Conversation
|
Note Gemini is unable to generate a summary for this pull request due to the file types involved not being currently supported. |
|
Assigning reviewers: R: @Abacn for label build. Note: If you would like to opt out of this review, comment Available commands:
The PR bot will only process comments in the main thread (not review comments). |
2bd7b78 to
f611cf8
Compare
f611cf8 to
0f85fec
Compare
| TOX_TESTENV_PASSENV: "DOCKER_*,TESTCONTAINERS_*,TC_*,BEAM_*,GRPC_*,OMP_*,OPENBLAS_*,PYTHONHASHSEED,PYTEST_*" | ||
| # Aggressive retry and timeout settings for flaky CI | ||
| PYTEST_ADDOPTS: "-v --tb=short --maxfail=5 --durations=30 --reruns=5 --reruns-delay=15 --timeout=600 --disable-warnings" | ||
| PYTEST_ADDOPTS: "-v --tb=short --maxfail=5 --durations=30 --reruns=5 --reruns-delay=15 --timeout=900 --disable-warnings" |
There was a problem hiding this comment.
Not sure further increasing timeout and/or rerun could help. We already see the test suite stuck for >5h e.g. https://github.com/apache/beam/actions/runs/23779768061
There was a problem hiding this comment.
Ahh I didn't see this run, then icreasing pytest timeoutcan mask hangs and extend stuck runs so i revert the timeout bump and keep this PR focused on grpc stability env vars only as a short term mitigation @Abacn
| TC_MAX_TRIES: "15" | ||
| TC_SLEEP_TIME: "5" | ||
| # Additional gRPC stability for flaky environment | ||
| GRPC_ARG_KEEPALIVE_TIME_MS: "60000" |
There was a problem hiding this comment.
Wondering if there is some racing condition for tests run in parllell cc: @tvalentyn. Nevertheless we can add these environment variables at the moment.
There was a problem hiding this comment.
if we can repro this stuckness locally, or ssh into GHA worker that is stuck, then we could spy on the python process with tools like pystack, and then examine stracktraces.
There was a problem hiding this comment.
that could give some clues.
There was a problem hiding this comment.
so should I run with SSH access on a stuck self-hosted runner and collect pystack traces there, or add temporary CI instrumentation (faulthandler + periodic pystack dumps as artifacts) so the next stuck run captures stack traces automatically? @tvalentyn
There was a problem hiding this comment.
so should I run with SSH access on a stuck self-hosted runner
is it an option? is it easy?
how often does this issue reproduce?
overall, i'd pick whatever option is more straightforward.
There was a problem hiding this comment.
There was a problem hiding this comment.
Good findings!
re: 1 -- is there something specific to YAML suite? do you have some suggestions (e.g. skip copying tmp?) I
re: 2 -- Could you try setting this environment variable and seeing if it helps:
There was a problem hiding this comment.
re: 1 --The issue ties to YAML external providers: they clone a venv in yaml_provider via clonevirtualenv so source is either a fresh mini venv, or in .dev the whole tox venv (path from base_python) and in CI that tree is live, so paths like tmp/ can disappear mid-copy, fits what we saw so I suggesst ignore tmp/, pycache/, .pytest_cache/, pip/build temps (or replace clonevirtualenv with a copy that supports ignore / only needed dirs) and for .dev/CI, prefer _create_venv_from_scratch (or a template venv) instead of cloning the full tox env, slower, stabler. Optional unit test with a fake venv + volatile tmp/ to guard regressions.
re: 2 -- subprocess_server waits on gRPC channel ready for expansion_service_main until pytest-timeout. faulthandler showed xdist/execnet sometimes, xdist off for one Python version didn’t fix other legs, parallelism isnt the whole story so I suggesst on child exit or long wait, surface exit code + stdout/stderr (and maybe fail faster with a clear error).
For ML YAML, preinstall/pin heavy deps in tox so slow pip/import/model startup doesn’t look like a hang
There was a problem hiding this comment.
I will add the same env var to Python PreCommit and see if it changes the expansion / xdist behavior, will update you back whether it helps
There was a problem hiding this comment.
@tvalentyn I added GRPC_ENABLE_FORK_SUPPORT: "0" in Python PreCommit and kept the YAML side stabilization in yaml_provider then I reran the workflow and it passed https://github.com/apache/beam/actions/runs/23892113364
|
Failed checks are unrelated |
related to #34739
Thank you for your contribution! Follow this checklist to help us incorporate your contribution quickly and easily:
addresses #123), if applicable. This will automatically add a link to the pull request in the issue. If you would like the issue to automatically close on merging the pull request, commentfixes #<ISSUE NUMBER>instead.CHANGES.mdwith noteworthy changes.See the Contributor Guide for more tips on how to make review process smoother.
To check the build health, please visit https://github.com/apache/beam/blob/master/.test-infra/BUILD_STATUS.md
GitHub Actions Tests Status (on master branch)
See CI.md for more information about GitHub Actions CI or the workflows README to see a list of phrases to trigger workflows.