Skip to content

KAFKA-20710: Share coordinator - Fence stale periodic jobs #22603

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

Open
Shekharrajak wants to merge 2 commits into
base: trunk
Choose a base branch
from
+106 −8
Open

KAFKA-20710: Share coordinator - Fence stale periodic jobs #22603

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
c395358
Add periodic job generation test
Shekharrajak Jun 17, 2026
03e4e12
Fence stale periodic jobs
Shekharrajak Jun 17, 2026
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
38 changes: 30 additions & 8 deletions ...coordinator/src/main/java/org/apache/kafka/coordinator/share/ShareCoordinatorService.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -76,6 +76,7 @@
import java.util.concurrent.ConcurrentHashMap;
import java.util.concurrent.Executors;
import java.util.concurrent.atomic.AtomicBoolean;
import java.util.concurrent.atomic.AtomicLong;
import java.util.function.IntSupplier;

import static org.apache.kafka.coordinator.common.runtime.CoordinatorOperationExceptionHelper.handleOperationException;
Expand Down Expand Up @@ -141,6 +142,8 @@ public class ShareCoordinatorService implements ShareCoordinator {
*/
private volatile boolean shouldRunPeriodicJob;

private final AtomicLong periodicJobGeneration = new AtomicLong();

@Shekharrajak Shekharrajak Jun 17, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

protects duplicate timer chains - similar to epoch concept.


public static class Builder {
private final int nodeId;
private final ShareCoordinatorConfig config;
Expand Down Expand Up @@ -319,9 +322,9 @@ public void startup(
log.info("Startup complete.");
}

private void setupPeriodicJobs() {
setupRecordPruning();
setupSnapshotColdPartitions();
private void setupPeriodicJobs(long generation) {
setupRecordPruning(generation);
setupSnapshotColdPartitions(generation);
}

/**
Expand All @@ -331,11 +334,15 @@ private void setupPeriodicJobs() {
*/
// Visibility for tests
void setupRecordPruning() {
setupRecordPruning(periodicJobGeneration.get());
}

private void setupRecordPruning(long generation) {
log.debug("Scheduling share-group state topic prune job.");
timer.add(new TimerTask(config.shareCoordinatorTopicPruneIntervalMs()) {
@Override
public void run() {
if (!shouldRunPeriodicJob) {
if (!shouldRunPeriodicJob(generation)) {
return;
}
List<CompletableFuture<Void>> futures = new ArrayList<>();
Expand All @@ -346,8 +353,11 @@ public void run() {
if (exp != null) {
log.error("Received error in share-group state topic prune.", exp);
}
if (!shouldRunPeriodicJob(generation)) {
return;
}
// Perpetual recursion, failure or not.
setupRecordPruning();
setupRecordPruning(generation);
});
}
});
Expand Down Expand Up @@ -415,11 +425,15 @@ private CompletableFuture<Void> performRecordPruning(TopicPartition tp) {
*/
// Visibility for tests
void setupSnapshotColdPartitions() {
setupSnapshotColdPartitions(periodicJobGeneration.get());
}

private void setupSnapshotColdPartitions(long generation) {
log.debug("Scheduling cold share-partition snapshotting.");
timer.add(new TimerTask(config.shareCoordinatorColdPartitionSnapshotIntervalMs()) {
@Override
public void run() {
if (!shouldRunPeriodicJob) {
if (!shouldRunPeriodicJob(generation)) {
return;
}
List<CompletableFuture<Void>> futures = runtime.scheduleWriteAllOperation(
Expand All @@ -432,7 +446,10 @@ public void run() {
if (exp != null) {
log.error("Received error while snapshotting cold partitions.", exp);
}
setupSnapshotColdPartitions();
if (!shouldRunPeriodicJob(generation)) {
return;
}
setupSnapshotColdPartitions(generation);
});
}
});
Expand Down Expand Up @@ -1137,8 +1154,9 @@ public void onMetadataUpdate(MetadataDelta delta, MetadataImage newImage) {
// 1 1 no op on flag, do not call jobs
if (enabled ^ shouldRunPeriodicJob) {
shouldRunPeriodicJob = enabled;
long generation = periodicJobGeneration.incrementAndGet();
if (enabled) {
setupPeriodicJobs();
setupPeriodicJobs(generation);
}
}
}
Expand Down Expand Up @@ -1182,4 +1200,8 @@ private boolean isShareGroupsEnabled(MetadataImage image) {
boolean shouldRunPeriodicJob() {
return shouldRunPeriodicJob;
}

private boolean shouldRunPeriodicJob(long generation) {
return shouldRunPeriodicJob && periodicJobGeneration.get() == generation;

@Shekharrajak Shekharrajak Jun 17, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

It must not start work if it belongs to old generation.

}
}
76 changes: 76 additions & 0 deletions ...dinator/src/test/java/org/apache/kafka/coordinator/share/ShareCoordinatorServiceTest.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2066,6 +2066,82 @@ public void testPeriodicJobsDoNotRunWhenShareGroupsDisabled() throws Interrupted
service.shutdown();
}

@Test
public void testPeriodicJobsDoNotDuplicateAfterDisableEnableWithInFlightJobs() throws InterruptedException {
CoordinatorRuntime<ShareCoordinatorShard, CoordinatorRecord> runtime = mockRuntime();
PartitionWriter writer = mock(PartitionWriter.class);
MockTime time = new MockTime();
MockTimer timer = spy(new MockTimer(time));

Metrics metrics = new Metrics();

ShareCoordinatorService service = spy(new ShareCoordinatorService(
new LogContext(),
ShareCoordinatorTestConfig.testConfig(),
runtime,
new ShareCoordinatorMetrics(metrics),
time,
timer,
writer
));

CompletableFuture<Optional<Long>> firstPruneFuture = new CompletableFuture<>();
CompletableFuture<Optional<Long>> secondPruneFuture = new CompletableFuture<>();
CompletableFuture<Void> firstSnapshotFuture = new CompletableFuture<>();
CompletableFuture<Void> secondSnapshotFuture = new CompletableFuture<>();

when(runtime.<Optional<Long>>scheduleWriteOperation(
eq("write-state-record-prune"),
any(),
any()
))
.thenReturn(firstPruneFuture)
.thenReturn(secondPruneFuture);

when(runtime.<Void>scheduleWriteAllOperation(
eq("snapshot-cold-partitions"),
any()
))
.thenReturn(List.of(firstSnapshotFuture))
.thenReturn(List.of(secondSnapshotFuture));

service.startup(() -> 1);

@Shekharrajak Shekharrajak Jun 17, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

two periodic jobs:
- record prune job: write-state-record-prune
- cold partition snapshot job: snapshot-cold-partitions


MetadataImage disabledImage = mock(MetadataImage.class, RETURNS_DEEP_STUBS);
when(disabledImage.features().finalizedVersions().getOrDefault(eq(ShareVersion.FEATURE_NAME), anyShort())).thenReturn((short) 0);

MetadataImage enabledImage = mockMetadataImageWithShareGroupsEnabled();
service.onMetadataUpdate(mock(MetadataDelta.class), enabledImage);

verify(timer, times(2)).add(any());
timer.advanceClock(30001L);

@Shekharrajak Shekharrajak Jun 17, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

manually advances mock time so the scheduled timer tasks fire

verify(runtime, times(1)).scheduleWriteOperation(

@Shekharrajak Shekharrajak Jun 17, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

both delayed tasks run once when we advance the timer

eq("write-state-record-prune"),
any(),
any()
);
verify(runtime, times(1)).scheduleWriteAllOperation(
eq("snapshot-cold-partitions"),
any()
);

service.onMetadataUpdate(mock(MetadataDelta.class), disabledImage);
assertFalse(service.shouldRunPeriodicJob());

service.onMetadataUpdate(mock(MetadataDelta.class), enabledImage);
assertTrue(service.shouldRunPeriodicJob());
verify(timer, times(4)).add(any());

@Shekharrajak Shekharrajak Jun 17, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

Without the fix, this became 6 because old prune and old snapshot completions each rescheduled one stale job.

@Shekharrajak Shekharrajak Jun 17, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

2(old) + 4(new)


firstPruneFuture.complete(Optional.empty());
firstSnapshotFuture.complete(null);

verify(timer, times(4)).add(any());

checkMetrics(metrics);

service.shutdown();
}

@Test
public void testShareStateTopicConfigs() {
CoordinatorRuntime<ShareCoordinatorShard, CoordinatorRecord> runtime = mockRuntime();
Expand Down