DevOps Classroomnotes 29/Apr/2023

Kubernetes as a Service (Managed K8s)

• Every Cloud provider offers k8s as a service
  • Azure = AKS
  • AWS = EKS
  • GCP = GKE
• K8s as a service basically means the master nodes will be managed by cloud provider
  • Typical k8s cluster
    
  • K8s as a Service
    
• Advantages:
  • less administration
  • nodes can be scaled
  • inbuilt support for cloud integrations
• In this course we will be using
  • AKS
  • EKS

Setting up basic k8s cluster in Azure (AKS)

• In this setups we configure kubectl on
  • dev systems
  • build servers
    
• install kubectl Refer Here or we can use azure cli to set it up
• Install azure cli: Refer Here
• Execute az login
• Refer Here for azure aks
• Create a resource group and continue according to docs Refer Here

Exposing Applications running in cluster to externally as well as internally when scaled

• Every pod gets a unique ip and name.
• Connecting from one pod to other on the basis of name/ip might not be a good idea as pods are controlled by replicasets or other controllers
• K8s has a service which helps us in connecting to pods with similar behaviour but by using labels.
• Each service gets a ip address and this is virtual ip which helps in forwarding traffic to one of the pod based on labels. This ip is called as cluster ip
• Services can be exposed to external world
  
• Service is similar to layer 4 load balancer
• Refer Here for official docs

Internal Communication using k8s service

• Consider the following
  • We have an alpine pod which needs to consume nginx
  • but nginx is a replica set and there can be n replicas
• Lets create a nginx-rs
  
  
• Create an alpine pod and login into that
  
• ping nginx-svc by its ip address and try accessing the web page using curl
  
• access nginx-svc by using name
  
• now do nslookup based on name
  
• Look into environment variables in alpine pod (Alpine was created post nginx service creation)
  
• Look into environment variables in nginx pods (These were created prior to nginx service)
  
• Refer Here for internal communication manifests

External Communication using k8s service

• Some user external to k8s cluster wants to access nginx
  
• Kubernetes has the service publishing types Refer Here
  • Cluster ip: internal communication
  • Node Port: k8s will expose the application on a port on every node in k8s cluster.
    
  • LoadBalancer: This is generally used with managed k8s clusters
    
  • ExternalName: Creates a CNAME record that can be used in your DNS Servers
• We have created a manifest with loadBalancer
  
• Refer Here for the spec