Exercise 5: Health checks

In Kubernetes, liveness and readiness probes are essential for smoothly running applications. A probe is generally a REST GET call, but there are other types of probes available. Liveness probes are used to determine when to restart a container. For example, an application that is unhealthy and no longer responding to an API call would be restarted by OpenShift. Readiness probes determine when a container is ready to start receiving traffic. If a readiness probe fails, then the load balancer would deregister that service.

Create Readiness and Liveness Probes

The /info endpoint on the Example Health application is a great way to check whether the application is running and responding to API calls -- it responds with a simple JSON payload.

Go ahead and click on the deployment view. You should see something like node-s-2-i-openshift. Click Actions > Edit Deployment. Find the containers line probably line 38. Under the resources line, so line 46 past the following yaml. This will add a liveness probe to your deployment!

    livenessProbe:
      initialDelaySeconds: 5
      periodSeconds: 2
      httpGet:
        path: /info
        port: 8080

Now, next step is to add a readiness probe. Luckly it's on the same page, imediatly under the livenessProbe stanza you entered, paste the following:

    readinessProbe:
    initialDelaySeconds: 5
    timeoutSeconds: 2
    httpGet:
      path: /info
      port: 8080

This will make sure when a new deployment happens that it won't start until the /info path is available.

If you want to verify it, you can run the following command to check the status, you should see something like this in the output:

oc describe deployment node-s-2-i-openshift

Liveness:     http-get http://:8080/info delay=5s timeout=1s period=2s #success=1 #failure=3
Readiness:    http-get http://:8080/info delay=5s timeout=2s period=10s #success=1 #failure=3

If all works, everything should be the same. Let's check that the probes are really working though.

Inject Failure

Let's edit the probe with a typo to see what happens when it fails. Edit the health check and change the path for the readiness probe to /badpath. Wait a few minutes and check your deployment - you'll notice that 0/1 containers are ready:

Badpath

Dive into your events and you'll see that the probe is failing, causing the platform to try and repeatedly restart your pod.

Deeper Dive

Using health checks gives your OpenShift service layer better reliability and helps you start with a strong foundation.