Skip to content

Graceful shutdown

DataPress catches SIGTERM and SIGINT (Ctrl+C) and drains in-flight requests before exiting. The grace period is configurable.

How it works

  1. The signal handler closes the listening socket — no new connections.
  2. Already-accepted requests get up to shutdown_timeout_secs seconds to finish.
  3. After the grace period (or once the last in-flight request completes), workers are stopped and the process exits cleanly.

The startup log records which signal triggered shutdown:

INFO  Received SIGTERM, shutting down gracefully (up to 30s for in-flight requests)...
INFO  Shutdown complete.

On non-Unix platforms (Windows), only Ctrl+C is honoured — SIGTERM isn't a thing there.

Configure the grace period

[server]
shutdown_timeout_secs = 30    # default

DataPressConfig(backend="datafusion", port=8000,
                shutdown_timeout_secs=30)

Choose the value based on your workload:

  • Short (5–10 s) — interactive UI workloads where every handler returns quickly. Faster pod restarts during rolling deploys.
  • Default (30 s) — bulk queries, medium aggregations.
  • Long (120+ s) — large parquet exports, expensive aggregations on big datasets. Match this to the longest realistic handler duration.

Tuning with the orchestrator

The orchestrator's "terminationGracePeriodSeconds" must be strictly greater than shutdown_timeout_secs, otherwise the pod gets SIGKILL'd before DataPress can finish draining.

spec:
  terminationGracePeriodSeconds: 45    # > shutdown_timeout_secs (30)
  containers:
    - name: datapress
      # ...
      env:
        - name: RUST_LOG
          value: info

For zero-downtime rolling deploys, also ensure:

  • The readiness probe (/readyz) is wired up so the load balancer stops sending new traffic the moment a pod begins shutting down.
  • The new pod is fully ready before the old one is torn down.