Readiness Probes

By default, Kubernetes assumes that when a Pod reaches the Ready stage, it is capable of serving requests.
In real-world scenarios, the application may take a few milliseconds (or more) to actually become ready due to initial startup delays.
We need a way to tie the Ready condition to the actual state of the application running inside the container.
Kubernetes provides several ways to define when applications are ready to serve requests:
- Set up an HTTP endpoint (for web servers)
- Use a TCP socket check (e.g., for databases)
- Write and execute a custom script inside the container to verify readiness

How to Define a Readiness Probe

In the Pod YAML definition, under the spec.containers section, define a readinessProbe:

containers:
  - name: web-server
    image: web-server-image
    readinessProbe:
      httpGet:
        path: /api/ready
        port: 8080

readinessProbe:
  tcpSocket:
    port: 3306

readinessProbe:
  exec:
    command:
      - python3
      - check_status.py

initialDelaySeconds: If your application takes some time to warm up, use this field to specify the delay before the first probe is initiated.
periodSeconds: Specifies how often the probe should be performed.
failureThreshold: By default, Kubernetes tries 3 times checking container is not ready. You can increase this threshold to allow more retries before failing.