Repository navigation

Resume benchmark fleet jobs after Spot interruptions - #1145

Merged

laughingman7743 merged 17 commits into

masterfrom

feat/1136-spot-resilience

Oct 12, 2026

Merged

Resume benchmark fleet jobs after Spot interruptions#1145
laughingman7743 merged 17 commits into
masterfrom
feat/1136-spot-resilience

Commits on Oct 12, 2026
- Resume benchmark fleet jobs after Spot interruptions
  
  laughingman7743
  and
  claude
  committed
  1c75926
- Skip the run after an early notice and keep done markers on report errors
  
  laughingman7743
  and
  claude
  committed
  1dfcab1
- Keep a lease whose conditional write succeeded before a retry
  
  laughingman7743
  and
  claude
  committed
  af75752
- Stop stale lease owners and wait for dead workers' claims
  
  laughingman7743
  and
  claude
  committed
  4b06a6c
- Bound the watch shutdown after a notice or a lost lease
  
  laughingman7743
  and
  claude
  committed
  e033f45
- Confirm live claims before exiting and keep the first done marker
  
  laughingman7743
  and
  claude
  committed
  6531dbf
- Treat a claim that expired after the run as lost before recording the job
  
  laughingman7743
  and
  claude
  committed
  9e349de
- Confirm claim ownership synchronously before recording a job
  
  laughingman7743
  and
  claude
  committed
  0f85514
- Recheck the claim before recording a job and before exiting
  
  laughingman7743
  and
  claude
  committed
  f0dc49f
- Keep each job summary below the attempt that the done marker names
  
  laughingman7743
  and
  claude
  committed
  26748f7
- Address code review findings on resume, leases, and the results grant
  
  laughingman7743
  and
  claude
  committed
  3eefc13
- Keep the host after an unconfirmed stopped attempt and retry uploads
  
  laughingman7743
  and
  claude
  committed
  ae77dab
- Keep cleanup errors when the claim is lost after the run
  
  laughingman7743
  and
  claude
  committed
  b5795b5
- Merge remote-tracking branch 'origin/master' into feat/1136-spot-resilience
  laughingman7743
  committed
  493b8e9
- Simplify the fleet's resume, leases, and worker start
  
  laughingman7743
  and
  claude
  committed
  7bbf1e1
- Release locally lost claims, contain downloads, and keep publication retryable
  
  laughingman7743
  and
  claude
  committed
  cd932b2
- Test a claim that stays live in S3 after it expires locally
  
  laughingman7743
  and
  claude
  committed
  0e5bbb7

Commits

Commits on Oct 12, 2026