Cluster API Provider Azure

This repository hosts an implementation of a provider for Azure for the OpenShift machine-api.

This provider runs as a machine-controller deployed by the machine-api-operator

Upstream Implementation

Other branches of this repository may choose to track the upstream Kubernetes Cluster-API Azure provider

In the future, we may align the master branch with the upstream project as it stabilizes within the community.

How to deploy and run azure actuator

Azure cloud resources

1. Configure RBAC rules so the actuator can CRUD resources

   $ az role definition update --role-definition azure-role.json

Deploy machine API plane with minikube

1. Install kvm

   Depending on your virtualization manager you can choose a different driver. In order to install kvm, you can run (as described in the drivers documentation):

   $ sudo yum install libvirt-daemon-kvm qemu-kvm libvirt-daemon-config-network
   $ systemctl start libvirtd
   $ sudo usermod -a -G libvirt $(whoami)
   $ newgrp libvirt

   To install to kvm2 driver:

   curl -Lo docker-machine-driver-kvm2 https://storage.googleapis.com/minikube/releases/latest/docker-machine-driver-kvm2 \
   && chmod +x docker-machine-driver-kvm2 \
   && sudo cp docker-machine-driver-kvm2 /usr/local/bin/ \
   && rm docker-machine-driver-kvm2

2. Deploying the cluster

   To install minikube v1.1.0, you can run:

   $ curl -Lo minikube https://storage.googleapis.com/minikube/releases/v1.1.0/minikube-linux-amd64 && chmod +x minikube && sudo mv minikube /usr/local/bin/
   

   To deploy the cluster:

   $ minikube start --vm-driver kvm2 --kubernetes-version v1.13.1 --v 5
   $ eval $(minikube docker-env)
   

3. Deploying machine API controllers

   For development purposes the azure machine controller itself will run out of the machine API stack. Otherwise, docker images needs to be built, pushed into a docker registry and deployed withing the stack.

   To deploy the stack:

   kustomize build config | kubectl apply --validate=false -f -
   

4. Deploy secret with Azure credentials

   Azure actuator assumes existence of a secret file (references in machine object) with base64 encoded credentials:

   apiVersion: v1
   kind: Secret
   metadata:
     name: test
     namespace: default
   type: Opaque
   data:
     azure_client_id: FILLIN
     azure_client_secret: FILLIN
     azure_region: ZWFzdHVzMg==   # eastus2 in base64
     azure_resource_prefix: b3M0LWNvbW1vbg== # os4-common in base64
     azure_resourcegroup: b3M0LWNvbW1vbg==
     azure_subscription_id: FILLIN
     azure_tenant_id: FILLIN

   $ kubectl apply -f secret.yaml

Test locally built azure actuator

1. Tear down machine-controller

   Deployed machine API plane (machine-api-controllers deployment) is (among other controllers) running machine-controller. In order to run locally built one, simply edit machine-api-controllers deployment and remove machine-controller container from it.

2. Build and run azure actuator outside of the cluster

   $ go build -o bin/machine-controller-manager sigs.k8s.io/cluster-api-provider-azure/cmd/manager

   $ .bin/machine-controller-manager --kubeconfig ~/.kube/config --logtostderr -v 5 -alsologtostderr

   If running in cointainer with podman and encountering permission issues, see hacking-guide.

3. Deploy k8s apiserver through machine manifest:

   To deploy user data secret with kubernetes apiserver initialization (under config/master-user-data-secret.yaml):

   $ kubectl apply -f config/master-user-data-secret.yaml

   To deploy kubernetes master machine (under config/master-machine.yaml):

   $ kubectl apply -f config/master-machine.yaml

4. Pull kubeconfig from created master machine

   All virtual machines created by machine templates under config can be accessed by using config/sshkey private key.

   The master public IP can be accessed from Azure Portal. Once done, you can collect the kube config by running:

   $ ssh -i config/sshkey capi@PUBLICIP 'sudo cat /root/.kube/config' > kubeconfig
   $ kubectl --kubeconfig=kubeconfig config set-cluster kubernetes --server=https://PUBLICIP:8443
   

   Once done, you can access the cluster via kubectl. E.g.

   $ kubectl --kubeconfig=kubeconfig get nodes

Deploy k8s cluster in Azure with machine API plane deployed

1. Generate bootstrap user data

   To generate bootstrap script for machine api plane, simply run:

   $ ./examples/generate-bootstrap.sh

   The script requires AZURE_SUBSCRIPTION_ID, AZURE_TENANT_ID, AZURE_CLIENT_ID and AZURE_CLIENT_SECRET environment variables to be set. It generates config/bootstrap.yaml secret for master machine under config/master-machine.yaml.

   The generated bootstrap secret contains user data responsible for:

   • deployment of kube-apiserver
   • deployment of machine API plane with azure machine controllers
   • generating worker machine user data script secret deploying a node
   • deployment of worker machineset

2. Deploy machine API plane through machine manifest:

   First, deploy generated bootstrap secret:

   $ kubectl apply -f config/bootstrap.yaml

   Then, deploy master machine (under config/master-machine.yaml):

   $ kubectl apply -f config/master-machine.yaml

3. Pull kubeconfig from created master machine

   All virtual machines created by machine templates under config can be accessed by using config/sshkey private key.

   The master public IP can be accessed from Azure Portal. Once done, you can collect the kube config by running:

   $ ssh -i config/sshkey capi@PUBLICIP 'sudo cat /root/.kube/config' > kubeconfig
   $ kubectl --kubeconfig=kubeconfig config set-cluster kubernetes --server=https://PUBLICIP:8443
   

   Once done, you can access the cluster via kubectl. E.g.

   $ kubectl --kubeconfig=kubeconfig get nodes