GitHub user ahmedahamid opened a pull request:
https://github.com/apache/samza/pull/498
SAMZA-1476: Fix TestStatefulTask flaky test
Unable to reproduce the issue even after constraining CPU/memory resources
available to the JVM, this fix addresses a potential root cause â a
`CountDownLatch` shared between 2 threads, main test thread and Kafka producer
thread, but not marked volatile even though it is reinitialized by the main
test thread. This could cause the reported issue since each of the 2 threads
could end up invoking `countDown`/`await` on a different `CountDownLatch`
object.
It is worthwhile to mention that without this fix, a different test,
`TestShutdownStatefulTask`, should have also exhibited some flakiness since it
shares the same `TestTask` base with `TestStatefulTask` and exercises exactly
the same portion of code that includes the failing assertion. Since no such
flakiness was reported for `TestShutdownStatefulTask` however, the assumption
made by this fix is that it might have not been encountered or reported.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/ahmedahamid/samza dev/fix-1476
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/samza/pull/498.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #498
----
commit 973d224e439f736da27a495f8470f4086b08b5cb
Author: Ahmed Abdul Hamid <ahabdulh@...>
Date: 2018-05-01T18:40:22Z
SAMZA-1476: Fix TestStatefulTask flaky test
Unable to reproduce the issue even after constraining CPU/memory resources
available to the JVM, this fix addresses a potential root cause â a
CountDownLatch shared between 2 threads, main test thread and Kafka producer
thread, but not marked volatile even though it is reinitialized by the main
test thread. This could cause the reported issue since each of the 2 threads
could end up invoking countDown/await on a different reference.
----
---
