Lab 2 - Probes¶

20 min Beginner

Problem¶

The energy-shield service has two problems in production.

Container health issues¶

After running for a short while, the application goes into an unhealthy state and starts answering requests with errors. The process doesn't crash, so the container keeps running and OpenShift never restarts it.

The application has an internal health endpoint that reports whether it's healthy: /healthz on port 8080.

• Add a liveness probe that checks /healthz on port 8080. If the endpoint returns an error, the cluster restarts the container.

Container startup issues¶

New Pods take a few seconds after startup before they can serve requests, and some users hit errors during that window.

• Add a readiness probe that checks /healthz on port 8080, so the Pod only receives traffic when it's ready.
• On both probes, set an initial delay of 5 seconds and check every 5 seconds.

Setup¶

oc new-project lab2

Here is the Pod manifest. Add the probes, then create the Pod in the cluster:

apiVersion: v1
kind: Pod
metadata:
  name: energy-shield-service
spec:
  containers:
    - name: energy-shield
      image: registry.k8s.io/e2e-test-images/agnhost:2.56
      args: ["liveness"] # (1)!
      ports:
        - containerPort: 8080

1. The liveness mode of this test image simulates our buggy service. /healthz returns 200 OK for about 10 seconds, then returns 500 forever, while the process keeps running.

Verification¶

Watch the Pod for about a minute:

oc get pod energy-shield-service -w

With working probes you should see:

• READY shows 0/1 for the first few seconds, then 1/1 once the readiness probe passes.
• About 20 seconds in, the app goes unhealthy. READY drops back to 0/1 (readiness probe), and a few seconds later the RESTARTS count goes up (liveness probe).
• The cycle repeats. Each restart gives the app a fresh start, and it only gets traffic while it's healthy.

Check the events to see the probes in action:

oc describe pod energy-shield-service | grep -A 10 Events

Look for Liveness probe failed: HTTP probe failed with statuscode: 500 followed by Container energy-shield failed liveness probe, will be restarted.

Cleanup¶

oc delete project lab2