In Kubernetes, health probes are mechanisms that allow the kubelet to determine the health of a container.

There are two main types of probes: Liveness and Readiness probes

These ensure your application runs smoothly and reliably

1. Liveness Probe

• Purpose: Determines if the container is still running. If the liveness probe fails, Kubernetes restarts the container.

• Use Case:

  • When the application gets stuck (e.g., deadlocks) and needs to be restarted.
  • Ensures the application is always in a functioning state.

• Behavior:

  • If the liveness probe fails, Kubernetes kills the container and restarts it, following the container’s restart policy.

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

• HTTP GET checks if the /healthz endpoint responds.
• Restarts the container if the check fails.

2. Readiness Probe

Purpose: Determines if the container is ready to serve requests. If the readiness probe fails, the container is removed from the service’s endpoint, meaning traffic will not be sent to it.

Use Case:

• When the application needs time to initialize or become fully ready to process traffic.
• Ensures only healthy instances handle incoming requests.

Behavior:

• If the readiness probe fails, the container remains running but is temporarily removed from the load balancer.

readinessProbe:
  httpGet:
    path: /readiness
    port: 8080
  initialDelaySeconds: 5
  periodSeconds: 10

• HTTP GET checks if the /readiness endpoint is available.
• Marks the pod as ready when the endpoint responds successfully.

![[Pasted image 20250128162505.png]]

3. Startup Probe

A third type of probe introduced in Kubernetes v1.16:

• Purpose: Specifically designed to handle slow-starting applications. Used to check whether the application has started successfully.

• Behavior:

  • Once the startup probe succeeds, both liveness and readiness probes take over.

• Use Case:

  • If your application takes a long time to initialize and liveness probes restart the container prematurely.

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