Termination Grace Period

The amount of time Kubernetes waits for a container to gracefully shut down before forcefully terminating it.

Default Value

30 seconds

How It Works

1. Pod receives SIGTERM signal
2. Container has terminationGracePeriodSeconds to shut down gracefully
3. If container doesn't exit, SIGKILL is sent to force termination
4. Pod is removed

Setting Termination Grace Period

spec:
  terminationGracePeriodSeconds: 60
  containers:
  - name: app
    image: myapp:1.0

Typical Use Cases

Web Application

Need time to drain connection pools and close active requests.

terminationGracePeriodSeconds: 30

Database

Needs time to flush writes and close gracefully.

terminationGracePeriodSeconds: 120

Long-Running Job

Needs significant time to save state.

terminationGracePeriodSeconds: 300

Best Practices

1. Combine with preStop hook - preStop should complete before grace period expires
2. Match with app shutdown time - set grace period longer than app's shutdown time
3. Default 30s is usually sufficient - increase only if needed
4. Monitor actual shutdown times - adjust based on real behavior

Example: Web Server with Graceful Shutdown

apiVersion: v1
kind: Pod
metadata:
  name: web-server
spec:
  terminationGracePeriodSeconds: 45
  containers:
  - name: app
    image: nginx:latest
    lifecycle:
      preStop:
        exec:
          command: ["/bin/sh", "-c", "nginx -s quit; while killall -0 nginx 2>/dev/null; do sleep 1; done"]
  restartPolicy: Always

This gives nginx up to 45 seconds to gracefully shut down before being killed.