Answers to Five Certified Kubernetes Application Developer CKAD Exam Questions (2021)

While doing some work with the open-source Kubernetes platform (K8s) (minikube, in this case) and reviewing CKAD exam material, I came across a page on Matthew Palmer’s website entitled “Practice Exam for Certified Kubernetes Application Developer (CKAD) Certification” and which contains five practice questions, which I’ll go over here. If you see a problem with anything I’ve done below, including inefficient solutions, please let me know in the comments.

We will focus on not just showing a possible solution to each problem but also verifying our work.

Since we’re limited to the kubernetes.io website when taking this test, this site is referenced where appropriate — in particular the kubectl cheat sheet page is not a bad place to start and we can search for what we need here as well.

Lastly, with each of these questions, we need to pay particular attention to the namespace requirement.

This article relies on minikube running on Ubuntu along with the Oh My Zsh shell.

The k command is defined as follows:

alias k='kubectl'

Let’s take a look at the first question.

Practice CKAD Exam Questions #1: Create two pods in a namespace

The first question from [1] is as follows:

“Create a namespace called ggckad-s0 in your cluster.
Run the following pods in this namespace.

1. A pod called pod-a with a single container running the kubegoldenguide/simple-http-server image

2. A pod called pod-b that has one container running the kubegoldenguide/alpine-spin:1.0.0 image, and one container running nginx:1.7.9

Write down the output of kubectl get pods for the ggckad-s0 namespace.“

Autogenerate the configuration files

We can hack together a pod by hand or we can autogenerate it, which should be the preferred option as we don’t want to waste time with this if we can avoid it. The following commands can be used to autogenerate each yaml file: one for the kubegoldenguide/simple-http-server container, one for the kubegoldenguide/alpine-spin:1.0.0 container, and one for the nginx:1.7.9 container.

kubectl run nginx --image=kubegoldenguide/simple-http-server --dry-run=client -o yaml > simple-http-server.yaml

kubectl run nginx --image=kubegoldenguide/alpine-spin:1.0.0 --dry-run=client -o yaml > alpine-spin.yaml

kubectl run nginx --image=nginx:1.7.9 --dry-run=client -o yaml > nginx.yaml

Modify the autogenerated configuration files

We can use these files to create the pod-a and pod-b yaml files, which are required for this question.

The configuration file for pod-a appears as follows (see: q1_dot_1_pod-a_configuration.yaml):

				
					apiVersion: v1
kind: Pod
metadata:
  name: pod-a
  labels:
    role: myrole
spec:
  containers:
    - name: web
      image: kubegoldenguide/simple-http-server
      ports:
        - name: web
          containerPort: 80
          protocol: TCP

And the configuration file for pod-b appears below (see: q1_dot_2_configuration.yaml):

				
					apiVersion: v1
kind: Pod
metadata:
  name: pod-b
  labels:
    role: myrole
spec:
  containers:
    - name: alpine-spin-container
      image: kubegoldenguide/alpine-spin:1.0.0
      ports:
        - name: web
          containerPort: 80
          protocol: TCP
    - name: nginx-container
      image: nginx:1.7.9
      ports:
        - name: web
          containerPort: 80
          protocol: TCP

We need to first create the namespace as per the instructions:

k create namespace ggckad-s0

Next, we can apply this configuration:

kubectl apply -f pod-[a or b].yaml --namespace ggckad-s0

The final instruction for this question indicates that we need to “[w]rite down the output of kubectl get pods for the ggckad-s0 namespace.” — this is accomplished as follows:

k get pods --namespace ggckad-s0

which yields the following:

Verify the solution by checking K8s logs and K8s events

Below we have the entire script that we use in this example. We check the logs for all containers in order to see that they’re running correctly.

k create namespace ggckad-s0
k apply -f pod-a.yaml --namespace ggckad-s0
k logs pod-a --namespace ggckad-s0
k apply -f pod-b.yaml --namespace ggckad-s0
k logs pod-b alpine-spin-container --namespace ggckad-s0
k logs pod-b nginx-container --namespace ggckad-s0

Running this script should yield the following output:

We can use the get events command to see the events that were collected.

k get events --namespace ggckad-s0

Running this command should yield the following output:

Notice that the nginx-container has started however there are no messages in the log that indicate that it’s actually running so we’ll check this by mapping port 19999 on the local machine to port 80 in the pod-b pod, which should point to the nginx-container which has the Nginx web server running on port 80.

kubectl port-forward pods/pod-b 19999:80 --namespace ggckad-s0

When we execute the line above, the output should look like what we have below (TPF):

Finally, we can now test that Nginx is running by browsing localhost:19999.

And that’s it for this question.

Practice CKAD Exam Questions #2: Expose configuration settings

The second question from [1] is as follows:

“All operations in this question should be performed in the ggckad-s2 namespace.
Create a ConfigMap called app-config that contains the following two entries:

‘connection_string’ set to ‘localhost:4096’
‘external_url’ set to ‘google.com’

Run a pod called question-two-pod with a single container running the kubegoldenguide/alpine-spin:1.0.0 image, and expose these configuration settings as environment variables inside the container.“

We need to first create the namespace as per the instructions:

k create namespace ggckad-s2

Create a ConfigMap

We need a ConfigMap, which can be created via the CLI as follows:

kubectl create configmap app-config --from-literal connection_string=localhost:4096 --from-literal external_url=google.com --dry-run -oyaml --namespace ggckad-s2

and which is defined in the following gist (see: app-config.yaml):

				
					apiVersion: v1
data:
  connection_string: localhost:4096
  external_url: google.com
kind: ConfigMap
metadata:
  creationTimestamp: null
  name: app-config
  namespace: ggckad-s2

Apply the configuration file

Next, we can apply this configuration making sure to include the specified namespace:

k apply -f ./app-config.yaml --namespace ggckad-s2

And the configuration file for the question-two-pod appears below (see question-two-pod.yaml):

				
					apiVersion: v1
kind: Pod
metadata:
  name: question-two-pod
  labels:
    role: myrole
spec:
  containers:
    - name: web
      image: kubegoldenguide/alpine-spin:1.0.0
      ports:
        - name: web
          containerPort: 80
          protocol: TCP
      envFrom:
        - configMapRef:
            name: app-config

We need to apply the pod configuration file, and we do so via the following command:

k apply -f ./question-two-pod.yaml --namespace ggckad-s2

Finally, we need to verify our work.

Verify the solution by inspecting the container environment variables

Verifying this solution is simple: In step one we’ll get CLI access to the container and in step two we’ll check the environment, if the environment variables have been set then we’re finished.

Step one appears as follows:

k exec -it question-two-pod -c web --namespace ggckad-s2 -- /bin/sh

and step two can be accomplished using the env command and visually inspecting the results.

And that’s it for this question.

Practice CKAD Exam Questions #3: Set user IDs and file system group IDs in containers

The third question from [1] is as follows:

“All operations in this question should be performed in the ggckad-s2 namespace. Create a pod that has two containers. Both containers should run the kubegoldenguide/alpine-spin:1.0.0 image. The first container should run as user ID 1000, and the second container with user ID 2000. Both containers should use file system group ID 3000.“

We need to first create the namespace as per the instructions:

k create namespace ggckad-s2

Apply the configuration file

And the configuration file for the question-three-pod appears below — review the namespace declaration — this allows us to apply the configuration file and not include the namespace via the command line (CLI); also, and more importantly, pay particular attention to the fsGroup and runAsUser key/value pairs (see: question-three-pod.yaml):

				
					apiVersion: v1
kind: Pod
metadata:
  name: alpine-spin
  # Note the namespace here does not appear in previous examples so
  # for example one and two we'll need to manually assign it on the
  # CLI.
  namespace: ggckad-s2
  labels:
    role: myrole
spec:
  securityContext:
    fsGroup: 3000
  containers:
    - name: alpine-spin-a
      securityContext:
        runAsUser: 1000
      image: kubegoldenguide/alpine-spin:1.0.0
      ports:
        - name: web
          containerPort: 80
          protocol: TCP
    - name: alpine-spin-b
      securityContext:
        runAsUser: 2000    
      image: kubegoldenguide/alpine-spin:1.0.0
      ports:
        - name: web
          containerPort: 80
          protocol: TCP

We can apply this configuration as follows:

k apply -f ./question-three-pod.yaml

Finally, we need to verify our work.

Verify the solution using the id command

We need to get CLI access to our container and then verify that the settings are correct and the following command can be used to do exactly that:

k exec -it question-three-pod -c alpine-spin-[a or b] --namespace ggckad-s2 -- /bin/sh

Once we have access we can check that the fsGroup and runAsUser key/value pairs have been set correctly — we’ll get this information by using the id command.

And that’s it for this question.

Practice CKAD Exam Questions #4: Configuring liveness and readiness probes

The fourth question from [1] is as follows:

“All operations in this question should be performed in the ggckad-s4 namespace. This question will require you to create a pod that runs the image kubegoldenguide/question-thirteen. This image is in the main Docker repository at hub.docker.com.

This image is a web server that has a health endpoint served at ‘/health’. The web server listens on port 8000. (It runs Python’s SimpleHTTPServer.) It returns a 200 status code response when the application is healthy. The application typically takes sixty seconds to start.

Create a pod called question-13-pod to run this application, making sure to define liveness and readiness probes that use this health endpoint.“

Unlike previous questions, this one specifically tells us that the kubegoldenguide/question-thirteen image is in the main Docker repository at hub.docker.com. According to the Kubernetes Images:Image Names documentation, “[i]f you don’t specify a registry hostname, Kubernetes assumes that you mean the Docker public registry” so no further action is necessary with respect to this detail.

Our liveness and readiness probe configuration settings

In our proposed solution, we’ll add a buffer and start the livenessProbe at 75 seconds, which is 15 seconds after the typical expected startup time, as indicated in the question four details. The 15 second buffer is an arbitrarily chosen number and the assumption is that, realistically, the server should have been started by this time; a similar example appears in [5] and refer also to [7] and [8].

Note that as of Kubernetes 1.16 it is possible to define a startupProbe such that the livenessProbe will not be started until an initial OK is returned.

Our pod configuration can be seen here (see: question-four-pod.yaml).

				
					apiVersion: v1
kind: Pod
metadata:
  name: question-13-pod
  namespace: ggckad-s4
  labels:
    role: myrole
spec:
  containers:
    - name: question-thirteen
      image: kubegoldenguide/question-thirteen
      ports:
        - name: web
          containerPort: 8000
          protocol: TCP
      readinessProbe:
        httpGet:
          path: /health 
          port: 8000
        initialDelaySeconds: 60
        periodSeconds: 5
      livenessProbe:
        httpGet:
          path: /health
          port: 8000
        initialDelaySeconds: 75
        periodSeconds: 10
        timeoutSeconds: 5

Finally, we need to verify our work.

Verify our solution by checking logs

In this example, we should expect to see log messages for the readinessProbe up until ~ 75 seconds after the pod was started, at which point we should see log messages appear for both the readinessProbe as well as the livenessProbe.

kubectl logs question-13-pod -c question-thirteen --v 4 --namespace ggckad-s4

Unfortunately, the output from this command does not tell us which probe is calling the endpoint.

As an exercise for the reader, can we determine exactly which probe is calling the health endpoint? The Kubernetes documentation seems to indicate that this is possible however at this time I do not have an example which demonstrates this.

And that’s it for this question.

Practice CKAD Exam Questions #5: Create a deployment with replicas

The fifth question from [1] is as follows:

“All operations in this question should be performed in the ggckad-s5 namespace. Create a file called question-5.yaml that declares a deployment in the ggckad-s5 namespace, with six replicas running the nginx:1.7.9 image.

Each pod should have the label app=revproxy. The deployment should have the label client=user. Configure the deployment so that when the deployment is updated, the existing pods are killed off before new pods are created to replace them.“

We need to check the labels for both the deployment and pods and then we need to, as per [1] “[c]onfigure the deployment so that when the deployment is updated, the existing pods are killed off before new pods are created to replace them”.

We have two deployment strategies available: RollingUpdate (the default), and Recreate [14]. Given the definitions for RollingUpdate and Recreate in this case we do not want to use the default and instead need to assign Recreate to the .spec.strategy.type.

Our deployment configuration

Our deployment configuration can be seen here (see: question-five-deployment.yaml):

				
					apiVersion: apps/v1
kind: Deployment
metadata:
  name: question-5
  namespace: ggckad-s5
  labels:
    client: user
spec:
  replicas: 6
    strategy:
      type: Recreate
  selector:
    matchLabels: {
      app: revproxy
    }
  template:
    metadata: {
      labels: {
        app: revproxy
      }
    }
    spec:
      containers:
      - name: web
        #
        # We start with nginx:1.7.9 but will cause a rolling update when we change this
        # to latest.
        #
        image: nginx:1.7.9
        ports:
          - name: web
            containerPort: 80
            protocol: TCP

We can apply the file the same as we did in the previous question:

k apply -f ./question-5.yaml

Finally, we need to verify our work.

Verify our solution by checking the rollout status

We need to check the labels for both the deployment and pods and we also need to ensure that the requirement is met that when the deployment is updated existing pods are killed off before new pods are created to replace them.

Review deployment labels

We can use the following command to view the deployment labels:

k get deployments --show-labels --namespace ggckad-s5

And we can use the following command to view the various pod labels:

k get pods --show-labels --namespace ggckad-s5

Below we can see the output of applying the question-5.yaml file and then executing these two commands; the yellow arrows show that the labels have been set as per the specification:

Note that our deployment relies on the nginx:1.7.9 image. This image is old so if we update this to nginx:latest and apply the question-5.yaml file again then we see that new containers are deployed while the old containers are gradually terminated. The following image demonstrates this behavior:

Review the rollout status

Next, we repeat the same exercise only here we check the rollout status using the following command [11]:

kubectl rollout status deployment.v1.apps/question-5 --namespace ggckad-s5

and we can see from a different angle that the old replicas are being terminated while the new ones are starting.

Lastly we can reproduce the same behavior by executing the following command:

kubectl rollout restart deployment question-5 --namespace ggckad-s5

and in this case we don’t even need to update the nginx version in the question-5.yaml file.

Let’s take a look at an update when the deployment strategy type is set to RollingUpdate — this is the default and it is also the incorrect choice given the specification.

An incorrect solution using the RollingUpdate deployment strategy

The same example but with a RollingUpdate deployment strategy would be incorrect. We include the output of this example below for comparison purposes.

Next, we repeat the same exercise only here we check the rollout status using the following command [11]:

kubectl rollout status deployment.v1.apps/question-5 --namespace ggckad-s5

and we can see from a different angle that the old replicas are being terminated while the new ones are starting.

And that’s it for this question, and for this article for that matter — on to the conclusion!

Practice CKAD Exam Questions Conclusion

This article took several hours to write and was done in part as an investigation into how these problems could be solved as an introduction into studying for the CKAD exam for me, specifically.

There are several good articles available if you’re beginning the process of preparing for this certification which provide guidance about the test structure and how to study for the exam and I include several links below.

Which Kubernetes certification is right for you?
Certified Kubernetes Application Developer (CKAD) exam preparation tips
How to pass CKAD exam in first attempt ? Tips & Tricks in Kubernetes
CKAD Exam Preparation Study Guide (Certified Kubernetes Application Developer)
CKAD Exam Guide
Be fast with Kubectl 1.19 CKAD/CKA
Practice Enough With These 150 Questions for the CKAD Exam (noted in the comments, apparently these are not questions that would appear on the exam however this is still a good list)

If you have other resources or references that you’ve found helpful in preparing to sit for this exam, please include them in the comments.

A Poem About Passing The CNCF CKAD Exam

I’ve been working with ChatGPT lately and while refreshing this article I instructed it to write a poem using iambic pentameter about passing the Cloud Native Computing Foundation (CNCF) Certified Kubernetes Application Developer Exam (CKAD) — for fun and a bit of humor, I’ve included the result here.

To pass the CNCF CKAD,
A journey lies ahead, of skill and will,
But fear not, for with focus, you’ll be glad,
And mastery of K8s, you’ll fulfill.

The exam will test your skills and your mind,
With challenges that’ll put you to the test,
You’ll need to be both swift and well-aligned,
And know the tools that K8s puts to the test.

You’ll learn about deployments, pods, and more,
And how to scale your apps with K8s power,
With services, ingresses, and secrets galore,
And networking that spans across the tower.

You’ll write YAML with clarity and ease,
And code with vim and shell, as it demands,
And master K8s commands that surely please,
With efficiency and focus that commands.

So study hard and learn the K8s way,
And pass the CKAD with mastery,
And join the ranks of those who’ve paved the way,
To a world of cloud-native harmony.

Kubernetes Quiz

Refer to Question 1: Why don’t we need to include the specific container in the port-forward command, copied below for convenience, when both the alpine-spin-container as well as the nginx-container both rely on containerPort 80?

kubectl port-forward pods/pod-b 19999:80 –namespace ggckad-s0

Note: port-forward only allows us to specify the pod and not the pod + container.
Refer to Question 1: What would happen if we tried to start pod-b but instead of the alpine-spin-container we had two nginx-container s?
What is the difference between a deployment strategy of “Recreate” juxtaposed with “RollingUpdate”?