DevOps Classroomnotes 10/Sep/2023

Controllers

• Pod tries to keep containers running, but for us we need to keep Pods running according to some state, Lets understand first two categories
  • Replicas:
    • Here we have two resources ReplicationController, ReplicaSet
    • Here our desired state (spec) will be
      • number of replicas
      • pod spec
      • label selector
    • These objects try maintain the desired
  • Jobs:
    • These will run the Pods which have finite execution time period
    • Here we have two resources
      • Job
      • CronJob

ReplicaSet

• This Resource is responsible for maintaing the desired state of number of replicase of pod
  
• Refer Here for official docs

Activity:

• Create a replica set with 3 replicas of jenkins/jenkins
• Refer Here for the manifests
  
• Lets try set based selectors Refer Here for the changes
• Exercise: Write a Replication controller for creating 3 httpd pods

Jobs and Cron Jobs

• Refer Here for Offical jobs documentation and Refer Here for Official CronJob Documentation
• Lets write a Cron Job which runs alpine pod with some script for every 5 mins
• Refer Here for the changes
  

Namespace

• Namespace in k8s is a logical space or logical cluster in which resources will be created
• Any resource which has a value of Namespace = true belongs in a namespace and with Namespace=false is shared across namespaces.
• Refer Here for official docs

Service in K8s

• Every Pod when created gives a unique ip address and Name
• When Pods are scaled
• Lets create a replicaset with 3 nginx pods with label app:nginx

---
apiVersion: apps/v1
kind: ReplicaSet
metadata:
  name: nginx-rs
  labels:
    app: nginx
    purpose: svcdemo
spec:
  replicas: 3
  minReadySeconds: 1
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
        - name: nginxc
          image: nginx:1.25
          ports:
            - containerPort: 80
              protocol: TCP

• We need to access nginx with in cluster, but pod ips are not reliable, so as shown in the below image, lets create a k8s service resource which gives a consistent ip and name to access all the pods matching labels (equality based selection)
  
• service in k8s is of 4 types
  • Cluster IP: This gives an ipaddress or name which can be accessed with in cluster
  • None: TBD
  • External: TBD
  • LoadBalancer: TBD
• Refer Here for the changeset
• Lets create nginx rs and and service and view endpoints. As discussed in the class when pods get update the endpoints are reflected. Service will forward the requests to endpoints
• We will try correcting the access issue and external access in the next session