Abdullah Şamil Güser

Nodes

Master Node

• Kubernetes services and controller are located on the control plane
• They are also called the master components

• etcd is a key-value data store that stores the state of the cluster
• The API Server (kube-apiserver) is the only component that interacts with etcd

kube-apiserver

• It exposes a REST interface and client tools like the kubernetes CLI communicates through that API.
• Save state to the datastore (etcd)
• All clients interact with it, never directly to the datastore

etcd

• Acts as a datastore for storing cluster state
• It’s not a database but a key-value store
• It’s the single source of truth inside the kube cluster

kube-control-manager

• The controller of controllers!
• Runs other kube controllers
  • Node controller
  • Replication controller
  • Endpoints controller
  • Service account and Token controller

cloud-control-manager

Interacts with the cloud providers controllers

• Node : For checking the cloud provider to determine if a node has been deleted in the cloud after it stops responding
• Route : For setting up routes in the underlying cloud infrastructure
• Service : For creating, updating and deleting cloud provider load balancers
• Volume : For creating, attaching, and mounting volumes, and interacting with the cloud provider to orchestrate volumes

kube-scheduler

• Watches newly created pods that have no node assigned, and selects a node for them to run on
• Factors taken into account for scheduling decisions include
  • Individual and collective resource requirements
  • Hardware/software/policy constraints
  • Affinity and anti-affinity specifications
  • Data locality

Addons

You can install various add-ons on the masternode these add-ons provide additional functionalities in your kubernetes cluster.

• DNS
• Web UI (dashboard)
• Cluster-level logging
• Container resource monitoring

Worker Nodes

• A node is a physical or virtual machine
• A group of nodes forms a cluster
• There’s a special node called the masternode, it’s sometimes called the control plane where the kubernetes services are installed
• The nodes running the containers are called the worker nodes

• When a worker node is added to the cluster some kubernetes services are installed automatically
• These are services necessary to run pods and they are managed by the master components on the masternode

kubelet

• Manage the pods lifecycle
• Ensures that the containers described in the Pod specs are running and healthy

kube-proxy

• A network proxy
• Manages network rules on nodes

Container runtime

• K8s supports several container runtimes
• Must implement the Kubernetes Container Runtime Interface
  • Moby
  • Containerd
  • Cri-O
  • Rkt
  • Kata
  • Virtlet

Note : Container runtime - K8s V1.19+

• Docker images run as is. It’s business as usual!
• What’s changed is what you can do inside the cluster
  • You can no longer access the Docker engine inside the cluster
  • Docker commands won’t run if you ssh into a node
  • Use crictl instead

Nodes pool

• A node pool is a group virtual machines, all with the same size
• A cluster can have multiple node pools
  • These pools can host different sizes of VMs
  • Each pool can be autoscaled independently from the other pools
• Docker Desktop is limited to 1 node

CLI Cheat Sheet

# Get a list of all the installed nodes. Using Docker Desktop, there should be only one.
kubectl get nodes

# Get some info about the node.
kubectl describe node