Skip to content

Runs and diagnostics

Job details contain run history, and /runs/:id opens a run detail page. The page shows status, attempt information, timestamps, row counts, cancellation state, and a bounded diagnostic log. Statuses include pending, running, cancel-requested, cancelled, retry-scheduled, succeeded, and failed.

ReplicaDB run detail and bounded diagnostics
Run detail shows status, metrics, and bounded operational logs.

Cancel run persists cancellation intent before the best-effort local signal. The UI displays a persisted cancellation warning when the sink outcome may be indeterminate. Retry run starts a new attempt from the beginning; it is not resume behavior. Retry availability depends on the run status and permission.

Trigger a manual run from job detail only after reviewing its definition and bindings. Run detail identifies the attempt, current status, timestamps, row counts, and retry eligibility. A pending or running state describes durable work that a worker may claim or already owns; it is not proof that all sink writes are reversible.

Request cancellation when continuing the run is riskier than an indeterminate sink. Wait for the terminal outcome before retrying. A retry creates a new attempt linked to the previous run and re-resolves current datasource profiles; it never resumes from a row offset. Compare the old and new attempts before declaring recovery complete.

Run logs are operational data. They are bounded, redacted, and not a place for credentials, keyrings, lease tokens, or resolved datasource security. A log load failure offers a retry action and does not turn an absent log into a successful run.

Use the detail’s status and warnings before interpreting row counts or logs. For complete mode, interruption can leave the sink empty, truncated, or partially populated; verify source and sink outcomes before a retry. A missing or truncated log is evidence of bounded diagnostics, not evidence that a run succeeded. Escalate persistent queue, lease, or worker-health symptoms through failure recovery and health and metrics.