Skip to main content

Pods

Updated Apr 07, 2022 ·

Overview​

A Pod is the smallest unit in Kubernetes. It can contain one or more containers that always run together. If you run single containers, think of a Pod as just one container.

Kubernetes manages Pods by:

  • Deciding when and where to run them
  • Handling traffic routing
  • Scaling Pods based on metrics

Additionally, Kubernetes:

  • Starts Pods based on resource needs
  • Restarts Pods if they or their hosts fail
  • Assigns an IP address and DNS name to each Pod

Pod Commands​

To view Pods in the default namespace:

kubectl get pods

To view Pods in all namespaces:

kubectl get pods -A

To view Pods in a specific namespace (e.g., kube-system):

kubectl get pods -n kube-system

To get details about a Pod:

kubectl describe <pod-name>

To view logs of a Pod:

kubectl logs <pod-name>

To delete a Pod:

kubectl delete pod <pod-name>

To delete a Pod created from a manifest:

kubectl delete -f <pod-manifest-file.yml>

IPs and Ports​

Each Pod is assigned a single IP address, regardless of how many containers are in the Pod. To allow access to a Pod, we must specify which port to publish.

  • Kubernetes cannot update ports on a running Pod.
  • If you need to change the port, delete the Pod.
  • Then update the port in the manifest, and recreate the Pod.
  • By default, Kubernetes uses TCP as the protocol.

For more details, check out Pods in Kubernetes.

Pod Declaration​

A Pod declaration defines all the properties of a Pod. Some common attributes include:

  • The container image to use
  • The container ports to publish
  • The restart policy for handling container failures
  • CPU and memory resources

Here’s an example of a simple Pod manifest that launches an NGINX server:

apiVersion: v1
kind: Pod
metadata:
  name: app
  labels:
  - app: webserver
spec:
  containers:
  - name: app-nginx
    image: nginx:latest
    ports:
    - containerPort: 80

Multi-container Pod​

In a multi-container Pod, each container should run a process that stays alive for the Pod's entire lifecycle. For example, if a Pod has:

  • A web application
  • A logging agent

Both containers need to run together. If either container fails, the Pod restarts.

However, some tasks may only need to run once, such as pulling code from a repository before the main application starts. These tasks can be handled using initContainers.

For more information, please see Probes and InitContainers.