Skip to content

Multi-Tenancy In EKS Cluster Using Vcluster

Posted on:October 9, 2023Β atΒ 03:57 PM

image

Abstract

  • Why should we consider using Vcluster? Our requirement is to establish multiple environments for developers to facilitate development, testing, as well as regression and performance tests.
  • The crucial aspect is ensuring that these environments closely mimic the structure of our staging and production environments, which are based on Kubernetes. Instead of relying on Kubernetes namespaces to create these environments, We opt to offer developers a solution that provides them with an environment that closely resembles a real Kubernetes cluster. This is where Vcluster comes into play.
  • Watch Demo: https://www.youtube.com/watch?v=vWNkGyLajJE
  • Source code: https://github.com/vumdao/multi-tenancy-using-vcluster-in-eks/tree/master

Table Of Contents

Open Table Of Contents

πŸš€ vcluster overview

πŸš€ Solution overview

πŸš€ Bootstrap EKS cluster using CDK EKS Blueprints

  • The bootstrap provisions EKS cluster with required AddOns using CDK EKS blueprints

      new VpcCniAddOn(),
      new MetricsServerAddOn(),
      new KarpenterAddOn(),
      new AwsLoadBalancerControllerAddOn(),
      new EbsCsiDriverAddOn(),
  • Cluster provider

    • Fargate to deploy Karpenter
    • Karpenter simplifies Kubernetes infrastructure with the right nodes at the right time.

πŸš€ Create vcluster

  • Create two vclusters with namepsace app1 and app2
    $ ./demo/create-vcl.sh app1
    $ ./demo/create-vcl.sh app2

πŸš€ Expose vcluster using Network Laoad Balancer

  • Create NLB service

    βœ— k apply -f demo/app1/service.yaml
    service/app1-lb created
    
    βœ— k apply -f demo/app2/service.yaml
    service/app2-lb created
    
    βœ— k get svc -n app1 app1-lb
    NAME      TYPE           CLUSTER-IP       EXTERNAL-IP                                                                    PORT(S)         AGE
    app1-lb   LoadBalancer   172.20.150.105   k8s-app1-app1lb-bb32c11098-3381306256798df4.elb.ap-southeast-1.amazonaws.com   443:32392/TCP   30h
    
    βœ— k get svc -n app2 app2-lb
    NAME      TYPE           CLUSTER-IP      EXTERNAL-IP                                                                    PORT(S)         AGE
    app2-lb   LoadBalancer   172.20.78.127   k8s-app2-app2lb-4690ffbcfe-bfb88a1245728e8a.elb.ap-southeast-1.amazonaws.com   443:31510/TCP   49s
  • Create CName record point to the NLB DNS

    ➜  multi-tenancy-in-eks-using-vcluster git:(master) βœ— ./demo/r53-record.sh create app2
    ➜  multi-tenancy-in-eks-using-vcluster git:(master) βœ— ping app2-eks.simflexcloud.com
    PING k8s-app2-app2lb-4690ffbcfe-bfb88a1245728e8a.elb.ap-southeast-1.amazonaws.com (13.250.162.120): 56 data bytes
  • Now we can connect to the vcluster app1 and app2 using their expose endpoint

πŸš€ Deploy applications on vcluster

  • Deploy echo and guestbook project

    βœ— ka2 apply -f demo/app2/vcluster
    ingress.networking.k8s.io/echo created
    deployment.apps/echo created
    service/echo created
    ingress.networking.k8s.io/guestbook created
    service/redis-leader created
    deployment.apps/redis-leader created
    service/redis-follower created
    deployment.apps/redis-follower created
    service/frontend created
    deployment.apps/frontend created
  • Get ALB DNS and point to the Web app endpoint

    • https://app1.simflexcloud.com

    • https://app2.simflexcloud.com

      βœ— ka2 get ingress
      NAME        CLASS   HOSTS   ADDRESS                                                          PORTS   AGE
      echo        alb     *       k8s-app2-dbb948e3be-939359744.ap-southeast-1.elb.amazonaws.com   80      10s
      guestbook   alb     *       k8s-app2-dbb948e3be-939359744.ap-southeast-1.elb.amazonaws.com   80      10s
      βœ— ./demo/r53-record.sh create app2 k8s-app2-dbb948e3be-939359744.ap-southeast-1.elb.amazonaws.com

πŸš€ Cleanup

  • Delete vcluster

    βœ— vcluster delete dev -n dev
    info   Delete vcluster dev...
    done √ Successfully deleted virtual cluster dev in namespace dev
    done √ Successfully deleted virtual cluster pvc data-dev-0 in namespace dev
  • Destroy all AWS resources within this project