Chapter 1 showed the kubelet restarting a container when its process exits. A process can also be alive and useless: deadlocked, out of file handles, stuck in a loop. From the outside it looks fine. A liveness probe lets the kubelet ask.
The check
yaml
containers:
- name: web
image: nginx:1.27
livenessProbe:
httpGet:
path: /
port: 80
periodSeconds: 10
failureThreshold: 3Every periodSeconds the kubelet sends GET / to the container's port. Status 200 to 399 is a pass. Anything else, or no answer within timeoutSeconds (default 1), is a failure.
probe fails
10s
fails
10s
fails
restart the container
The Pod stays. Its IP stays. Only the container is killed and started again, and RESTARTS goes up by one.
Three kinds of probe
| Kind | Passes when |
|---|---|
httpGet | the response status is 200 to 399 |
tcpSocket | something accepts a connection on port |
exec | the command exits 0 inside the container |
Reading it
kubectl describe pod prints the probe on one line and the failures as Events:
text
Liveness: http-get http://:80/healthz delay=0s timeout=1s period=10s #success=1 #failure=3
Events:
Warning Unhealthy Liveness probe failed: HTTP probe failed with statuscode: 404
Normal Killing Container web failed liveness probe, will be restartedA rising RESTARTS column together with Unhealthy events is a liveness problem, and the message says why.