Skip to content
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

KAFKA-17506 KRaftMigrationDriver initialization race #17147

Merged
merged 10 commits into from
Sep 11, 2024
Merged

KAFKA-17506 KRaftMigrationDriver initialization race #17147

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
a4802b0
fixup deflake workflow
mumrah Sep 10, 2024
01399c1
add repeated test
mumrah Sep 10, 2024
2f87c13
ensure we initialize
mumrah Sep 10, 2024
4181969
increase repeat
mumrah Sep 10, 2024
8790626
Handle another init race
mumrah Sep 10, 2024
8512458
revert temp code
mumrah Sep 10, 2024
91be734
add a test
mumrah Sep 10, 2024
abfa140
Merge remote-tracking branch 'origin/trunk' into KAFKA-17506-KRaftMig…
mumrah Sep 10, 2024
e711122
Revert deflake changes
mumrah Sep 10, 2024
87c4234
improve test case
mumrah Sep 11, 2024
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
7 changes: 5 additions & 2 deletions metadata/src/main/java/org/apache/kafka/metadata/migration/KRaftMigrationDriver.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -359,6 +359,9 @@ public String name() {
@Override
public void onControllerChange(LeaderAndEpoch newLeaderAndEpoch) {
curLeaderAndEpoch = newLeaderAndEpoch;
if (migrationState.equals(MigrationDriverState.UNINITIALIZED)) {
eventQueue.append(new RecoverMigrationStateFromZKEvent());
}
eventQueue.append(new KRaftLeaderEvent(newLeaderAndEpoch));
}

Expand Down Expand Up @@ -519,8 +522,8 @@ public void run() throws Exception {
KRaftMigrationDriver.this.image = image;
String metadataType = isSnapshot ? "snapshot" : "delta";

if (migrationState.equals(MigrationDriverState.INACTIVE)) {
// No need to log anything if this node is not the active controller
if (EnumSet.of(MigrationDriverState.UNINITIALIZED, MigrationDriverState.INACTIVE).contains(migrationState)) {
Copy link
Contributor

Choose a reason for hiding this comment

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

Ah, nice! Not only onControllerChange, but all OnMetadataChange will cause race condition.

Copy link
Member Author

Choose a reason for hiding this comment

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

Yea, this race is harmless since onMetadataChange does not alter the state machine state and most of its logic is guarded behind if (!migrationState.allowDualWrite()) {. But it's good to have this for completeness.

// No need to log anything if this node is not the active controller or the driver has not initialized
completionHandler.accept(null);
return;
}
Expand Down
26 changes: 26 additions & 0 deletions metadata/src/test/java/org/apache/kafka/metadata/migration/KRaftMigrationDriverTest.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -232,6 +232,32 @@ CompletableFuture<Void> enqueueMetadataChangeEventWithFuture(
return future;
}

@Test
public void testOnControllerChangeWhenUninitialized() throws InterruptedException {
CountingMetadataPropagator metadataPropagator = new CountingMetadataPropagator();
CapturingMigrationClient.newBuilder().build();
CapturingMigrationClient migrationClient = CapturingMigrationClient.newBuilder().build();
MockFaultHandler faultHandler = new MockFaultHandler("testBecomeLeaderUninitialized");
KRaftMigrationDriver.Builder builder = defaultTestBuilder()
.setZkMigrationClient(migrationClient)
.setPropagator(metadataPropagator)
.setFaultHandler(faultHandler);
try (KRaftMigrationDriver driver = builder.build()) {
// Fake a complete migration with ZK client
migrationClient.setMigrationRecoveryState(
ZkMigrationLeadershipState.EMPTY.withKRaftMetadataOffsetAndEpoch(100, 1));

// simulate the Raft layer running before the driver has fully started.
driver.onControllerChange(new LeaderAndEpoch(OptionalInt.of(3000), 1));

// start up the driver. this will enqueue a poll event. once run, this will enqueue a recovery event
driver.start();

// Even though we contrived a race above, the driver still makes it past initialization.
TestUtils.waitForCondition(() -> driver.migrationState().get(30, TimeUnit.SECONDS).equals(MigrationDriverState.WAIT_FOR_CONTROLLER_QUORUM),
"Waiting for KRaftMigrationDriver to enter WAIT_FOR_CONTROLLER_QUORUM state");
}
}
/**
* Don't send RPCs to brokers for every metadata change, only when brokers or topics change.
* This is a regression test for KAFKA-14668
Expand Down