Skip to content

fix(benchmarks): cold-slot worker isolation in sweep_extra.sh (#3023) - #3221

Merged
Xore merged 1 commit into
mainfrom
issue-3023-cold-slot
Sep 19, 2026
Merged

Xore merged 1 commit into
mainfrom
issue-3023-cold-slot

Conversation

@Xore

@Xore Xore commented Sep 19, 2026

Copy link
Copy Markdown
Owner

Closes #3023

This updates the phase-2 cold-slot sweep protocol to:

  • record the exact prior worker state in a state file;
  • stop only workers that are running;
  • restore exactly that prior state via EXIT/INT/TERM traps;
  • fail loudly when worker-state operations fail; and
  • remain idempotent across cleanup paths.

Validation: bash -n analysis/ghidra/benchmarks/corpus/sweep_extra.sh passes. DIFF.md documents the rationale.

This makes phase-2 sweeps comply with the #2641 cold protocol by default, with no retroactive effect on already-published numbers.

@github-actions

Copy link
Copy Markdown

Dependency Review

✅ No vulnerabilities or license issues or OpenSSF Scorecard issues found.

Scanned Files

None

@Xore
Xore merged commit a9b3e7a into main Sep 19, 2026
109 of 110 checks passed
@Xore
Xore deleted the issue-3023-cold-slot branch September 19, 2026 19:20
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.

ops(llm): #2582 has recurred — hp-llm-worker unhealthy for ~11h, and the #1947 phase-2 sweep is running contended, not cold

1 participant