Running Multiple Containers in a Single Pod

Every pod so far had exactly one container. That is the normal case, and it should stay your default. But a pod is not "a container with extra steps". It is a small environment that can hold several containers, which then share the same network address, can share files, and are always scheduled together on the same node.

This lesson builds a pod with two containers so you can see how that works and how kubectl handles it. The next lesson turns it into the most common real use: the sidecar.

What containers in one pod share

Use several containers in one pod only when they are tightly coupled: one cannot do its job without the other being right next to it. A Spring Boot app and its MySQL database are not tightly coupled; they belong in separate pods (and separate Deployments) so they can be scaled, updated and restarted independently.

Step 1: Write the manifest

The demo runs two well known images:

• nginx: a popular web server and reverse proxy, listening on port 80
• busybox: a tiny Linux image with many common commands (often called the Swiss Army knife of Linux). It runs a small shell loop that prints a message every 10 seconds

vi multi-container.yml

apiVersion: v1
kind: Pod
metadata:
  name: telusko-multicontainer-pod
  labels:
    app: multicontainer-pod
spec:
  containers:
    - name: nginx-container
      image: nginx
      ports:
        - containerPort: 80

    - name: busy-box-container
      image: busybox
      command:
        - sh
        - -c
        - while true; do echo 'Hello From Telusko K8S learning'; sleep 10; done

The blank line between the two containers is only for readability; YAML ignores it.

command here is the Kubernetes equivalent of Docker's ENTRYPOINT, and args is the equivalent of CMD. If you have done the Docker course, CMD vs ENTRYPOINT explains the idea.

Step 2: Apply it, and fix the field name

kubectl apply -f multi-container.yml

In the lecture the first attempt failed with Pod in version "v1" cannot be handled as a Pod, because the field was written as commands. The correct field name is command (singular). As with every unknown field error, compare the spelling with kubectl explain pod.spec.containers.

After fixing:

kubectl get pods

NAME                         READY   STATUS    RESTARTS   AGE
telusko-multicontainer-pod   2/2     Running   0          4m31s

READY 2/2 means both containers in the pod are ready. You may first see 0/2 ContainerCreating while the images are pulled. If you ever see 1/2, one container is failing; kubectl describe pod shows which one.

Step 3: Logs from a specific container

With more than one container, kubectl needs to know which one you mean. Use -c:

kubectl logs telusko-multicontainer-pod -c busy-box-container

Hello From Telusko K8S learning
Hello From Telusko K8S learning
Hello From Telusko K8S learning
...

A new line every 10 seconds. And the nginx startup log:

kubectl logs telusko-multicontainer-pod -c nginx-container

If you get the container name wrong, kubectl tells you the valid names. You can also find them with:

kubectl describe pod telusko-multicontainer-pod          # look under Containers:
kubectl get pod telusko-multicontainer-pod -o jsonpath='{.spec.containers[*].name}'

Useful log options:

Step 4: Go inside a container, and prove they share the network

Just as docker exec opens a shell in a container, kubectl exec does it in a pod. Again, -c picks the container:

kubectl exec -it telusko-multicontainer-pod -c busy-box-container -- sh

Inside the busybox shell, call nginx on localhost:

wget -qO- http://localhost:80 | head -5
exit

You get the nginx welcome page HTML. busybox has no web server of its own; it reached nginx through localhost because both containers share the pod's network namespace. Two separate pods could not do that.

Without -c, kubectl exec and kubectl logs use the first container in the list (or the one marked as default with the kubectl.kubernetes.io/default-container annotation) and print a note telling you so.

When multi-container pods make sense

The sidecar is the one you will meet most often, and it now has first-class support in Kubernetes. That is the next lesson.

Clean up

kubectl delete -f multi-container.yml

Deleting the pod removes both containers. If you are finished for the day: eksctl delete cluster --name telusko-learning --region ap-south-1 --wait.

Try It Yourself

1. Create the pod and confirm READY 2/2.
2. Follow the busybox logs with -f for 30 seconds, then stop with Ctrl+C.
3. Exec into busybox and fetch the nginx page from localhost. Then try to start a second web server on port 80 inside busybox (httpd -f -p 80) and read the error.
4. Remove the while loop from the busybox command so it only runs echo hello. Apply (delete the pod first) and watch RESTARTS grow and the status become CrashLoopBackOff. Why does this happen? (Covered in Kubernetes Pod Statuses.)

Official references

• Kubernetes: Pods (how pods manage multiple containers)
• Kubernetes: Communicate Between Containers in the Same Pod Using a Shared Volume
• kubectl reference: logs

Written By: Shramik Masti