Monitoring executions
Two views into executions:
- Cross-job:
/scheduler/executions— every execution in the workspace, newest first. - Per-job:
/scheduler/:jobId/executions— just the executions for one schedule.
Filters
Filter executions by status (PENDING, RUNNING, COMPLETED, FAILED), by date range, or by schedule. The filter set is sticky to the URL — share a link and the recipient sees the same view.
What each row shows
| Column | What it means |
|---|---|
| Status | Lifecycle state — RUNNING means a worker is currently processing it |
| Schedule | Source schedule name; click to jump to its detail page |
| Started / Completed | Timestamps in your local timezone |
| Duration | Wall-clock time the executor took |
| Retry attempt | 0 for the first run; higher for retries |
| Worker | Worker pod that picked the job up; useful when investigating execution skew |
Inspecting a single execution
Click a row to open execution detail. You see:
- The full executor output (JSON)
- The exact config snapshot used at fire time (so editing the schedule later doesn't change the historical record)
- Error message + stack trace if failed
- Distributed-trace IDs (
correlationId,traceId,spanId) — copy these into Grafana / Tempo to see the full call chain
Acting on an execution
- Retry a failed execution — fires a brand-new execution with the same config snapshot.
- Cancel a running execution — best-effort signal to the worker; not all executor types support immediate cancel.
- Open dead-letter queue entry — if retries were exhausted, the execution is in the DLQ. See 05-dead-letter-queue.md.
Real-time updates
Both pages subscribe to live execution updates over GraphQL Subscriptions. New executions, status changes, and completions appear without manual refresh. If the connection drops, the page falls back to a periodic refresh; a small banner tells you when this happens.