Note - Kubernetes in Action

I took notes during reading “Kubernetes in Action”.

It can be found on Amazon.

Part 1: Overview

Part 2: Core concepts

Chapter 3:

• Po/pods/pod is similar. For example:

  kubectl get po
  kubectl get pod
  kubectl get pods

  It’s the same with service/services.

Introducing pods

A pod is a Kubernetes abstraction that represents a group of one or more application containers (such as Docker or rkt), and some shared resources for those containers. You can have several related docker containers run on a pod.

A Pod always runs on a Node. A Node is a worker machine in Kubernetes and may be either a virtual or a physical machine, depending on the cluster.

Create pod with yaml/json

• Create pod from yaml file

  kubectl create -f [yaml/json file]

• Get a list of pods

  kubectl get po

• Get describes of pods

  kubectl get po [pod_name] -o yaml

• Get log of container (-c is optional, when you want to view the log of a specific container)

  kubectl logs [pod_name] -c [container_name]

• Forward port 80 from pod to port 8888 local (effective way to test pod)

  kubectl port-forward [pod_name] 8888:80

Organizing resources with labels

• Label is an arbitrary key-value attached to a resources. It can be defined under metadata.labels tag.
• Each pod is recommended to be labeled with 2 labels: app (which microservice it belong to), env (which environment: staging, dev,..)
• Get the list of pods with label

  kubectl get po --show-labels

• Get the pods with specific label by column (in this case is app & rel)

  kubectl get po -L app,rel

• Get the pods with label selector

  ///Get the pods have env label
  kubectl get po -l env
  ///Get the pods have specific env label 
  kubectl get po -l env=staging

• Change label of pods

  kubectl label po [pod_name] app=frontend
  kubectl label po [pod_name] --overwrite app=backend

Constrain pod scheduling with labels and selectors

All the pods is scheduled randomly across the worker nodes by default.

• Add the label to specific nodes

  //get list of nodes
  kubectl get nodes --show-labels
  //add the label
  kubectl label node [node_name] gpu=true

• Schedule pods to specific nodes which have gpu label is true by editing in yaml/json file. Should do like this instead of scheduling to a specific node because the node can be offline.

  ...
  spec:
    nodeSelector:
      gpu: "true"
  ...

Annotation

Mostly the same with label. But annotation is used for large blobs of data (256KB in total); mostly use of annotations is adding descriptions for each pod.

kubectl annotate pod [pod_name] mycompany.com/someannotation="this pos is used for..."

Namespaces