Job worker restarts when SQLite pool acquisition times out on a busy host #951

Open
opened 2026-10-02 21:02:04 +00:00 by kayg · 1 comment
Owner

Observed during the focused #655 browser run on the busy build host. The local debug server logged:

2026-10-02T20:36:14.254664Z ERROR calternal_server::wire: job worker failed; restarting error=SQLite operation failed: pool timed out while waiting for an open connection

The same run logged SQLite pool acquisition waits of 23 seconds. No API 5xx or server crash was observed. The worker restarts, but the current evidence does not establish whether queued background work retains its full retry state. Keep this for the merge round robustness check; no benchmark or broad adversarial matrix was run in this UI job.

Command: cd apps/web && bun run test:e2e:taskday-655-657, against a real local server and the worktree production build. This is separate from the test polling defect fixed in #655.

Observed during the focused #655 browser run on the busy build host. The local debug server logged: ```text 2026-10-02T20:36:14.254664Z ERROR calternal_server::wire: job worker failed; restarting error=SQLite operation failed: pool timed out while waiting for an open connection ``` The same run logged SQLite pool acquisition waits of 23 seconds. No API 5xx or server crash was observed. The worker restarts, but the current evidence does not establish whether queued background work retains its full retry state. Keep this for the merge round robustness check; no benchmark or broad adversarial matrix was run in this UI job. Command: `cd apps/web && bun run test:e2e:taskday-655-657`, against a real local server and the worktree production build. This is separate from the test polling defect fixed in #655.
Author
Owner

The latest real-server #655 browser run also logged pending user Home deletion list failed error=authentication database is busy at 2026-10-02T21:19:02Z, after pool acquisition waits of 25–29 seconds. The process stayed alive. The run then stopped at the all-day completion Undo assertion (Task status remained done; expected todo); I cannot establish whether the shortcut dispatched its writer or whether resource contention caused that failure. These are recorded separately from SLOW-only query logs. No existing expectations were changed.

The latest real-server #655 browser run also logged `pending user Home deletion list failed error=authentication database is busy` at 2026-10-02T21:19:02Z, after pool acquisition waits of 25–29 seconds. The process stayed alive. The run then stopped at the all-day completion Undo assertion (`Task status remained done; expected todo`); I cannot establish whether the shortcut dispatched its writer or whether resource contention caused that failure. These are recorded separately from SLOW-only query logs. No existing expectations were changed.
Sign in to join this conversation.
No labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
kayg/calternal#951
No description provided.