PreStop

PreStop is a lifecycle hook used in container orchestration, most commonly in Kubernetes, to run cleanup actions just before a container is terminated. It is defined in a Pod spec under the container's lifecycle field and can be implemented as an exec command or an HTTP GET request.

During termination, Kubernetes first sends a SIGTERM to the container, then executes the preStop hook, and only after the termination grace period elapses does it forcibly terminate the container. This sequence gives the application time to perform graceful shutdown tasks, such as closing network connections, flushing logs, draining work queues, or deregistering from services.

If the preStop hook completes before the grace period ends, the container can exit cleanly; if it takes too long, the container is killed when the grace period expires. The hook is optional and must complete within the overall terminationGracePeriodSeconds for the Pod.

PreStop supports two action types: exec, which runs a command inside the container, and httpGet, which performs an HTTP request to a specified path and port. Example: lifecycle: preStop: exec: command: ["/bin/sh", "-c", "/app/cleanup.sh"]

Best practices include keeping preStop short to avoid delaying pod termination, ensuring cleanup is idempotent, and avoiding blocking operations on the main application thread. For long-running cleanup, configure a suitable terminationGracePeriodSeconds and consider performing work asynchronously outside the container where possible.