vvcephei commented on a change in pull request #9414:
URL: https://github.com/apache/kafka/pull/9414#discussion_r526377375
##########
File path:
streams/src/test/java/org/apache/kafka/streams/integration/EOSUncleanShutdownIntegrationTest.java
##########
@@ -140,9 +140,6 @@ public void
shouldWorkWithUncleanShutdownWipeOutStateStore() throws InterruptedE
IntegrationTestUtils.produceSynchronously(producerConfig, false,
input, Optional.empty(),
singletonList(new KeyValueTimestamp<>("k1", "v1", 0L)));
- TestUtils.waitForCondition(stateDir::exists,
- "Failed awaiting CreateTopics first request failure");
Review comment:
Hmm, I see. Maybe I misread it. I thought we were asserting that Streams
actually creates the directory when it starts processing, so that the later
assertion (L159) that it cleans up the directory is actually valid.
Then again, L159 seems to assert the opposite thing that its own comment
states, so maybe this whole test is just wacky.
The only clue about what we're trying to do here is the name, which almost
doesn't make sense at all: `shouldWorkWithUncleanShutdownWipeOutStateStore`. I
guess it's saying that we should delete the state store on unclean shutdown? I
think that's not what we do anyway. I think we just guarantee that we write an
empty checkpoint file. So maybe we should instead change the assertion to:
* the directory is absent
* OR the directory is present, but the checkpoint file is missing
* OR the directory and checkpoint file are present, but the checkpoint file
is empty
Any of those should indicate a successful unclean shutdown.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
