- Configuration in Kubernetes
- Launching a container using a configuration file
- Validating configuration
- Environment variables and variable expansion
- Viewing pod status
- Viewing pod output
- Deleting pods
- What’s next?
Configuration in Kubernetes
In addition to the imperative-style commands, such as kubectl run and kubectl expose, described elsewhere, Kubernetes supports declarative configuration. Oftentimes, configuration files are preferable to imperative commands, since they can be checked into version control and changes to the files can be code reviewed, which is especially important for more complex configurations, producing a more robust, reliable and archival system.
In the declarative style, all configuration is stored in YAML or JSON configuration files using Kubernetes’s API resource schemas as the configuration schemas. kubectl can create, update, delete, and get API resources. The apiVersion (currently v1?), resource kind, and resource name are used by kubectl to construct the appropriate API path to invoke for the specified operation.
Launching a container using a configuration file
Kubernetes executes containers in Pods. A pod containing a simple Hello World container can be specified in YAML as follows:
apiVersion: v1
kind: Pod
metadata:
name: hello-world
spec: # specification of the pod's contents
restartPolicy: Never
containers:
- name: hello
image: "ubuntu:14.04"
command: ["/bin/echo", "hello", "world"]
The value of metadata.name, hello-world, will be the name of the pod resource created, and must be unique within the cluster, whereas containers[0].name is just a nickname for the container within that pod. image is the name of the Docker image, which Kubernetes expects to be able to pull from a registry, the Docker Hub by default.
restartPolicy: Never indicates that we just want to run the container once and then terminate the pod.
The command overrides the Docker container’s Entrypoint. Command arguments (corresponding to Docker’s Cmd) may be specified using args, as follows:
command: ["/bin/echo"]
args: ["hello","world"]
This pod can be created using the create command:
$ kubectl create -f ./hello-world.yaml
pods/hello-world
kubectl prints the resource type and name of the resource created when successful.
Validating configuration
We enable validation by default in kubectl since v1.1.
Let’s say you specified entrypoint instead of command. You’d see output as follows:
error validating "./hello-world.yaml": error validating data: found invalid field Entrypoint for v1.Container; if you choose to ignore these errors, turn validation off with --validate=false
Using kubectl create --validate=false to turn validation off, it creates the resource anyway, unless a required field is absent or a field value is invalid. Unknown API fields are ignored, so be careful. This pod was created, but with no command, which is an optional field, since the image may specify an Entrypoint.
View the Pod API
object
to see the list of valid fields.
Environment variables and variable expansion
Kubernetes does not automatically run commands in a shell (not all images contain shells). If you would like to run your command in a shell, such as to expand environment variables (specified using env), you could do the following:
apiVersion: v1
kind: Pod
metadata:
name: hello-world
spec: # specification of the pod’s contents
restartPolicy: Never
containers:
- name: hello
image: "ubuntu:14.04"
env:
- name: MESSAGE
value: "hello world"
command: ["/bin/sh","-c"]
args: ["/bin/echo \"${MESSAGE}\""]
However, a shell isn’t necessary just to expand environment variables. Kubernetes will do it for you if you use $(ENVVAR) syntax:
command: ["/bin/echo"]
args: ["$(MESSAGE)"]
Viewing pod status
You can see the pod you created (actually all of your cluster’s pods) using the get command.
If you’re quick, it will look as follows:
$ kubectl get pods
NAME READY STATUS RESTARTS AGE
hello-world 0/1 Pending 0 0s
Initially, a newly created pod is unscheduled – no node has been selected to run it. Scheduling happens after creation, but is fast, so you normally shouldn’t see pods in an unscheduled state unless there’s a problem.
After the pod has been scheduled, the image may need to be pulled to the node on which it was scheduled, if it hadn’t been pulled already. After a few seconds, you should see the container running:
$ kubectl get pods
NAME READY STATUS RESTARTS AGE
hello-world 1/1 Running 0 5s
The READY column shows how many containers in the pod are running.
Almost immediately after it starts running, this command will terminate. kubectl shows that the container is no longer running and displays the exit status:
$ kubectl get pods --show-all
NAME READY STATUS RESTARTS AGE
hello-world 0/1 ExitCode:0 0 15s
Viewing pod output
You probably want to see the output of the command you ran. As with docker logs, kubectl logs will show you the output:
$ kubectl logs hello-world
hello world
Deleting pods
When you’re done looking at the output, you should delete the pod:
$ kubectl delete pod hello-world
pods/hello-world
As with create, kubectl prints the resource type and name of the resource deleted when successful.
You can also use the resource/name format to specify the pod:
$ kubectl delete pods/hello-world
pods/hello-world
Terminated pods aren’t currently automatically deleted, so that you can observe their final status, so be sure to clean up your dead pods.
On the other hand, containers and their logs are eventually deleted automatically in order to free up disk space on the nodes.
What’s next?
Learn about deploying continuously running applications.