KAFKA-19452: Fix flaky test LogRecoveryTest.testHWCheckpointWithFailuresMultipleLogSegments #20121
+1
−1
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
github-actions
bot
added
core
junrao
reviewed
Jul 7, 2025
| @@ -224,6 +224,9 @@ class LogRecoveryTest extends QuorumTestHarness { | |||
| server1.startup() | |||
| updateProducer() | |||
|
|
|||
| waitUntilTrue(() => server2.replicaManager.onlinePartition(topicPartition).get.inSyncReplicaIds.size == 2, | |||
There was a problem hiding this comment.
Hmm, why is this needed? Later on, there is another waitUntilTrue(). That should be enough, right?
TestUtils.waitUntilTrue(() =>
server1.replicaManager.localLogOrException(topicPartition).highWatermark == hw,
"Failed to update high watermark for follower after timeout")
Also, from the jira, the failure happened in the following line.
leader = awaitLeaderChange(servers, topicPartition, oldLeaderOpt = Some(leader), timeout = 30000L)
There was a problem hiding this comment.
@junrao My previous fix was not addressing the specific failure point mentioned in the JIRA issue, and the waitUntilTrue I added was likely redundant for the reasons you mentioned. My apologies.
I updated the commit with a fix for the actual issue.
There was a problem hiding this comment.
@junrao Can you please take a look again?
karuturi
force-pushed
the
fix/KAFKA-19452-flaky-log-recovery-test
branch
from
6c911ef to
755494a
Compare
…resMultipleLogSegments The test was flaky because awaitLeaderChange was timing out. The original code called awaitLeaderChange(servers, ...) where servers is a list containing both server1 and server2. This was problematic because server1 was still offline. The function was trying to poll a dead broker, which led to unpredictable timeouts and caused the test to fail intermittently. The fix is to make the awaitLeaderChange call more specific, so it only polls the broker that is actually running(server2).
karuturi
force-pushed
the
fix/KAFKA-19452-flaky-log-recovery-test
branch
from
755494a to
8c32f2b
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The test was flaky because awaitLeaderChange was timing out.
The original code called awaitLeaderChange(servers, ...) where servers
is a list containing both server1 and server2. This was problematic
because server1 was still offline. The function was trying to poll a
dead broker, which led to unpredictable timeouts and caused the test to
fail intermittently.
The fix is to make the awaitLeaderChange call more specific, so it only
polls the broker that is actually running(server2).
ran the test in a loop to ensure its working fine
for i in {1..10}; do ./gradlew core:test --tests kafka.server.LogRecoveryTest && echo "Run $i passed" || exit 1; done