Repository navigation
test(notebooks): raise per-line install timeout to 1200s (PRI-392) - #1335
klemens-floege wants to merge 2 commits into
Conversation
The catboost/xgboost line downloads ~560 MiB; 600s timed out on a slow nightly runner (PRI-392). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit d5fe32a. Configure here.
| @@ -0,0 +1 @@ | |||
| The nightly notebook install check now allows 1200 s per install line instead of 600 s, so a slow runner downloading the `catboost`/`xgboost` wheels no longer fails it. | |||
There was a problem hiding this comment.
Changelog covers CI-only timeout
Medium Severity
The changelog/1335.fixed.md fragment describes a nightly CI timeout, not a user-facing change. Towncrier concatenates these fragments into the published changelog, so this would surface install-check internals, 1200 s/600 s timings, and the catboost/xgboost incident diagnosis to users.
Triggered by learned rule: Towncrier fragments: user-facing change only; keep concise
Reviewed by Cursor Bugbot for commit d5fe32a. Configure here.
There was a problem hiding this comment.
@klemens-floege I think if we change the CI we can use the no-changelog-needed tag.
adrian-prior
left a comment
adrian-prior
left a comment
There was a problem hiding this comment.
Are you sure that this is the reason? Why would a worker need more than 5 min to download 500MB? I would assume that this would take less than 30 seconds.
|
I have little intution how long the download times can take on these CI workers, btw it is often much less than 5min, but we did get a timeout here |
|
If you think the error was transient I would be happy to close the PR + ticket |
|
I would move to closing for now and if this repeats itself we can raise the timeout. I am unsure if this would have run through even in 1200s. Feels like maybe a bad worker to me. |
2 participants
Generated by Claude Code
The nightly notebook install check timed out on Python 3.11. The
catboost xgboostinstall line inTabPFN_Demo_Local.ipynbdownloads about 560 MiB, mostlynvidia-nccl-cu12throughxgboost. A slow runner did not finish in 600 s, so this PR raises the timeout for each install line to 1200 s.1200 s stays below the workflow's 30-minute
timeout-minutes, so the test fails before the job is killed. The ticket investigation ran the same install on a cold cache in 19 s, and both packages resolved to the latest PyPI versions.Linear: https://linear.app/priorlabs/issue/PRI-392/nightly-notebook-install-check-failed-python-311
Fixes #1334
🤖 Generated with Claude Code