Skip to main content

Scheduled & manual jobs

Jobs run to completion on the same runtime as services. Use them for migrations, backups, batch work, or anything that runs and exits.

Create a job​

POST /v1/jobs (Operator) with an image source (Git or external image), a command, optional env, an optional cron schedule, and max_retries:

curl -fsS -X POST \
  -H "Authorization: Bearer $TOKEN" \
  -H 'Content-Type: application/json' \
  -d '{
        "project_id":"'"$PID"'",
        "name":"nightly-backup",
        "source":{"ExternalImage":{"image":"ghcr.io/you/backup:latest"}},
        "command":["/bin/backup","--all"],
        "schedule":"0 3 * * *",
        "max_retries":2
      }' \
  https://your-node.example.com/v1/jobs

The cron schedule uses standard 5-field syntax (here: 03:00 daily). Omit it for a manual-only job.

Trigger a run by hand​

curl -fsS -X POST -H "Authorization: Bearer $TOKEN" \
  https://your-node.example.com/v1/jobs/$JOB_ID/run

Inspect runs​

curl -fsS -H "Authorization: Bearer $TOKEN" \
  https://your-node.example.com/v1/jobs/$JOB_ID/runs

Each JobRun records its status (Pending, Running, Succeeded, Failed, Skipped), attempt number, exit code, and timestamps. The scheduler guards concurrency so a long run does not overlap its next tick. See ADR-010.