Pod Lifecycle: Init Containers, Probes & Termination

In this tutorial, you'll learn about Pod Lifecycle: Init Containers, Probes & Termination. We cover key concepts, practical examples, and best practices to help you understand and apply this topic effectively.

Every Kubernetes pod goes through a lifecycle from pending to running to terminated, with init containers running setup tasks, probes checking health, and hooks handling graceful shutdown.

What You'll Learn

This tutorial covers pod phases, init containers, startup and liveness probes, readiness probes, container lifecycle hooks, and graceful termination.

Why It Matters

Misconfigured probes cause cascading failures in production. Init containers that fail block deployments. Without proper termination handling, applications lose data during rolling updates or node drains.

Real-World Use

Netflix uses sophisticated liveness and readiness probes in their Spinnaker-managed deployments to ensure zero-downtime canary deployments, with init containers warming cache before the main application starts.

Pod Phases

A pod moves through five phases: Pending, Running, Succeeded, Failed, and Unknown.

# Watch pod phase transitions
kubectl get pods -w

# Examine pod status details
kubectl describe pod my-app

The describe output shows the phase, conditions, container states, and recent events.

Init Containers

Init containers run before the application containers start. They perform setup tasks like database migrations or permissions configuration.

apiVersion: v1
kind: Pod
metadata:
  name: app-with-init
spec:
  initContainers:
  - name: db-migrate
    image: migrate/migrate:latest
    command: ['migrate', 'up']
    env:
    - name: DB_URL
      value: "postgres://db:5432/app"
  containers:
  - name: app
    image: myapp:1.0

Init containers run sequentially. If one fails, the pod restarts the entire init sequence.

# Check init container status
kubectl get pods -o wide

# View init container logs
kubectl logs app-with-init -c db-migrate

Container Probes

Kubernetes supports three probe types: startup, readiness, and liveness.

Startup Probe

Startup probes check if the application has started. They are ideal for slow-starting containers.

startupProbe:
  httpGet:
    path: /health/startup
    port: 8080
  initialDelaySeconds: 0
  periodSeconds: 10
  failureThreshold: 30

Liveness Probe

Liveness probes check if the container is still running. If they fail, the kubelet restarts the container.

livenessProbe:
  httpGet:
    path: /healthz
    port: 8080
  initialDelaySeconds: 5
  periodSeconds: 10
  failureThreshold: 3

Readiness Probe

Readiness probes check if the container is ready to serve traffic. Failed probes remove the pod from service endpoints.

readinessProbe:
  httpGet:
    path: /ready
    port: 8080
  initialDelaySeconds: 3
  periodSeconds: 5
  failureThreshold: 2

Container Lifecycle Hooks

Kubernetes provides postStart and preStop hooks for lifecycle events.

preStop Hook for Graceful Shutdown

The preStop hook runs before the container terminates, allowing graceful shutdown.

containers:
- name: app
  image: myapp:1.0
  lifecycle:
    preStop:
      exec:
        command: ["/bin/sh", "-c", "sleep 10 && /usr/local/bin/drain-connections"]

Termination Grace Period

The terminationGracePeriodSeconds controls how long Kubernetes waits after sending SIGTERM before sending SIGKILL.

spec:
  terminationGracePeriodSeconds: 60

Graceful Termination Flow

When a pod is deleted, Kubernetes sends SIGTERM, waits the grace period, then sends SIGKILL.

# Delete pod gracefully
kubectl delete pod my-app

# Delete pod immediately
kubectl delete pod my-app --grace-period=0 --force

Practice Questions

1. What is the difference between liveness and readiness probes? Liveness probes restart unhealthy containers. Readiness probes control traffic routing to pods.

2. When would you use a startup probe instead of increasing initial delay? Startup probes are better for applications with variable startup times, like Java applications or legacy services.

3. What happens if an init container fails? The pod restarts all init containers from the beginning until they all succeed.

4. What is the purpose of the preStop hook? It runs custom cleanup commands before the container receives SIGTERM, enabling graceful shutdown.

5. How does terminationGracePeriodSeconds affect shutdown? It sets the time between SIGTERM and SIGKILL, giving the application time to clean up connections.