Repository navigation
Fix CI floor OOM: update per_shard_peak to 3GiB + fallback to 2 - #5843
Closed
gunbai-bot[bot] wants to merge 1 commit into
Closed
gunbai-bot[bot] wants to merge 1 commit into
gunbai-bot[bot] wants to merge 1 commit into
Conversation
…llback Width=3 has been OOM-killing the 8GiB cgroup on every CI run since #5836 enforced the real cap (runs 28253594624, 28256000608). The prior measurement (2.1GiB from run 27893441091) was taken on a larger-budget runner against an older corpus. The OOM observation proves per_shard_actual > 2.43GiB. Updated to 3GiB (within observed [2.43, 3.65]GiB honest range); derives width=2 at 8GiB: floor(8GiB * 4/5 / 3GiB) = 2, fits: 2*3GiB+0.7GiB=6.7GiB. Also drops conservative_fallback_width from 4 to 2 — fallback=4 at 8GiB and 3GiB/shard would also OOM. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Contributor
Author
|
Closing as superseded — the CI floor OOM fix landed via #5836 (real runner cgroup cap, 8 GiB), which supersedes this per_shard_peak approach. — sent from loyal-bee-794 |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
gunbc_ci_floor_per_shard_rss_measurement.per_shard_peak: 2100028416 → 3221225472 (2.0GiB → 3.0GiB)gunbc_ci_floor_conservative_fallback_width: 4 → 2Root cause: Since #5836 enforced the real 8GiB cgroup cap,
memory_aware_spawn_widthderives width=3 (floor(8GiB × 4/5 / 2.0GiB) = 3). But 3 workers OOM-kill the 8GiB cgroup every run — this has blocked main CI for 5+ consecutive pushes.Evidence: The OOM observation (runs 28253594624, 28256000608; both exit-137) proves per_shard_actual > (8GiB - 0.7GiB executor_base) / 3 = 2.43GiB. The prior 2.0GiB measurement (run 27893441091) was taken on a larger-budget runner against an older, smaller corpus.
Fix: Updated to 3GiB — within the observed honest range [2.43, 3.65]GiB. Derives width=2: floor(8GiB × 4/5 / 3GiB) = 2. Fits: 2×3GiB + 0.7GiB = 6.7GiB < 8GiB. Also drops conservative_fallback_width from 4 to 2 — the old fallback of 4 would also OOM at 3GiB/shard on an 8GiB cap.
All existing
ci_floor_planwitnesses remain valid (fallback witness checks== gunbc_ci_floor_conservative_fallback_width, width witness checks< breadth— both still hold).🤖 Generated with Claude Code