Kubernetes Core Concepts

Cluster Architecture

Master Node

• Manage
• Plan
• Schedule
• Monitor Nodes

Worker Nodes

• Host Application as Containers

Kubernetes Architecture

Master Node

• kube-apiserver
• Controller-Manager
• Node-Manager
• Replication-Manager

Worker Nodes

• kubelet is manager of a node
• Kube Proxy

kube-api server Execute Process

1. Authenticate User
2. Validate Request
3. Retrieve data
4. Update ETCD
5. Scheduler
6. Kubelet

api-server options

1

cat /etc/kubernetes/manifests/kube-apiserver.yaml

1

cat /etc/systemd/system/kube-apiserver.service

1

ps -aux | grep kube-apiserver

ETCD

ETCD is a distributed reliable key-value store that is Simple, Secure & Fast

Install ETCD

1. Download Check official ETCD. The direct link is here.

2. Run

1

./etcd

Operate ETCD

1

./etcdctl set key1 value1

1
2

./etcdctl get key1
# value1

ETCD in Kubernetes

ETCD Cluster

• Nodes
• PODs
• Configs
• Secrets
• Accounts
• Roles
• Bindings
• Others

Setup

Manual

1

wget DOWNLOAD_URL

1

# etcd.service

kubeadm

1

kubectl get pods -n kube-system

1

kebuectl exec etcd-master -n kube-system etcdctl get / --prefix -keys-only

ETCD Commands

To Set Version

1

export ETCDCTL_API=3

ETCD Version 2

1
2
3
4
5

etcdctl backup
etcdctl cluster-health
etcdctl mk
etcdctl mkdir
etcdctl set

ETCD Version 3

1
2
3
4

etcdctl snapshot save 
etcdctl endpoint health
etcdctl get
etcdctl put

kube-controller-manager

List

• Node-Controller
• Replication-Controller
• ReplicaSet
• Service-Account-Controller
• CronJob
• Job-Controller
• Deployment-Controller
• Namespace-Controller
• Endpoint-Controller
• PV-Protection-Controller
• Stateful-Set

Main Functions

• Watch Status
• Remediate Situation

Option Examples

Node Monitor Period = 5s
Node Monitor Grace Period = 40s
POD Eviction Timeout = 5m

kube-controller-manager Options

1

cat /etc/kubernetes/manifests/kube-controller-manager.yaml

1

cat /etc/systemd/system/kube-controller-manager.service

1

ps -aux | grep kube-controller-manager

Kube Scheduler

kube-scheduler just decide which POD goes where.

Rule

Required CPU: 10

Node: 4

Node: 4

Node: 12

Node: 16

1. Filter Nodes -> Except 4 Nodes
2. Rank Nodes -> Picked Node: 16

kube-scheduler Options

1

cat /etc/kubernetes/manifests/kube-scheduler.yaml

1

cat /etc/systemd/system/kube-scheduler.service

1

ps -aux | grep kube-scheduler

Kubelet

Functions

• Register Node
• Create PODs
• Monitor Node and PODs

Options

1

ps -aux | grep kubelet

Kube Proxy

It’s to configure IP from the service name.