Resource Creation Tips for the Kubernetes CKA / CKD Certification Exam

In addition to having to understand the relevant Kubernetes concepts, one needs to be able to quickly create Kubernetes resources to be successful in the exam. Here are some tips that I have picked up as part of my preparations.

Apparently, there were a number of changes in the imperative resource creation commands between versions 1.17.X and 1.18.X; this article is written against version 1.18.3 and as such is aligned with the current exams.

Also, I cannot recommend enough the practice exams available at CKA CKAD Simulator. Before using the simulator, I did not appreciate the need to be able to quickly create resources.

The first time-saver is simply creating an alias for the kubectl command at the start of the exam.

alias k="kubectl"

While this seemed unnecessary at first, after typing kubectl hundreds of times during a practice exam, I learned that this is indeed quite important.

Resources

Some (but not all) resources can be created using the create command; the full list of resource kinds that can be created can seen using the following command:

k create --help

note: Some kinds of resources, e.g., DaemonSet will require copying a sample configuration from the Kubernetes documentation into a file, updating it, and applying it.

Then for each resource kind, one can get the specific options by using --help, e.g.,:

k create configmap --help

The bottom of the help output provides the required and common optional options (the top provides the complete list of options):

kubectl create configmap NAME [--from-file=[key=]source]
[--from-literal=key1=value1] [--dry-run=server|client|none] [options]

note: In some cases, options can be repeated. For example, the ConfigMap --from-literal=key1=value1 option can be repeated.

As not all resource configuration values can be set using the command line options, it is common to use the --dry-run=client -o yaml options with the create command to output the resource configuration into a file; where it can be then updated before applying it (with k apply -f my-resource.yaml).

Other benefits of using configuration files are:

• Resources can be deleted using the configuration file, e.g., with the k delete -f my-resource.yaml command
• Resources configuration files can also be later updated and re-applied

Another big time-saver is to set a shell variable at the start of the exam as follows:

do="--dry-run=client -o yaml"

and then use it as in the following example:

k create configmap example --from-literal=a=apple $do > my-resource.yaml

One important observation is that the generated configuration files do not include the .metadata.namespace key / value; even if you supply the -n my-namespace option. So, if you need to create resources in a Namespace other than default, you can do either:

• Add the .metadata.namespace key / value to the configuration file
• Supply the namespace when applying the configuration file, e.g., k apply -f my-resource.yaml -n my-namespace

note: There is also a way to set the namespace used for your current configuration context; k config set-context --current --namespace=my-namespace. But, I choose to not use this approach as I would undoubtedly forget what namespace is currently being used.

Pods

You may have noticed that you cannot create Pods using the create command and yet creating pods is a common activity. There is another command with numerous options to do this:

k run --help

The bottom of the help output provides the required and common optional options:

kubectl run NAME --image=image [--env="key=value"] [--port=port]
[--dry-run=server|client] [--overrides=inline-json] [--command] -- [COMMAND]
[args...] [options]

note: The env option can be repeated but the port option cannot. The overrides option is more obscure (I have never used it).

The top of the help output provides the complete list of optional options; some of which are more commonly used:

• --expose=true : Will also create a Service exposing the Pod
• --labels=a=apple,b=banana : Create the Pod with the supplied label
• --limits=cpu=200m,memory=512Mi : Create the Pod with supplied limits
• --requests=cpu=100m,memory=256Mi : Create the Pod with supplied requests
• --restart=Never : Create the Pod that never restarts containers
• --serviceaccount=my-sa : Create the Pod with the supplied Service Account name

The following options are not to be used when performing a dry run but are useful in special cases:

• -it : A combination of the -i and -t options; can be used to attach to the container, e.g., displays its standard output
• --rm : Wait for the Pod to succeed / fail and then delete it

For example, one common scenario of creating a Pod without a dry run is to hit a URL from a web browser in a container, e.g.:

k run tmp --image=busybox --restart=Never -it --rm -- wget -O- -T 3 https://www.google.com

note: The wget -O- option redirects output to standard out; the -T 3 option sets the timeout to three seconds.

As you likely observed, you will likely have to modify the generated Pod configuration file before applying it. Here are some common scenarios:

• Specify the container’s name
• Create a Pod with multiple containers

Other scenarios require adding new keys; often by copying snippets from the Kubernetes documentation. There are a number of such cases, e.g., adding Pod Volumes, i.e., spec.volumes.

Pods can be deleted using the configuration file as we described earlier. However it is useful to use the --force=true option to avoid having to wait up to ten seconds for a graceful deletion (during the exam, every second counts). For example:

k delete -f my-resource.yaml --force=true

In general, Kubernetes does not allow one to update a Pod. There are some exceptions; as one can see from a representative error if one tries to incorrectly update a Pod:

The Pod "nginx" is invalid: spec: Forbidden: pod updates may not change fields other than `spec.containers[*].image`, `spec.initContainers[*].image`, `spec.activeDeadlineSeconds` or `spec.tolerations` (only additions to existing tolerations)

To keep things simple, however, it is easier to simply delete the Pod and recreate it if it needs updating.

Deployments

Unlike many of the other create commands (and the run command for creating Pods), there are few options available when creating Deployments as you can see by running:

k create deployment --help

The bottom of the help output provides the required and pretty much only common option:

kubectl create deployment NAME --image=image [--dry-run=server|client|none]
[options]

As such, you will likely have to modify the generated Deployment configuration file before applying it. Here are some common scenarios:

• Replicas, i.e., spec.replicas
• Pod template, i.e., spec.template

One trick for modifying the Deployment’s template is to also generate a Pod configuration file (see Pods above) and reading it into the Deployment configuration file.

note: You will need to update the imported Pod’s labels to match the Deployment’s selector. You will also need to delete its name.

Services

While one can create Services using the create command, it is much quicker to use the expose command in conjunction with running Pod or Deployment. This is because this command will automatically create the Service label selector based on the Pod or Deployment.

note: As you may recall, we can also create a Service exposing a Pod at Pod creation time with the --expose=true option.

Running the following command gives us the command options:

k expose --help

The bottom of the help output provides the required and common optional options (the top provides the complete list of options):

kubectl expose (-f FILENAME | TYPE NAME) [--port=port]
[--protocol=TCP|UDP|SCTP] [--target-port=number-or-name] [--name=name]
[--external-ip=external-ip-of-service] [--type=type] [options]

For example, to create a NodePort service called my-service on port 80 targeting a deployment called nginx running an nginx container (exposes port 80) we can execute:

k expose deployment nginx --port=80 --target=port=80 --name=my-service --type=NodePort

Wrap-Up

Hopefully you found this helpful. Good luck on your exams.