Container Orchestration

• Orchestration in Compute: This is about
  • automated configuration
  • management
  • co-ordination of systems, applications and services
• Container Orchestration is
  • automation of operational effort required to run containerized workloads and services
  • managing
    • container lifecycle
      • provisioning
      • deployment
      • scaling
    • networking
    • storage

Cluster

• Refer Here for cluster definitions
• Refer Here for cluster architectures
• Most popular architecture
  

Rest API

• This is a standard way of exposing functionality over http(s)
• Refer Here for what rest api
• Verbs:
  • GET
  • PUT
  • POST
  • DELETE

Kubernetes (k8s)

• Basic kubernetes cluster
  
• Kubernetes cluster
  
• Kubernetes has components categorized based on where they are executing
  • Control Plane
    • kube-api Server
    • kube-scheduler
    • kube-controller-manager
    • etcd store
  • Worker Node
    • kubelet
    • container engine
    • kube-proxy
• Any system which wants to interact to kubernetes should have kubectl
  

kube-api server component

• This is most important component of the k8s control plane
• kube-api server exposes REST API which enables clients to send HTTP requests to kube-api server
• kube-api server responds over http requests and writes the resource information to etcd store
• kube-api server exposes the k8s objects
• kube-api server is responsible for all the communication
• The api server is over https and requires authentication

etcd store

• This is memory of k8s cluster Refer Here
• etcd cluster can scale across multiple nodes unlike traditional databases

kube-scheduler

• This polls the kube-api server for any new objects in etcd store which are not scheduler
• For any workloads unscheduled k8s scheduler will find a suitable worker node and then creates the workload

kube-controller-manager

• This ensures the desired state is met.
• This is combination of multiple controller
  • NodeController
  • Replication Controller
  • Namespace Controller
  • EndpointController
  • ServiceAccountController

kubelet

• This is an agent on the worker node
• This reacts to requests/orders from control plane components and speaks with container runtime and gets the work done
• If it fails responds back to control plane with status

Container Runtime

• this is container engine
• this could be docker or any other CRI compatible runtime

kube-proxy

• this is responsible for networking
• this implementations will be based on CNI

Getting familiar with tmux

• Install tmux on linux