Running pipelines and executions
Run on the pipeline list enqueues work. Open Executions for status. The table polls about every 3 seconds while any run is active.
Status
| Status | Meaning |
|---|---|
scheduled | Waiting for a worker |
claimed | A worker owns it, run starting |
running | Pipeline executing |
completed | Finished; inspect records_out / errors_count |
failed | Exhausted retries or unrecoverable error (last_error) |
cancelled | Cancel requested and acknowledged |
Cancel is available while scheduled, claimed, or running. It sets cancel_requested; the worker notices on the next heartbeat.
Default max_retries is 3 with exponential backoff. The list API is capped at 50 rows.
Counters
records_out— records written to the sink after a successful drainerrors_count— stage errors (bad records). A run can complete with errors_count > 0- Sink write happens after the idle drain. Abandoned runs can leave partial sink output; there is no automatic rollback
Workers page
Queue cards, connected instances (id, slots, last seen), and active leases. Useful when a run stays claimed. Details: Operators: Workers.