2. K8s Architecture

Node Processes

• Each Node has multiple Pods in it

• Worker Nodes do the actual work

• 3 processes must be installed on every Node

  1. Container Runtime

     • Docker or other technology

  2. Kublet

     • Interacts with both container and node

     • It starts the pod with a container inside

  3. Kube proxy

     • Forwards the requests between nodes

---

Master Processes

• 4 processes must be installed on every master node which controls cluster state and worker nodes

  1. API Server

     • Could be UI, API, CLI, Kubectl)

     • Acts as cluster gateway

     • Acts as gatekeeper for authentication

  2. Scheduler

     • API Server after validation of request hands over the request to scheduler

     • It just decides on which node new pod should be scheduled based on current resources of nodes

     • Kublet is the actual component which does the scheduling after getting request form scheduler

  3. Controller Manager

     • Detec cluster state changes like dying of podes

     • It makes a request to scheduler to restart the pod

  4. etcd

     • It is the cluster brain

     • cluster changes get stored in the key value store

     • application data is not stored in etcd

• On multiple master nodes

  • API server is load balanced

  • etcd forms distributed storage across all master nodes