Note: this is based on upmc-enterprises/kubeadm-aws.
This is a Terraform configuration for provisioning a simple nodeless Kubernetes cluster that uses Milpa and Kiyot. Milpa is a nodeless engine developed by Elotl and Kiyot is a CRI that plugs Milpa into Kubernetes. The script will create a new VPC and subnet then create a nodeless kubernets cluster inside the VPC.
Create a file at ~/env.tfvars:
$ cp env.tfvars.example ~/env.tfvars
$ vi ~/env.tfvars
Once you have filled in the necessary variables in env.tfvars, apply the configuration:
$ terraform init # Only needed the first time.
[...]
$ terraform apply -var-file ~/env.tfvars
[...]
Apply complete! Resources: 24 added, 0 changed, 0 destroyed.
Outputs:
master_ip = 3.81.184.107
worker_ip = 54.90.138.204
This will create a cluster with one master and one worker.
Inline-style:
SSH into the master node and check the status of the cluster:
ubuntu@ip-10-0-100-66:~$ kubectl cluster-info
Kubernetes master is running at https://10.0.100.66:6443
KubeDNS is running at https://10.0.100.66:6443/api/v1/namespaces/kube-system/services/kube-dns:dns/proxy
To further debug and diagnose cluster problems, use 'kubectl cluster-info dump'.
ubuntu@ip-10-0-100-66:~$ kubectl get nodes
NAME STATUS ROLES AGE VERSION
ip-10-0-100-135 Ready <none> 62s v1.14.0
ip-10-0-100-66 Ready master 97s v1.14.0
ubuntu@ip-10-0-100-66:~$
At this point, the cluster is ready to use.
Stateless pods (no persistent volume claims or locally attached volumes) will be scheduled via Milpa. For example:
ubuntu@ip-10-0-100-66:~$ kubectl run nginx --image=nginx
will run the pod(s) created by the deployment via Milpa.
If you have both Milpa and non-Milpa workers in your cluster, you can also use a nodeSelector to assign a pod to a non-Milpa worker. Label your non-Milpa worker node with e.g. mycompany.com/worker, then add a nodeselector:
spec:
nodeSelector:
mycompany.com/worker: ""
Make sure all pods and services are removed. On the master:
ubuntu@ip-10-0-100-66:~$ for ns in $(kubectl get namespaces | tail -n+2 | awk '{print $1}'); do kubectl delete --all deployments --namespace=$ns; kubectl delete --all services --namespace=$ns; kubectl delete --all daemonsets --namespace=$ns; kubectl delete --all pods --namespace=$ns; done
No resources found
service "kubernetes" deleted
No resources found
No resources found
[...]
ubuntu@ip-10-0-100-66:~$
Then you can log out from the master, and use Terraform to tear down the infrastructure:
$ terraform destroy -var-file ~/env.tfvars
[...]
Plan: 0 to add, 0 to change, 24 to destroy.
Do you really want to destroy all resources?
Terraform will destroy all your managed infrastructure, as shown above.
There is no undo. Only 'yes' will be accepted to confirm.
Enter a value: yes
[...]
Destroy complete! Resources: 24 destroyed.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent