Skip to content

test(serve): de-flake test_sample_deployment CPU timeout (OPS-7728) #11727

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
dmitry-tokarev-nv merged 6 commits into from
Jul 16, 2026
+7 −1
Merged

test(serve): de-flake test_sample_deployment CPU timeout (OPS-7728) #11727

Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
9345134
test(serve): de-flake test_sample_deployment CPU timeout (OPS-7728)
dmitry-tokarev-nv Jul 15, 2026
4203c19
test(serve): drop internal Linear ref from timeout comment
dmitry-tokarev-nv Jul 15, 2026
6d434a4
test(serve): lower de-flake timeout 600->450s to bound PR-gate worst …
dmitry-tokarev-nv Jul 15, 2026
202f8bf
ci: [DO NOT MERGE] in-PR vllm-efa CPU-test repro to validate de-flake…
dmitry-tokarev-nv Jul 15, 2026
be2411e
revert: remove in-PR vllm-efa CPU-test diagnostic (OPS-7728)
dmitry-tokarev-nv Jul 16, 2026
88263dd
Merge branch 'main' into dtokarev/ops-7728-deflake-sample-deployment
dmitry-tokarev-nv Jul 16, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 7 additions & 1 deletion tests/serve/test_sample.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,7 +32,13 @@
script_args=["--model-name", "Qwen/Qwen3-0.6B"],
marks=[
pytest.mark.gpu_0,
pytest.mark.timeout(300),
# CPU-mode vLLM startup for this deployment runs ~237s; the old
# 300s cap left only ~20% headroom and flaked ~40% of recent
# post-merge runs on vllm-runtime-efa / CPU Test (the pod was
# killed mid-startup under runner contention). 450s is ~2x the
# measured baseline -- enough headroom to de-flake while keeping
# the pre-merge-gate worst-case bounded.
pytest.mark.timeout(450),
pytest.mark.pre_merge,

@devin-ai-integration devin-ai-integration Bot Jul 15, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🔍 Pre-merge test with 600s timeout is very slow for the PR gate

The pytest guidelines (.ai/pytest-guidelines.md:283-284) state that tests averaging over 60 seconds should default to post_merge unless they guard a critical path. This test averages ~237s and is marked pre_merge — that was already the case before this PR (the scheduling marker is unchanged), but doubling the timeout ceiling to 600s makes the worst-case CI impact more significant. The pre-merge placement is pre-existing and outside the scope of this PR's intent, but the reviewer may want to consider whether this test should move to post_merge.

Open in Devin Review

Was this helpful? React with 👍 or 👎 to provide feedback.

pytest.mark.unified,
pytest.mark.vllm,
Expand Down
Loading