aIbrahiim commented on PR #38237: URL: https://github.com/apache/beam/pull/38237#issuecomment-4270860505
> Would you mind sharing some context? Looks like the test has been been flaky though not permared on HEAD. How this is related to the flaky test? Yeah sure happy to add more context and sorry forgot to add that on the PR description as well, so the stack trace looked like the ApproximateUnique snippet test was just timing out / flaking, but I dont think the transform itself was the real problem so what was happening is those snippet tests use TestPipeline under the hood, and TestPipeline was still inheriting the global --test-pipeline-options that pytest sets up for the cloud job. so sometimes you would get this tiny unit snippet test accidentally running with runner settings meant for heavier integration-style runs, and then you hit weird gRPC / deadline stuff under load which matches what was found so the fix is basically is only use those global pytest pipeline options for tests that are actually integration tests (is_integration_test=True) and by that way snippet tests stay on the normal local path, and we dont get this accidental coupling to remote runner config @Abacn -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
