Skip to content

nightly: raise timeouts so the rerun pass can run#411

Merged
TeoSlayer merged 1 commit into
mainfrom
ci/nightly-timeout-headroom
Jul 22, 2026
Merged

nightly: raise timeouts so the rerun pass can run#411
TeoSlayer merged 1 commit into
mainfrom
ci/nightly-timeout-headroom

Conversation

@TeoSlayer

@TeoSlayer TeoSlayer commented Jul 22, 2026

Copy link
Copy Markdown
Collaborator

Two combined fixes for the 3 consecutive red nightlies (Jul 19–21):

  1. main was still on the old serialized config (-parallel 1, no reruns) — the validated gotestsum config from feat/nightly-selfhosted (parallel 6 + --rerun-fails=3 fresh-process isolation + the nightly-beta prerelease job) never landed. This adopts it.
  2. Timeout headroom: go-test -timeout 40/45m → 60/65m, job timeouts → 80/85m, so the first pass is never killed mid-run and the rerun pass gets to isolate the ~45 known load-starvation failures (they pass individually — isolated 2026-07-15; -parallel 3 measured strictly worse: 59 fails/60m vs 45/39m at p6).

Long-term fix remains making the integration tests hermetic; this makes the nightly meaningful + green-capable now, and green nights start cutting v*-nightly.<date> prereleases.

🤖 Generated with Claude Code

…+ timeout headroom

Two combined fixes for the 3 consecutive red nightlies (Jul 19-21):

1. main was still running the OLD serialized config (-parallel 1, no
   reruns) — the validated gotestsum config from feat/nightly-selfhosted
   (parallel 6, --rerun-fails=3 to re-run load-starvation failures in a
   fresh process, nightly-beta release job) never landed. Adopt it.
2. Raise go-test -timeout 40/45m -> 60/65m and job timeout-minutes ->
   80/85 so the first pass is never killed mid-run and the rerun pass
   has room. Evidence: first pass takes ~39m at -parallel 6 with ~45
   load-dependent failures that pass individually (isolated 2026-07-15);
   -parallel 3 measured strictly worse (59 fails/60m), so parallelism
   tuning is a dead end — completion headroom is the fix.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
@TeoSlayer
TeoSlayer force-pushed the ci/nightly-timeout-headroom branch from 8cc4e94 to cad9b25 Compare July 22, 2026 06:23
@TeoSlayer
TeoSlayer merged commit ed31fc7 into main Jul 22, 2026
13 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants