PR #60583: fix(cron): resume interrupted recurring jobs on first restart

• Repository: openclaw/openclaw
• Author: joelnishanth
• State: closed | merged: True
• Link: https://github.com/openclaw/openclaw/pull/60583

Description (problem / solution / changelog)

Closes #60495

Summary

• When a gateway restart interrupts a running cron/every job, the job is now re-executed as part of startup catch-up on the first restart instead of requiring a second restart to recover.
• One-shot (schedule kind "at") jobs preserve the existing behavior and are not retried after interruption.

Root Cause

The start function in src/cron/service/ops.ts collected all jobs with a stale runningAtMs marker into startupInterruptedJobIds and passed them to runMissedJobs as skipJobIds. This explicitly excluded interrupted jobs from startup catch-up. After runMissedJobs skipped them, recomputeNextRuns advanced their nextRunAtMs into the future, so the normal timer tick also would not fire them. Only a second restart (where runningAtMs was already cleared) allowed runMissedJobs to pick them up via the missed-slot detection path.

Fix

Changed the skip list to only include interrupted one-shot jobs. Interrupted recurring jobs (cron/every) now have their runningAtMs cleared and are immediately eligible for startup catch-up, since their nextRunAtMs is still past-due (the incomplete run never reached applyJobResult).

Changes

src/cron/service/ops.ts — replaced blanket startupInterruptedJobIds skip with interruptedOneShotIds that only collects one-shot jobs.

src/cron/service/ops.test.ts — updated the stale-running-marker test to expect the interrupted cron job to be replayed on first restart.

src/cron/service.restart-catchup.test.ts — updated the stale-marker test to expect the interrupted cron job to run and verify its lastStatus and lastRunAtMs are set.

Test Plan

• npx vitest run src/cron/service.restart-catchup.test.ts — 8/8 pass
• npx vitest run src/cron/service/ops.test.ts — 5/5 pass
• npx vitest run src/cron/service.issue-regressions.test.ts — 43/43 pass
• npx vitest run src/cron/service.issue-13992-regression.test.ts — 6/6 pass
• npx vitest run src/cron/service.every-jobs-fire.test.ts — 3/3 pass
• No lint errors on touched files

Risks and Mitigations

• Interrupted recurring jobs now re-execute on first restart. This is the intended fix. Previously they were silently skipped, causing data gaps for users relying on scheduled tasks. The interrupted run never completed, so retrying is the correct behavior.
• One-shot behavior preserved. Interrupted one-shots are still skipped, matching the existing design choice that one-shots are fire-and-forget.

— Joel Nishanth · offlyn.AI

Made with Cursor

Changed files

• CHANGELOG.md (modified, +1/-0)
• src/cron/service.restart-catchup.test.ts (modified, +10/-8)
• src/cron/service/ops.test.ts (modified, +5/-4)
• src/cron/service/ops.ts (modified, +13/-4)

PR #60615: fix(cron): replay interrupted recurring jobs on first restart

• Repository: openclaw/openclaw
• Author: oliviareid-svg
• State: closed | merged: False
• Link: https://github.com/openclaw/openclaw/pull/60615

Description (problem / solution / changelog)

Summary

When OpenClaw restarts while a recurring cron job is running, the job is not retried on startup. It is only caught on a second restart when runMissedJobs() finds the next scheduled time past due again.

Root cause: In start() (src/cron/service/ops.ts), all interrupted jobs — regardless of schedule kind — are added to startupInterruptedJobIds and skipped by runMissedJobs(). After the skip, recomputeNextRuns() advances nextRunAtMs to the next future slot without executing the missed run.

Fix: Only add one-shot (kind === "at") jobs to the skip set. Recurring jobs (cron/every) are now eligible for startup catchup via runMissedJobs(), which already handles them correctly when their nextRunAtMs is past due.

• Updated restart-catchup test to verify interrupted recurring jobs ARE replayed
• One-shot (at) job test remains unchanged (still skipped on restart)

Fixes #60495

Test plan

• Existing test "replays interrupted recurring cron jobs on startup" passes (updated from previous "clears stale running markers" test)
• Existing test "does not replay interrupted one-shot jobs on startup" still passes
• Manual: create a cron job, restart while running, verify it retries on first restart

🤖 Generated with Claude Code

Changed files

• src/cron/service.restart-catchup.test.ts (modified, +10/-5)
• src/cron/service/ops.ts (modified, +5/-1)