Operations¶

Logs¶

Targetlog writes JSON lines to stderr: startup, background work, warnings and errors (requests are not logged one by one). TARGETLOG_LOG_LEVEL=debug adds more detail.

Metrics¶

Prometheus metrics are served on TARGETLOG_METRICS_LISTEN (default :9090) at /metrics, on a separate port so they never leak through the ingress. The Helm chart can create a PodMonitor.

Backups¶

All state is in PostgreSQL, including print jobs, sessions and realtime events. Back up the database (for example with the CloudNativePG backup of the cluster); the containers hold nothing worth saving.

Updates¶

Deploy a newer image. Migrations run at startup, so a new version migrates the database before it serves requests. Read the changelog before upgrading.

Background work¶

Every replica runs small janitors (expired sessions, old print jobs, requeueing unconfirmed jobs). The reservation ticker announces reservation starts and ends every 30 seconds; a PostgreSQL advisory lock makes sure only one replica does it at a time.