Handle SIGTERM in your application for Kubernetes

Approach

On SIGTERM, stop accepting new work, finish in-flight work, close connections and exit before the grace period ends. Combine with a preStop delay for load balancer propagation.

Solution

import signal
import threading

shutting_down = threading.Event()

def on_sigterm(signum, frame):
    shutting_down.set()            # readiness endpoint starts returning 503, workers stop taking jobs

signal.signal(signal.SIGTERM, on_sigterm)

def worker_loop(queue):
    while not shutting_down.is_set():
        job = queue.get(timeout=1)
        if job:
            job.process()          # finish the current job before exiting

apiVersion: v1
kind: Pod
metadata: {name: worker}
spec:
  terminationGracePeriodSeconds: 120
  containers:
    - name: worker
      image: ghcr.io/acme/worker:2.0.0
      lifecycle: {preStop: {sleep: {seconds: 5}}}

Pitfalls

• The app runs under a shell that doesn't forward signals (use exec form).
• Jobs longer than the grace period; checkpoint and resume, or raise the grace period.

Variations

• For queue consumers, use visibility timeouts so unfinished messages return to the queue.
• For HTTP servers, use the framework's graceful shutdown (server.close, Shutdown(ctx)).

Further reading: Pod termination