I have 2 PR blocked by what it appears as flaky Pi...
# pinot-dev
j
I have 2 PR blocked by what it appears as flaky Pinot-controller unit test under
Pinot Unit Test Set 2 (temurin-21) with skip bytebuffers:true
. Inspecting other PRs, it seems like this unit test suite also fails for the same reason. • https://github.com/apache/pinot/actions/runs/11378305857/job/31692173933?pr=14105 • https://github.com/apache/pinot/actions/runs/11378343759/job/31692198040?pr=14241 • https://github.com/apache/pinot/pull/14244 When I tried to run the same unit test on my laptop, they worked fine. Anyone have any recommendation on how to proceed?
a
Running them again.
j
ty
e
+1, I’ve been experiencing a lot of flaky tests (different failures each run) and it’s a source of frustration. Is Pinot using Apache-run test infrastructure? If so, do we need to do any maintenance on the servers? Do we need more resources? I’m getting failures like this within different suites on each run, which seem clearly unrelated to the outcome of tests but instead point to the test environment:
Copy code
Runtime java.net.BindException: Address already in use
j
I think there's some effort in mitigating this already
🙌 1
image.png
I saw one PR from @Xiang Fu which reduces test workload of a unit test
x
Which test fails for the Address already in use
e
@Xiang Fu at the moment it’s
QueryRoutingTest
maybe I should rebase / merge master to bring that IdealStateGroupCommitTest update into my branch, thanks @Jack Luo for pointing that out 🙏
👍 1
x
yeah, I looked at the current master branch CI, it’s getting better
🌟 1
e
awesome, thanks for your efforts to improve that @Xiang Fu! Pulled those changes into my branch as well 🤞
x
🤞
j
Copy code
[INFO] org.apache.pinot.integration.tests.BasicAuthBatchIntegrationTest.testIngestionBatch -- Time elapsed: 8.014 s
[INFO] 
[INFO] Results:
[INFO] 
Error:  Failures: 
Error:    HybridClusterIntegrationTest.testReload:248->BaseClusterIntegrationTestSet.testReload:769->BaseClusterIntegrationTestSet.lambda$testReload$5:773 expected [0] but found [1]
[INFO] 
Error:  Tests run: 425, Failures: 1, Errors: 0, Skipped: 12
[INFO] 
[INFO] ------------------------------------------------------------------------
[INFO] BUILD FAILURE
[INFO] ------------------------------------------------------------------------
[INFO] Total time:  35:18 min
[INFO] Finished at: 2024-10-21T20:40:45Z
Any chance this is also an environment issue? My change has nothing to do with the failed integration test.
x
I don’t know what’s going on, let’s add more logs to check: https://github.com/apache/pinot/pull/14269
e
Is there a way to re-run just failed tasks? Via Github UI or similar?
x
commiter can run failed test
image.png,image.png
j
Can I apply to become a "commiter" so I can run-run tests?
x
Definitely, just apache commitership requires a process and may take certain amount of time
🚀 1
e
I’ve seen other OSS projects with bots that PR authors can message to help work around this issue. It can be helpful for allowing momentum for authors that are not yet committers. I’ve seen them operate by taking action on the author sending PR comments like “Run Pinot Integration Test Set 2” and then re-executing a test suite by the matching name. It could be something to explore as a middle ground
but of course, takes time and effort haha