Create a Deployment

The pod does not provide enough automation to revive a failed process. The deployment ensures that a given amount of pods is always started.

We will change the configuration to use a deployment instead of a manually created pod.

Delete the pod

1. Delete the previously created pod: kubectl delete pod calculator-backend
2. The result will be:

   pod "calculator-backend" deleted
   

Write the deployment definition

1. Delete all contents of your backend.yml.
2. Copy the contents from kubernetes-workshop/k8s-examples/deployment.yml.
3. Under .spec.template.spec you’ll find the similar fields like in the pod configuration, make the same changes as previously.
4. The two new configuration items livenessProbe and readinessProbe need to configured with the correct port of your service, otherwise the health checks will fail.
5. Ensure that the deployment has the name calculator-backend.
6. Also ensure that .template.metadata.labels.'app.kubernetes.io/name' is set to the same value, we will need this later.
7. The last required change is to set .spec.selector.matchLabels to the same content as the labels.

Apply the pod definition

Like previously, use kubectl apply -f k8s-files/.

The console should display the following:

deployment.apps "calculator-backend" created

Inspect the created resources

The previous command as created multiple resources.

1. With kubectl get deployment you can see the created deployment:

   NAME                 DESIRED   CURRENT   UP-TO-DATE   AVAILABLE   AGE
   calculator-backend   1         1         1            1           3m
   

2. With kubectl get pods the automatically created pod is returned:

   NAME                                  READY     STATUS    RESTARTS   AGE
   calculator-backend-557c7f66d8-zdm9b   1/1       Running   0          3m
   

3. In comparison to the last pod the name changed, -557c7f66d8-zdm9b was added as a suffix.
   557c7f66d8 is the identifier for this specific configuration inside the deployment.
   zdm9b is a random string that identifies this pod, a deployment can start multiple pods, each will get its own random id.

Access the pod

The pod can be accessed like before, just enter your pod name to get the URL:

$GCLOUD_SHELL_URL/api/v1/namespaces/default/pods/<pod-name>/proxy/

Trigger a restart

1. Run kubectl get pods -w to show changes on the pods.
2. Crash the pod again with an invalid expression, e.g. sqrt(Hello World)
3. The output should now look like this:

   NAME                                  READY     STATUS    RESTARTS   AGE
   calculator-backend-557c7f66d8-zdm9b   1/1       Running   0          12m
   calculator-backend-557c7f66d8-zdm9b   0/1       Error     0          12m
   calculator-backend-557c7f66d8-zdm9b   0/1       Running   1          12m
   calculator-backend-557c7f66d8-zdm9b   1/1       Running   1          12m
   

4. Afterwards you can test again to calculate results for valid expressions, the pod will now again answer your requests.