Now that Kafka supports Java 17, this PR updates the shell module to get
rid of older code. The changes mostly include:
- Collections.emptyList(), Collections.singletonList() and
- Arrays.asList() are replaced with List.of()
- Replace switch statement with switch expression
Reviewers: PoAn Yang <payang@apache.org>, Yung
<yungyung7654321@gmail.com>, Ken Huang <s7133700@gmail.com>, TengYao Chi
<frankvicky@apache.org>
The test incorrectly used force-terminate instead of
forceTerminateTransaction
Reviewers: Justine Olshan <jolshan@confuent.io>, Kuan-Po Tseng
<brandboat@gmail.com>, Ken Huang <s7133700@gmail.com>
### Issue:
API Responses missing latest version in [Kafka protocol
guide](https://kafka.apache.org/protocol.html)
#### For example:
These are missing:
- ApiVersions Response (Version: 4) — Only versions 0–3 are documented,
though version 4 of the request is included.
- DescribeTopicPartitions Response — Not listed at all.
- Fetch Response (Version: 17) — Only versions 4–16 are documented,
though version 17 of the request is included.
#### After the fix:
docs/generated/protocol_messages.html
<img width="1045" alt="image"
src="https://github.com/user-attachments/assets/5ea79ced-aab5-4c47-8e09-9956047c9bf1"
/>
Reviewers: dengziming <dengziming1993@gmail.com>, Ken Huang
<s7133700@gmail.com>, Chia-Ping Tsai <chia7712@gmail.com>
Now that Kafka Brokers support Java 17, this PR updates the share
coordinator module to get rid of older code. The changes mostly include:
- Collections.emptyList(), Collections.singletonList() and
- Arrays.asList() are replaced with List.of()
- Collections.emptyMap() and Collections.singletonMap() are replaced
with Map.of()
- Collections.singleton() is replaced with Set.of()
Reviewers: Andrew Schofield <aschofield@confluent.io>
Use Java to rewrite `PlaintextConsumerCommitTest` by new test infra and
move it to client-integration-tests module.
Reviewers: PoAn Yang <payang@apache.org>, Chia-Ping Tsai
<chia7712@gmail.com>
Replace `TopicMetadata` with `MetadataImage` in
`SubscribedTopicDescriberImpl` and `TargetAssignmentBuilder`.
Reviewers: Chia-Ping Tsai <chia7712@gmail.com>, David Jacot
<djacot@confluent.io>
This PR is based on the discussion here:
https://github.com/apache/kafka/pull/18929#discussion_r2083238645
Currently, the handle methods related to `shared‐group` requests are
inconsistent: some return `Unit` while others return
`CompletableFuture[Unit]` without a clear rationale. This PR
standardizes all of them to return `CompletableFuture[Unit]` and ensures
consistent error handling by chaining `.exceptionally(handleError)` to
each call site.
Reviewers: PoAn Yang <payang@apache.org>, Ken Huang
<s7133700@gmail.com>, Chia-Ping Tsai <chia7712@gmail.com>
Enable next system test with KIP-1071.
Some of the validation inside the test did not make sense for KIP-1071.
This is because in KIP-1071, if a member leaves or joins the group, not
all members may enter a REBALANCING state. We use the wrapper introduced
in [KAFKA-19271](https://issues.apache.org/jira/browse/KAFKA-19271)
to print a log line whenever the member epoch is bumped, which is the
only way a member can "indirectly" observe that other members are
rebalancing.
Reviewers: Bill Bejeck <bill@confluent.io>
In FETCH requests and TXN_OFFSET_COMMIT requests, on current trunk we
run into a race condition inside UnifiedLog, causing a
`NoSuchElementException` in
`UnifiedLog.fetchLastStableOffsetMetadata(UnifiedLog.java:651)`.
The cause is that the line a performing an `isPresent` check on a
volatile Optional before accessing it in `get`, leaving the door open to
a race condition when the optional changes between `isPresent` and
`get`. This change takes a copy of the volatile variable first.
This PR adds changes, so the IQ endpoint information is only sent to
streams group members when there has been a change in the assignments
requiring an update in the streams client host-partition ownership.
The existing IQ integration test passes with no modifications and
updated the `GroupMetadataManagerTest` to cover the new process.
Reviewers: Matthias Sax <mjsax@apache.org>, Lucas Brutschy
<lucasbru@apache.org>
Follow up https://github.com/apache/kafka/pull/19460/files#r2062664349
Reviewers: Ismael Juma <ismael@juma.me.uk>, PoAn Yang
<payang@apache.org>, TaiJuWu <tjwu1217@gmail.com>, Ken Huang
<s7133700@gmail.com>, Chia-Ping Tsai <chia7712@gmail.com>
This PR includes the system test file test_console_share_consumer.py
which tests the functioning of ConsoleShareConsumer
Reviewers: Andrew Schofield <aschofield@confluent.io>
This patch adds the `TopicId` field to the `OffsetCommitValue` record as
a tagged field. It will be later used on the offset fetch path to ensure
that the persisted offset matches the requested one.
Reviewers: Dongnuo Lyu <dlyu@confluent.io>, Sean Quah
<squah@confluent.io>, Lianet Magrans <lmagrans@confluent.io>
According to KIP-830, in Kafka 4 users currently relying on JmxReporter
and that are using additional reporters via metric.reporters will have
to include org.apache.kafka.common.metrics.JmxReporter in
metric.reporters.
Reviewers: Mickael Maison <mickael.maison@gmail.com>
With KIP-1071 enabled, the main consumer is created differently, side
stepping `KafkaClientSupplier`.
To allow injection test wrappers, we add an internal ConsumerWrapper,
until we define a new public interface.
Reviewers: Lucas Brutschy <lbrutschy@confluent.io>
Migrate MinInSyncReplicasConfigTest to the server module
Reviewers: PoAn Yang <payang@apache.org>, Yung
<yungyung7654321@gmail.com>, TengYao Chi <frankvicky@apache.org>, Ken
Huang <s7133700@gmail.com>
This PR fixes a regression bug introduced with KAFKA-17203. We need to
pass in mutable collections into `closeTaskClean(...)`.
Reviewers: Chia-Ping Tsai <chia7712@gmail.com>, Bruno Cadonna <bruno@confluent.io>, Lucas Brutschy <lbrutschy@confluent.io>
Updates the min version used by `ListOffsetsRequest` to
`ApiKeys.LIST_OFFSETS.oldestVersion()` rather than hardcoding `1`.
Reviewers: Ken Huang <s7133700@gmail.com>, PoAn Yang
<payang@apache.org>, TengYao Chi <frankvicky@apache.org>, Chia-Ping
Tsai <chia7712@gmail.com>
* Change `ListClientMetricsResourcesRequest.json` to
`ListConfigResourcesRequest.json`.
* Change `ListClientMetricsResourcesResponse.json` to
`ListConfigResourcesResponse.json`.
* Change `ListClientMetricsResourcesRequest.java` to
`ListConfigResourcesRequest.java`.
* Change `ListClientMetricsResourcesResponse.java` to
`ListConfigResourcesResponsejava`.
* Change `KafkaApis` to handle both `ListClientMetricsResourcesRequest`
v0 and v1 requests.
Reviewers: Andrew Schofield <aschofield@confluent.io>
---------
Signed-off-by: PoAn Yang <payang@apache.org>
Recent changes to BrokerMetadataPublisher seem to have caused a mismatch
between the BrokerMetadataPublisher and its test.
Reviewers: Apoorv Mittal <apoorvmittal10@gmail.com>, Sushant Mahajan
<smahajan@confluent.io>
### About
Minor improvements to Persister APIs in KafkaApis with respect to share
group protocol enablement
### Testing
The new code added has been tested with the help of unit tests
Reviewers: Sushant Mahajan <sushant.mahajan88@gmail.com>, Apoorv Mittal <apoorvmittal10@gmail.com>
Group Coordinator Shards are not unloaded when `__consumer_offsets`
topic is deleted. The unloading is scheduled but it is ignored because
the epoch is equal to the current epoch:
```
[2025-05-13 08:46:00,883] INFO [GroupCoordinator id=1]
Scheduling unloading of metadata for __consumer_offsets-0 with epoch
OptionalInt[0]
(org.apache.kafka.coordinator.common.runtime.CoordinatorRuntime)
[2025-05-13 08:46:00,883] INFO [GroupCoordinator id=1] Scheduling
unloading of metadata for __consumer_offsets-1 with epoch OptionalInt[0]
(org.apache.kafka.coordinator.common.runtime.CoordinatorRuntime)
[2025-05-13 08:46:00,883] INFO [GroupCoordinator id=1] Ignored unloading
metadata for __consumer_offsets-0 in epoch OptionalInt[0] since current
epoch is 0.
(org.apache.kafka.coordinator.common.runtime.CoordinatorRuntime)
[2025-05-13 08:46:00,883] INFO [GroupCoordinator id=1] Ignored unloading
metadata for __consumer_offsets-1 in epoch OptionalInt[0] since current
epoch is 0.
(org.apache.kafka.coordinator.common.runtime.CoordinatorRuntime)
```
This patch fixes the issue by not setting the leader epoch in this case.
The coordinator expects the leader epoch to be incremented when the
resignation code is called. When the topic is deleted, the epoch is not
incremented. Therefore, we must not use it. Note that this is aligned
with deleted partitions are handled too.
Reviewers: Dongnuo Lyu <dlyu@confluent.io>, José Armando García Sancio <jsancio@apache.org>
Adding test to specifically force the fencing path due to delayed
rebalance, and validate how the consumer recovers automatically.
Running this test and DEBUG log enabled, allows to see the details of
the fencing flow: consumer getting fenced due to rebalance exceeded,
resetting to epoch 0, rejoining on the next poll with the existing
subscription, and being accepted back in the group (so consumption
resumes)
This is aimed to help understand
[KAFKA-19233](https://issues.apache.org/jira/browse/KAFKA-19233)
Will add another one in separate PR to also involve commits in similar
fencing scenarios.
Reviewers: TengYao Chi <frankvicky@apache.org>
This fixes the paragraph splitting done in the pr-format.py script. If a
line contains a triple backtick, the script will consider it markdown
and adjust the formatting accordingly.
Reviewers: TengYao Chi <frankvicky@apache.org>, Ken Huang <s7133700@gmail.com>, Jhen-Yung Hsu <jhenyunghsu@gmail.com>
* Previously we had added code to `GroupCoordinatorService.onElection`
to reconcile pending share group initializing topics. This was done to
manage state in case of failovers and broker failures.
* However, we later modified share group heartbeat code to do the retry
to clean up the state and the `onElection` code is now redundant.
* In this PR we are cleaning up this code.
Reviewers: David Jacot <djacot@confluent.io>, Andrew Schofield <aschofield@confluent.io>
According to the recent changes in KIP-932, when the share session cache
is full and a broker receives a Share Fetch request with Initial Share
Session Epoch (0), then the error code `SHARE_SESSION_LIMIT_REACHED` is
returned after a delay of maxWaitMs. This PR implements this logic. In
order to add a delay between subsequent share fetch requests, the timer
is delayed operation purgatory is used. A new `IdleShareFetchTimeTask`
has been added which takes in a CompletableFuture<Void>. Upon the
expiration, this future is completed with null. When the future is
completes, a response is sent back to the client with the error code
`SHARE_SESSION_LIMIT_REACHED`
Reviewers: Andrew Schofield <aschofield@confluent.io>, Apoorv Mittal <apoorvmittal10@gmail.com>
* We have a few periodic timer tasks in `ShareCoordinatorService` which
run continuously.
* With the recent introduction of share group enabled config at feature
level, we would like these jobs to stop when the aforementioned feature
is disabled.
* In this PR, we have added the functionality to make that possible.
* Additionally the service has been supplemented with addition of a
static share group config supplier.
* New test has been added.
Reviewers: Andrew Schofield <aschofield@confluent.io>, Apoorv Mittal
<apoorvmittal10@gmail.com>
If delete.topic.enable is false on the brokers, deleteTopics will mark
the topics for deletion, but not actually delete them. The futures will
return successfully in this case.
It is not true as the server return exception now.
```java if (!config.deleteTopicEnable) { if (apiVersion < 3) { throw
new InvalidRequestException("Topic deletion is disabled.") } else {
throw new TopicDeletionDisabledException() } } ```
Reviewers: Nick Guo <lansg0504@gmail.com>, PoAn Yang
<payang@apache.org>, Ken Huang <s7133700@gmail.com>, Jhen-Yung Hsu
<jhenyunghsu@gmail.com>, Chia-Ping Tsai <chia7712@gmail.com>
The PR is a minor follow up on
https://github.com/apache/kafka/pull/19659.
KafkaApis.scala already have a check which denies new share fetch
related calls if the share group is not supported. Hence no new sessions
shall be created so the requirement to have share group enabled flag in
ShareSessionCache is not needed, unless I am missing something.
Reviewers: Abhinav Dixit <adixit@confluent.io>, Andrew Schofield <aschofield@confluent.io>
* Add `com.dynatrace.hash4j:hash4j:0.22.0` to dependencies.
* Add `computeTopicHash` to `org.apache.kafka.coordinator.group.Utils`.
* If topic name is non-existent, return 0.
* If topic name is existent, use streaming XXH3 to compute topic hash
with magic byte, topic id, topic name, number of partitions, partition
id and sorted racks.
* Add `computeGroupHash` to `org.apache.kafka.coordinator.group.Utils`.
* If topic map is empty, return 0.
* If topic map is not empty, use streaming XXH3 to compute group
metadata hash with sorted topic hashes by topic names.
* Add related unit test.
Reviewers: Ismael Juma <ismael@juma.me.uk>, Chia-Ping Tsai <chia7712@gmail.com>, Sean Quah <squah@confluent.io>, David Jacot <djacot@confluent.io>
---------
Signed-off-by: PoAn Yang <payang@apache.org>
Do not print the `"Unexpected error .."` log when the config
`delete.topic.enable` is false
Reviewers: PoAn Yang <payang@apache.org>, Ken Huang
<s7133700@gmail.com>, TengYao Chi <frankvicky@apache.org>, Chia-Ping
Tsai <chia7712@gmail.com>
In KRaft mode, controllerListenerName must always be specified, so we don't need an `Optional` to wrap it.
Reviewers: Chia-Ping Tsai <chia7712@gmail.com>, Ken Huang
<s7133700@gmail.com>, TengYao Chi <frankvicky@apache.org>
This patch belongs to the client-side changes required to enable 2PC as
a part of KIP-939.
New method is added to KafkaProducer: public PreparedTxnState
prepareTransaction()
This would flush all the pending messages and transition the producer
into a mode where only .commitTransaction, .abortTransaction, or
.completeTransaction could be called (calling other methods, e.g. .send
, in that mode would result in IllegalStateException being thrown). If
the call is successful (all messages successfully got flushed to all
partitions) the transaction is prepared. If the 2PC is not enabled, we
return the INVALID_TXN_STATE error.
A new state is added to the TransactionManager called
PREPARING_TRANSACTION. There are two situations where we would move into
this state:
1) When prepareTransaction() is called during an ongoing transaction
with 2PC enabled
2) When initTransaction(true) is called after a client failure
(keepPrepared = true)
Reviewers: Artem Livshits <alivshits@confluent.io>, Justine Olshan
<jolshan@confluent.io>
This PR introduces the following per-broker metrics:
-`kafka.controller:type=KafkaController,name=BrokerRegistrationState,broker=X`
-`kafka.controller:type=KafkaController,name=TimeSinceLastHeartbeatReceivedMs,broker=X`
and this metric:
`kafka.controller:type=KafkaController,name=ControlledShutdownBrokerCount`
Reviewers: Colin P. McCabe <cmccabe@apache.org>
Since topic name is sensitive information, it should return a
TOPIC_AUTHORIZATION_FAILED error for non-existing topic. The Fetch
request also follows this pattern.
Co-authored-by: John Doe <zh2725284321@gmail.com>, Ken Huang <s7133700@gmail.com>, Jhen-Yung Hsu <yungyung7654321@gmail.com>
Reviewers: Jun Rao <junrao@gmail.com>, Chia-Ping Tsai
<chia7712@gmail.com>, TaiJuWu <tjwu1217@gmail.com>, TengYao Chi
<frankvicky@apache.org>, Ken Huang <s7133700@gmail.com>, Jhen-Yung Hsu
<jhenyunghsu@gmail.com>
Add a warning message when using log.cleaner.enable to remind users that
this configuration is deprecated. Also, add a warning message for
log.cleaner.threads=0 because in version 5.0, the value must be greater
than zero.
Reviewers: Ken Huang <s7133700@gmail.com>, PoAn Yang
<payang@apache.org>, TengYao Chi <frankvicky@apache.org>, TaiJuWu
<tjwu1217@gmail.com>, Chia-Ping Tsai <chia7712@gmail.com>
We updated the validation rule for cleanup.policy in remote storage
mode.
If remote log storage is enabled, only cleanup.policy=delete is allowed.
Any other value (e.g. compact, compact,delete) will now result in a
config validation error.
Reviewers: Luke Chen <showuon@gmail.com>, Ken Huang
<s7133700@gmail.com>, PoAn Yang <payang@apache.org>, Chia-Ping Tsai
<chia7712@gmail.com>
The root case of flakiness is race condition between worker thread
(thread which is executing the test) and executor-testDelayedFuture
(thread which should execute callback).
It was fixed with TestUtils#waitForCondition to wait until callback will
be done
Test evidence:
Test was running 1000 times with repeated test.
Results: `~/p/kafka (bloku/kafka-19091) [1]> ./gradlew server:test
--tests DelayedFutureTest --fail-fast > res.txt` `~/p/kafka
(bloku/kafka-19091)> grep FAILED res.txt ` `~/p/kafka
(bloku/kafka-19091) [1]>`
res.txt: `> Task :server:test`
`Gradle Test Run :server:test > Gradle Test Executor 14 >
DelayedFutureTest > testDelayedFuture() > repetition 1 of 1000 PASSED`
...
`BUILD SUCCESSFUL in 37m`
Reviewers: Ken Huang <s7133700@gmail.com>, TaiJuWu <tjwu1217@gmail.com>,
PoAn Yang <payang@apache.org>, Chia-Ping Tsai <chia7712@gmail.com>
The PR adds ShareGroupListener which triggers on group changes, such as
when member leaves or group is empty. Such events are specific to
connection on respective broker. There events help to clean specific
states managed for respective member or group in various caches.
Reviewers: Andrew Schofield <aschofield@confluent.io>
There were some tests in share_consumer_test, where the default value of
metadata_quorum was quorum.zk. That is change to quorum.isolated_kraft in
this PR
Reviewers: Andrew Schofield <aschofield@confluent.io>
Guard against possible `NullPointerExceptions` in
`ConsumerNetworkThread.cleanup()` if
`ConsumerNetworkThread.initializeResources()` previously failed.
Reviewers: Chia-Ping Tsai <chia7712@gmail.com>
Sanskar Jhajharia
Kaushik Raina
Ritika Reddy
Andy Li
TengYao Chi
PoAn Yang
Ken Huang
Hong-Yi Chen
Yunchi Pang
Lucas Brutschy
Bill Bejeck
Jhen-Yung Hsu
Chirag Wadhwa
David Jacot
Mickael Maison
Federico Valeri
Matthias J. Sax
Bolin Lin
Andrew Schofield
Sushant Mahajan
Abhinav Dixit
Lianet Magrans
David Arthur
YuChia Ma
Apoorv Mittal
xijiu
Ming-Yen Chung
Kevin Wu
Xuan-Zhang Gong
Uladzislau Blok
Dmitry Werner
Kirk True