Kubernetes Overview

Container + Orchestration

Docker is not meant to run different Operating Systems/Kernels on the same hardware. The main purpose of docker, is to containerize applications and ship and run them.

Container orchestration: the process of automatically deploying and managing containers (usually based on the load needs).

Kubernetes Architecture

• Worker Node (aka minions): the worker machines where the containers are running. If node fails, the app is down (that's why we must have multiple nodes).

• Cluster: a set of nodes. If a node fails, the app is still available through the other nodes

• Master Node: watches over the nodes and is responsible for the containers orchestration.

• Components:

  • API Server: frontend for kubernetes
  • etcd: distributed reliable key-value store, to store data used to manage the cluster. Stores logs to ensure no conflicts between the masters.
  • Scheduler: distribute work and containers and assign them to nodes.
  • Controller: the brain behind orchestration, responsible to react when containers/nodes/endpoints go down.
  • Container Runtime: the underlying software used to run containers (currently it's docker)
  • kubelet: the agent running in each node in the cluster, responsible to assure the containers are running as expected. It interacts with the Master node via API Server.

| -------------- |       | ----------------- |
| kube-apiserver | <---> | kubelet           |
| etcd           |       |                   |
| controller     |       |                   |
| scheduler      |       |                   |
|                |       | container runtime |
| -------------- |       | ----------------- |
| Master Node    |       | Worker Node       |
| -------------- |       | ----------------- |

The kubectl is the CLI used to interact with kubernetes.

Examples:

• kubectl run hello-minikube - deploy an application
• kubectl cluster-info - view info about the cluster
• kubectl get nodes - list all the nodes that are part of the cluster

anki cards

TARGET DECK: kubernetes

The 6 main kubernetes components #flashcard

• Master Node:
  • API Server
  • etcd
  • controller
  • scheduler
• Worker Node
  • kubelet
  • container runtime

k8s component: API Server #flashcard

• handles communication with the worker node.
• runs on the Master Node

k8s component: etcd #flashcard

• key-value store with data used to manage the cluster
• stores logs to prevent conflicts between the nodes
• runs on the master node

k8s component: Scheduler #flashcard

• distributes work and containers, assigning them to nodes
• runs on the master node

k8s component: Controller #flashcard

• the brain behind orchestration
• responsible to react when containers/nodes/endpoints go down
• runs on the master node

k8s component: Container Runtime #flashcard

• underlying software used to run containers
• runs on the worker node

k8s component: kubelet #flashcard

• the agent running in each node
• responsible to assure containers runs as expected
• runs on the worker node
• interacts with master node via API Server