Worker and Queue Troubleshooting
Stalled, failed, duplicated, and repeatedly retried jobs.
This page covers Worker unhealthy states, delayed or repeatedly restarting jobs, and bounded acknowledgement failures.
Inspect job identity, attempts, dependency state, and idempotency before replay.
Source boundary
Source boundary: AutoCore current main 7a504f6e430c16d4fcb03ebdea3cc3fb7816df60, checked against immutable v1.0.0-rc.1 at 9edfb109f44cc80385784c694b96392cfc04e70f.
Classify the job as safe-to-retry, requires operator review, or payment/data-sensitive. Preserve failure evidence and use a bounded retry window.
| Check | Source-backed expectation |
|---|---|
| Scope | Do not bulk replay or purge jobs without an approved recovery plan. |
| Evidence | Use placeholders, timestamps, release identity, route, status, and request ID where relevant. |
| Safe action | pnpm --filter @autocore/worker typecheck |
| Stop condition | Unknown authorization, destructive impact, secret exposure, or conflicting evidence. |
Verification
Run the safe check, record its result, and compare the outcome with the documented contract. If the result depends on external configuration or provider availability, mark it as configuration-dependent rather than claiming a product guarantee.
pnpm --filter @autocore/worker typecheck
Safety boundary
Examples are non-production and use placeholders. Do not deploy, reset, force a migration, create secrets, activate providers, replay sensitive work, or expose private data from this page.