Workload Resources (ReplicaSets)

• Pods are not self healling.
• Generally we might require to run multiple instances of our application for various reasons
  • Redundancy: Fault tolerance
  • Scale: Higher request processing capacity
  • Sharding: Can handle different parts of computation in parallel.
• The central concept of workload resources is reconcilation loop i.e. notion of desired stated vs current state.
• Replicaset is also an k8s object which would have a unique name. This describes number of Pods and Pod Specification.
• K8s Will run a reconcilation loop to maintain the desired number of Pods
  

Labels

• Lets see the first purpose of labels
• Lets create pods imperatively in k8s cluster

kubectl run qtapp-dev --image=httpd --labels="ver=3,app=qtapp,env=dev"
kubectl run qtapp-qa --image=httpd --labels="ver=2.5,app=qtapp,env=qa"
kubectl run qtapp-staging --image=httpd --labels="ver=2.4,app=qtapp,env=staging"
kubectl run qtapp-prod --image=httpd --labels="ver=2.0,app=qtapp,env=prod"

* The labels for the pods are as shown below


* Label Selectors: Using label selectors we can query the pods or k8s objects

• Label Selector Operators

Replicasets contd

• Lets try to create a manifest for running 2 nginx pods in the cluster

---
apiVersion: apps/v1
kind: ReplicaSet
metadata:
  name: nginx-rs
  labels:
    app: nginx
    version: "1.23.0"
spec:
  minReadySeconds: 3
  replicas: 2
  selector:
    matchLabels:
      app: nginx
      version: "1.23.0"
  template:
    metadata:
      labels:
        app: nginx
        version: "1.23.0"
        size: large
    spec:
      containers:
        - name: nginx
          image: nginx:1.23.0
          ports:
            - containerPort: 80

• Refer Here for the changes
  
• k8s will try to maintain the desired state of replicas
  
• Inspecting replicaset
  
• Scaling replicasets:
  • Can be done from commandline but changing the spec is the best practice
  • In the future classes we would set automatic scaling based on metrics
    
• Exercise: Create a replicaset manifest for shaikkhajaibrahim/qttestapp:1.1 with 3 replicas.
  • solution: Refer Here for the changes
  • Fixed issue with version Refer Here
    

DaemonSets

• A Daemon set ensures that a copy of pod is running on a set of nodes in a k8s cluster
• Daemon set has pod specification and label selector
• Scenario 1: We want to run fluentd log collector on all the nodes in the k8s cluster
• Refer Here for the daemon set added
  
  
• Scenario 2: Running a daemon set on specific nodes:
  • Lets apply a label to one node with ssd=true & ssd=false on other
    
    
  • Refer Here for the changes
    
• Updating the daemonset we will discuss after discussing deployments