Debug School: Sarmila P

Day 3 - Sarmila

Sarmila P — Thu, 19 Oct 2023 04:00:07 +0000

Daemon Sets

Min/Max one pod per node eg: kubectl canal and proxy

Jobs

To compute a work one time

Cron Jobs

To execute regular scheduled actions

Config Maps

Save configs, keys in a config file used by application instead of hardcoding in the application properties

Services

Network Load Balancer to deal which pods to connect with

Service Types -

1.Cluster IP - Accessible only within the cluster
2.Node Port - Can be accessed from outside the cluster
3.Load Balancer - Balances the input request load (used aws load balancer - this needs to be linked with DNS)
*4.External *

Ingress

*Application Load Balancer *
To reduce the usage of Load balancers, routing rules can be specified to enable traffic, routing with security
2 main components: Rule, Controller (exposed to Load Balancer)
3 Types:
Name Based - Based on Domain Names
Path Based - Based on application path
TLS Based

Day 02 - Sarmila

Sarmila P — Tue, 17 Oct 2023 12:09:41 +0000

Namespaces

Namespace is a logical separation of resources within the cluster
Namespaces was introduced to overcome 2 problems, 1.To avoid problems while using the shared resources by diff applications 2.To restrict access over resources
Pods will come under namespaces
When pods are under a namespace ,namespace will be true
When pods are under a worker node, namespace will be false
Any namespace will be classified under one of the below category,
System - Created by Kubernetes
Public - Common resources
Default - When no namespace is provided
Admin - For admin roles
When namespace is deleted, all respective pods will get deleted

Accessing resources:

Create- create/ create -f yaml
Read- get/ get -f yaml
Update- edit / apply -f yaml
Delete- delete / delete -f yaml

Debugging/Troubleshooting strategies

Logs
attach
Exec
Port forward
copy
auth

Deployment in Kubernetes:
1.Replication - controls number of replications without writing many yaml files
2.Controller - controls replication
3.Versioning - controls the version
4.Rollout - To roll out the changes(Recreate - Delete and create all the container and Rolling update)
5.RollBack - To roll back to the older version

What is Pod - Sarmila

Sarmila P — Mon, 16 Oct 2023 11:50:07 +0000

Pod is an atomic unit of scheduling work
One pod can have multiple containers running
Pod will run as long as the containers are running (lifecycle includes pending, running, succeeded/Failed)
Best approach for pods is to have one container per pod
When multiple containers are running in the same file, resources will be shared
Every container inside a pod will be having the same ip address but port number will be unique
No pod will be ever redeployed
Containers inside pods can communicate within the pods using localhost

Day 01 - By Sarmila

Sarmila P — Mon, 16 Oct 2023 07:13:09 +0000

What is Kubernetes?

Kubernetes is an container orchestration system which helps in High availability, Scalability and Auto healing of the containers.
It was developed by Google and maintained as open source by CNCF.
It plays a huge role in maintaining containers in large scale.

Why Do we need Kubernetes? Explain in 10 lines

Kubernetes will be consider data center as a one machine
It makes deployment easy
It provides Self Healing of the containers
It also ensures maintaining high scalability and availability of containers in large scale
It can also provide features like storage and network orchestration
Helps in effective service discovery and load balancing
It provides security

How Kubernetes Works?

Kubernetes will be consider data center as a one machine (cluster) and it works based on master-worker model

Kubernetes Architecture. Explain each component with 1 line.

Master Node :

Api Server: Will receive json inputs and provides response (Eg: any requests, state of pods)
Cluster Store: Stores each changes (eg: cluster state, configs) in etcd database
Controller Manager: Watches for the changes in loop
Kube Scheduler: Will schedule work to worker pod

Worker Node:

Kublet: Will register node with cluster, keep watching api server for work and instantiates the node and reports back to master
Container Engine: Where our actual containerization works
Kube-proxy: It will act as a proxy server inside cluster