This project includes the Terraform configuration to deploy an RKE2 cluster in Azure.
- The terraform script does not work within Azure Cloudshell because of a Cloudshell/Terraform AzureRM provider issue.
- A .devcontainer is provided with all dependencies installed. It is not required to be used.
- It is expected that the cloud target has been set using az cloud set -name AzureUSGovernment and an az login and subscription setting has been performed.
- The .tfvar cloud variable values are determined by the Kubernetes azure cloud provider which utilises the go-autorest library which doesn't use Azure defined cloud names.
These regions were deployed and tested other regions might also work
- USGovVirginia
- USGovArizona
-
Start devcontainer in VScode or install prerequisites in your system. The devcontainer is in .devcontainer.
-
An example usage of this module can be found in the quickstart folder
This section assumes you have a publicly accessible cluster, i.e. you have set server_public_ip
to true
A script is provided to download the kubeconfig file needed to access the cluster, from KeyVault to the local machine, it also sets KUBECONFIG to point to the new kubeconfig
source scripts/fetch-kubeconfig.sh
Note. You must run this from the location where Terraform apply has been run and there is a terraform.tfstate file Note. You must source the script, also you may have to wait for a minute or two after deploying the cluster before the kubeconfig is ready
Now you can run kubectl commands against the cluster as normal, e.g. kubectl get nodes
or kubectl get pods -A
to see the status and health of the cluster.
A set of simple smoke tests is provided to validate the cluster is healthy and can communicate with Azure
See Smoke Tests for Bare Metal Kubernetes on Azure
If you set server_open_ssh_public
to true, then SSH will be allowed onto the server nodes, through the control plane load balancer.
Note. This is only recommended when troubleshooting RKE2 itself, and associated configuration such as the Azure cloud provider. For normal operation SSH access is not required.
This is done with a Azure Load Balancer NAT pool, the pool maps ports from 5000 onwards to port 22 on each of the instances, e.g.
- Port 5000 -> port 22 on instance 0
- Port 5001 -> port 22 on instance 1
- Port 5002 -> port 22 on instance 2
- etc
A script is provided that will download the SSH private key from KeyVault and tell you the public IP you need to use. The SSH username is rke2
./scripts/fetch-ssh-key.sh
Note. You must run this from the location where Terraform apply has been run and there is a terraform.tfstate file
Note. For reasons unknown sometimes the scale set takes some time to settle down, and even with a single instance, it might not be instance 0, it can be 1 or even 2, so try ports 5001 and 5002 if 5000 doesn't work