where the integration of Khepri mostly
takes place: we use a function called `rabbit_khepri:handle_fallback/1`
which selects the database and perform the query or the transaction.
Here is an example from `rabbit_db_vhost`:
* Up until RabbitMQ 3.12.x:
get(VHostName) when is_binary(VHostName) ->
get_in_mnesia(VHostName).
* Starting with RabbitMQ 3.13.0:
get(VHostName) when is_binary(VHostName) ->
rabbit_khepri:handle_fallback(
#{mnesia => fun() -> get_in_mnesia(VHostName) end,
khepri => fun() -> get_in_khepri(VHostName) end}).
This `rabbit_khepri:handle_fallback/1` function relies on two things:
1. the fact that the `khepri_db` feature flag is enabled, in which case
it always executes the Khepri-based variant.
4. the ability or not to read and write to Mnesia tables otherwise.
Before the feature flag is enabled, or during the migration, the
function will try to execute the Mnesia-based variant. If it succeeds,
then it returns the result. If it fails because one or more Mnesia
tables can't be used, it restarts from scratch: it means the feature
flag is being enabled and depending on the outcome, either the
Mnesia-based variant will succeed (the feature flag couldn't be enabled)
or the feature flag will be marked as enabled and it will call the
Khepri-based variant. The meat of this function really lives in the
`khepri_mnesia_migration` application [3] and
`rabbit_khepri:handle_fallback/1` is a wrapper on top of it that knows
about the feature flag.
However, some calls to the database do not depend on the existence of
Mnesia tables, such as functions where we need to learn about the
members of a cluster. For those, we can't rely on exceptions from
Mnesia. Therefore, we just look at the state of the feature flag to
determine which database to use. There are two situations though:
* Sometimes, we need the feature flag state query to block because the
function interested in it can't return a valid answer during the
migration. Here is an example:
case rabbit_khepri:is_enabled(RemoteNode) of
true -> can_join_using_khepri(RemoteNode);
false -> can_join_using_mnesia(RemoteNode)
end
* Sometimes, we need the feature flag state query to NOT block (for
instance because it would cause a deadlock). Here is an example:
case rabbit_khepri:get_feature_state() of
enabled -> members_using_khepri();
_ -> members_using_mnesia()
end
Direct accesses to Mnesia still exists. They are limited to code that is
specific to Mnesia such as classic queue mirroring or network partitions
handling strategies.
Now, to discover the Mnesia tables to migrate and how to migrate them,
we use an Erlang module attribute called
`rabbit_mnesia_tables_to_khepri_db` which indicates a list of Mnesia
tables and an associated converter module. Here is an example in the
`rabbitmq_recent_history_exchange` plugin:
-rabbit_mnesia_tables_to_khepri_db(
[{?RH_TABLE, rabbit_db_rh_exchange_m2k_converter}]).
The converter module — `rabbit_db_rh_exchange_m2k_converter` in this
example — is is fact a "sub" converter module called but
`rabbit_db_m2k_converter`. See the documentation of a `mnesia_to_khepri`
converter module to learn more about these modules.
[1] https://github.com/rabbitmq/ra
[2] https://github.com/rabbitmq/khepri
[3] https://github.com/rabbitmq/khepri_mnesia_migration
See #7206.
Co-authored-by: Jean-Sébastien Pédron <jean-sebastien@rabbitmq.com>
Co-authored-by: Diana Parra Corbacho <dparracorbac@vmware.com>
Co-authored-by: Michael Davis <mcarsondavis@gmail.com>
2023-01-05 20:57:50 +08:00
|
|
|
end.
|
|
|
|
|
|
2021-01-26 19:47:15 +08:00
|
|
|
-spec primary_replica_transfer_candidate_nodes() -> [node()].
|
2020-06-17 23:12:15 +08:00
|
|
|
primary_replica_transfer_candidate_nodes() ->
|
2023-01-18 01:39:20 +08:00
|
|
|
filter_out_drained_nodes_consistent_read(rabbit_nodes:list_running() -- [node()]).
|
2020-06-17 23:12:15 +08:00
|
|
|
|
2021-01-26 19:47:15 +08:00
|
|
|
-spec random_primary_replica_transfer_candidate_node([node()], [node()]) -> {ok, node()} | undefined.
|
|
|
|
|
random_primary_replica_transfer_candidate_node([], _Preferred) ->
|
2020-06-17 23:12:15 +08:00
|
|
|
undefined;
|
2021-01-26 19:47:15 +08:00
|
|
|
random_primary_replica_transfer_candidate_node(Candidates, PreferredNodes) ->
|
|
|
|
|
Overlap = sets:to_list(sets:intersection(sets:from_list(Candidates), sets:from_list(PreferredNodes))),
|
|
|
|
|
Candidate = case Overlap of
|
|
|
|
|
[] ->
|
|
|
|
|
%% Since ownership transfer is meant to be run only when we are sure
|
|
|
|
|
%% there are in-sync replicas to transfer to, this is an edge case.
|
|
|
|
|
%% We skip the transfer.
|
|
|
|
|
undefined;
|
|
|
|
|
Nodes ->
|
|
|
|
|
random_nth(Nodes)
|
|
|
|
|
end,
|
2020-06-17 23:12:15 +08:00
|
|
|
{ok, Candidate}.
|
2020-06-25 02:57:04 +08:00
|
|
|
|
2021-01-26 19:47:15 +08:00
|
|
|
random_nth(Nodes) ->
|
|
|
|
|
Nth = erlang:phash2(erlang:monotonic_time(), length(Nodes)),
|
|
|
|
|
lists:nth(Nth + 1, Nodes).
|
|
|
|
|
|
2020-05-19 16:24:13 +08:00
|
|
|
%%
|
|
|
|
|
%% Implementation
|
|
|
|
|
%%
|
|
|
|
|
|
|
|
|
|
local_listener_fold_fun(Fun) ->
|
|
|
|
|
fun(#listener{node = Node, ip_address = Addr, port = Port}, Acc) when Node =:= node() ->
|
|
|
|
|
RanchRef = rabbit_networking:ranch_ref(Addr, Port),
|
|
|
|
|
[Fun(RanchRef) | Acc];
|
|
|
|
|
(_, Acc) ->
|
|
|
|
|
Acc
|
|
|
|
|
end.
|
2023-01-02 12:17:36 +08:00
|
|
|
|
2020-05-19 16:24:13 +08:00
|
|
|
ok_or_first_error(ok, Acc) ->
|
|
|
|
|
Acc;
|
|
|
|
|
ok_or_first_error({error, _} = Err, _Acc) ->
|
|
|
|
|
Err.
|
