Skip to content

KAFKA-19827: Call ack commit callback at end of waiting calls #20758

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Oct 24, 2025
Merged

KAFKA-19827: Call ack commit callback at end of waiting calls #20758

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
a0bc52e
KAFKA-19827: Call ack commit callback at end of waiting calls
AndrewJSchofield Oct 23, 2025
bc42a4f
Review comment
AndrewJSchofield Oct 23, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
39 changes: 35 additions & 4 deletions ...-integration-tests/src/test/java/org/apache/kafka/clients/consumer/ShareConsumerTest.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -357,10 +357,41 @@ public void testAcknowledgementCommitCallbackSuccessfulAcknowledgement() throws
TestUtils.waitForCondition(() -> shareConsumer.poll(Duration.ofMillis(2000)).count() == 1,
DEFAULT_MAX_WAIT_MS, 100L, () -> "Failed to consume records for share consumer");

TestUtils.waitForCondition(() -> {
shareConsumer.poll(Duration.ofMillis(500));
return partitionOffsetsMap.containsKey(tp);
}, DEFAULT_MAX_WAIT_MS, 100L, () -> "Failed to receive call to callback");
// The callback should be called before the return of the poll, even when there are no more records.
ConsumerRecords<byte[], byte[]> records = shareConsumer.poll(Duration.ofMillis(2000));
assertEquals(0, records.count());
assertTrue(partitionOffsetsMap.containsKey(tp));

// We expect no exception as the acknowledgement error code is null.
assertFalse(partitionExceptionMap.containsKey(tp));
verifyShareGroupStateTopicRecordsProduced();
}
}

@ClusterTest
public void testAcknowledgementCommitCallbackSuccessfulAcknowledgementOnCommitSync() throws Exception {
alterShareAutoOffsetReset("group1", "earliest");
try (Producer<byte[], byte[]> producer = createProducer();
ShareConsumer<byte[], byte[]> shareConsumer = createShareConsumer("group1")) {

Map<TopicPartition, Set<Long>> partitionOffsetsMap = new HashMap<>();
Map<TopicPartition, Exception> partitionExceptionMap = new HashMap<>();
ProducerRecord<byte[], byte[]> record = new ProducerRecord<>(tp.topic(), tp.partition(), null, "key".getBytes(), "value".getBytes());

producer.send(record);
producer.flush();

shareConsumer.setAcknowledgementCommitCallback(new TestableAcknowledgementCommitCallback(partitionOffsetsMap, partitionExceptionMap));
shareConsumer.subscribe(Set.of(tp.topic()));

TestUtils.waitForCondition(() -> shareConsumer.poll(Duration.ofMillis(2000)).count() == 1,
DEFAULT_MAX_WAIT_MS, 100L, () -> "Failed to consume records for share consumer");

// The acknowledgement commit callback should be called before the commitSync returns
// once the records have been confirmed to have been acknowledged.
Map<TopicIdPartition, Optional<KafkaException>> result = shareConsumer.commitSync();
assertEquals(1, result.size());
assertTrue(partitionOffsetsMap.containsKey(tp));

// We expect no exception as the acknowledgement error code is null.
assertFalse(partitionExceptionMap.containsKey(tp));
Expand Down
14 changes: 13 additions & 1 deletion clients/src/main/java/org/apache/kafka/clients/consumer/internals/ShareConsumerImpl.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -609,6 +609,15 @@ public synchronized ConsumerRecords<K, V> poll(final Duration timeout) {
throw e.cause();
} finally {
kafkaShareConsumerMetrics.recordPollEnd(timer.currentTimeMs());

// Handle any acknowledgements which completed while we were waiting, but do not throw
// the exception because the fetched records would then not be returned to the caller
try {
handleCompletedAcknowledgements(false);
} catch (Throwable t) {
log.warn("Exception thrown in acknowledgement commit callback", t);
}
Comment on lines +615 to +619
Copy link
Collaborator

@TaiJuWu TaiJuWu Oct 28, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I find this change make testShareGroupMaxSizeConfigExceeded become flaky.
The following code will fix this issue but seem to violate the comments.

// Handle any acknowledgements which completed while we were waiting, but do not throw
// the exception because the fetched records would then not be returned to the caller
try {
    handleCompletedAcknowledgements(false);
} catch (GroupMaxSizeReachedException e) {
    log.warn("Exception thrown in acknowledgement commit callback", e);
    throw e;
} catch (Throwable t) {
    log.warn("Exception thrown in acknowledgement commit callback", t);
}

Copy link
Member

@chia7712 chia7712 Nov 1, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@TaiJuWu thanks for this find. The flaky behaviour and its root cause have already been logged in https://issues.apache.org/jira/browse/KAFKA-19840


release();
}
}
Expand Down Expand Up @@ -752,8 +761,11 @@ public Map<TopicIdPartition, Optional<KafkaException>> commitSync(final Duration
result.put(tip, Optional.of(exception));
}
});
return result;

// Handle any acknowledgements which completed while we were waiting
handleCompletedAcknowledgements(false);

return result;
} finally {
wakeupTrigger.clearTask();
}
Expand Down