Skip to content

Latest commit

 

History

History
331 lines (249 loc) · 13.7 KB

README.md

File metadata and controls

331 lines (249 loc) · 13.7 KB

Kubernetes Cloud Controller Manager for Hetzner Cloud

GitHub Actions status

The Hetzner Cloud cloud-controller-manager integrates your Kubernetes cluster with the Hetzner Cloud & Robot APIs.

Features

  • Node:
    • Updates your Node objects with information about the server from the Cloud & Robot API.
    • Instance Type, Location, Datacenter, Server ID, IPs.
  • Node Lifecycle:
    • Cleans up stale Node objects when the server is deleted in the API.
  • Routes (if enabled):
    • Routes traffic to the pods through Hetzner Cloud Networks. Removes one layer of indirection in CNIs that support this.
  • Load Balancer:
    • Watches Services with type: LoadBalancer and creates Hetzner Cloud Load Balancers for them, adds Kubernetes Nodes as targets for the Load Balancer.

Read more about cloud controllers in the Kubernetes documentation.

Node Metadata Example

apiVersion: v1
kind: Node
metadata:
  labels:
    node.kubernetes.io/instance-type: cx11
    topology.kubernetes.io/region: fsn1
    topology.kubernetes.io/zone: fsn1-dc8
  name: node
spec:
  podCIDR: 10.244.0.0/24
  providerID: hcloud://123456 # <-- Hetzner Cloud Server ID
status:
  addresses:
    - address: node
      type: Hostname
    - address: 1.2.3.4 # <-- Hetzner Cloud Server public ipv4
      type: ExternalIP

Deployment

This deployment example uses kubeadm to bootstrap an Kubernetes cluster, with flannel as overlay network agent. Feel free to adapt the steps to your preferred method of installing Kubernetes.

These deployment instructions are designed to guide with the installation of the hcloud-cloud-controller-manager and are by no means an in depth tutorial of setting up Kubernetes clusters. Previous knowledge about the involved components is required.

Please refer to the kubeadm cluster creation guide, which these instructions are meant to augment and the kubeadm documentation.

  1. The cloud controller manager adds its labels when a node is added to the cluster. For current Kubernetes versions, this means we have to add the --cloud-provider=external flag to the kubelet before initializing the control plane with kubeadm init. To do accomplish this we add this systemd drop-in unit /etc/systemd/system/kubelet.service.d/20-hcloud.conf:

    [Service]
    Environment="KUBELET_EXTRA_ARGS=--cloud-provider=external"
    

    Note: the --cloud-provider flag is deprecated since K8S 1.19. You will see a log message regarding this. For now (v1.26) it is still required.

  2. Now the control plane can be initialized:

    sudo kubeadm init --pod-network-cidr=10.244.0.0/16
  3. Configure kubectl to connect to the kube-apiserver:

    mkdir -p $HOME/.kube
    sudo cp -i /etc/kubernetes/admin.conf $HOME/.kube/config
    sudo chown $(id -u):$(id -g) $HOME/.kube/config
  4. Deploy the flannel CNI plugin:

    kubectl apply -f https://raw.githubusercontent.com/coreos/flannel/v0.9.1/Documentation/kube-flannel.yml
  5. Patch the flannel deployment to tolerate the uninitialized taint:

    kubectl -n kube-system patch ds kube-flannel-ds --type json -p '[{"op":"add","path":"/spec/template/spec/tolerations/-","value":{"key":"node.cloudprovider.kubernetes.io/uninitialized","value":"true","effect":"NoSchedule"}}]'
  6. Create a secret containing your Hetzner Cloud API token.

    kubectl -n kube-system create secret generic hcloud --from-literal=token=<hcloud API token>
  7. Deploy hcloud-cloud-controller-manager

    Using Helm (recommended):

    helm repo add hcloud https://charts.hetzner.cloud
    helm repo update hcloud
    helm install hccm hcloud/hcloud-cloud-controller-manager -n kube-system
    

    See the Helm chart README for more info.

    Legacy installation method:

    kubectl apply -f https://github.com/hetznercloud/hcloud-cloud-controller-manager/releases/latest/download/ccm.yaml

Networks support

When you use the Cloud Controller Manager with networks support, the CCM is in favor of allocating the IPs (& setup the routing) (Docs: https://kubernetes.io/docs/concepts/architecture/cloud-controller/#route-controller). The CNI plugin you use needs to support this k8s native functionality (Cilium does it, I don't know about Calico & WeaveNet), so basically you use the Hetzner Cloud Networks as the underlying networking stack.

When you use the CCM without Networks support it just disables the RouteController part, all other parts work completely the same. Then just the CNI is in charge of making all the networking stack things. Using the CCM with Networks support has the benefit that your node is connected to a private network so the node doesn't need to encrypt the connections and you have a bit less operational overhead as you don't need to manage the Network.

If you want to use the Hetzner Cloud Networks Feature, head over to the Deployment with Networks support documentation.

If you manage the network yourself it might still be required to let the CCM know about private networks. You can do this by adding the environment variable with the network name/ID in the CCM deployment.

          env:
            - name: HCLOUD_NETWORK
              valueFrom:
                secretKeyRef:
                  name: hcloud
                  key: network

You also need to add the network name/ID to the secret: kubectl -n kube-system create secret generic hcloud --from-literal=token=<hcloud API token> --from-literal=network=<hcloud Network_ID_or_Name> .

Kube-proxy mode IPVS and HCloud LoadBalancer

If kube-proxy is run in IPVS mode, the Service manifest needs to have the annotation load-balancer.hetzner.cloud/hostname where the FQDN resolves to the HCloud LoadBalancer IP.

See hetznercloud#212

Versioning policy

We aim to support the latest three versions of Kubernetes. After a new Kubernetes version has been released we will stop supporting the oldest previously supported version. This does not necessarily mean that the Cloud Controller Manager does not still work with this version. However, it means that we do not test that version anymore. Additionally, we will not fix bugs related only to an unsupported version.

With Networks support

Kubernetes Cloud Controller Manager Deployment File
1.28 main https://github.com/hetznercloud/hcloud-cloud-controller-manager/releases/latest/download/ccm-networks.yaml
1.27 main https://github.com/hetznercloud/hcloud-cloud-controller-manager/releases/latest/download/ccm-networks.yaml
1.26 main https://github.com/hetznercloud/hcloud-cloud-controller-manager/releases/latest/download/ccm-networks.yaml
1.25 main https://github.com/hetznercloud/hcloud-cloud-controller-manager/releases/latest/download/ccm-networks.yaml
1.24 v1.17.2 https://github.com/hetznercloud/hcloud-cloud-controller-manager/releases/download/v1.17.2/ccm-networks.yaml
1.23 v1.13.2 https://github.com/hetznercloud/hcloud-cloud-controller-manager/releases/download/v1.13.2/ccm-networks.yaml

Without Networks support

Kubernetes Cloud Controller Manager Deployment File
1.28 main https://github.com/hetznercloud/hcloud-cloud-controller-manager/releases/latest/download/ccm.yaml
1.27 main https://github.com/hetznercloud/hcloud-cloud-controller-manager/releases/latest/download/ccm.yaml
1.26 main https://github.com/hetznercloud/hcloud-cloud-controller-manager/releases/latest/download/ccm.yaml
1.25 main https://github.com/hetznercloud/hcloud-cloud-controller-manager/releases/latest/download/ccm.yaml
1.24 v1.17.2 https://github.com/hetznercloud/hcloud-cloud-controller-manager/releases/download/v1.17.2/ccm.yaml
1.23 v1.13.2 https://github.com/hetznercloud/hcloud-cloud-controller-manager/releases/download/v1.13.2/ccm.yaml

Unit tests

To run unit tests locally, execute

go test ./...

Check that your go version is up-to-date, tests might fail if it is not.

If in doubt, check which go version is installed in the ci.yaml GitHub Actions Workflow:

go-version: "1.21"

E2E Tests

The Hetzner Cloud cloud controller manager was tested against all supported Kubernetes versions. We also test against the same k3s releases (Sample: When we support testing against Kubernetes 1.20.x we also try to support k3s 1.20.x). We try to keep compatibility with k3s but never guarantee this.

You can run the tests with the following commands. Keep in mind, that these tests run on real cloud servers and will create Load Balancers that will be billed.

Test Server Setup:

1x CPX21 (Ubuntu 18.04)

Requirements: Docker and Go 1.21

  1. Configure your environment correctly
export HCLOUD_TOKEN=<specifiy a project token>
export K8S_VERSION=k8s-1.21.0 # The specific (latest) version is needed here
export USE_SSH_KEYS=key1,key2 # Name or IDs of your SSH Keys within the Hetzner Cloud, the servers will be accessable with that keys
export USE_NETWORKS=yes # if `yes` this identidicates that the tests should provision the server with cilium as CNI and also enable the Network related tests
## Optional configuration env vars:
export TEST_DEBUG_MODE=yes # With this env you can toggle the output of the provision and test commands. With `yes` it will log the whole output to stdout
export KEEP_SERVER_ON_FAILURE=yes # Keep the test server after a test failure.
  1. Run the tests
go test ./tests/e2e -tags e2e -v -timeout 60m

The tests will now run and cleanup themselves afterward. Sometimes it might happen that you need to clean up the project manually via the Hetzner Cloud Console or the hcloud-cli .

For easier debugging on the server we always configure the latest version of the hcloud-cli with the given HCLOUD_TOKEN and a few bash aliases on the host:

alias k="kubectl"
alias ksy="kubectl -n kube-system"
alias kgp="kubectl get pods"
alias kgs="kubectl get services"

The test suite is split in three parts:

  • General Part: Sets up the test env & checks if the HCCM Pod is properly running
    • Build Tag: e2e
  • Cloud Part: Tests regular functionality against a Cloud-only environment
    • Build Tag: e2e && !robot
  • Robot Part: Tests Robot functionality against a Cloud+Robot environment
    • Build Tag: e2e && robot

Local test setup

This repository provides skaffold to easily deploy / debug this controller on demand

Requirements

  1. Install hcloud-cli
  2. Install k3sup
  3. Install cilium
  4. Install docker

You will also need to set a HCLOUD_TOKEN in your shell session

Manual Installation guide

  1. Create an SSH key

Assuming you already have created an ssh key via ssh-keygen

hcloud ssh-key create --name ssh-key-ccm-test --public-key-from-file ~/.ssh/id_rsa.pub 
  1. Create a server
hcloud server create --name ccm-test-server --image ubuntu-20.04 --ssh-key ssh-key-ccm-test --type cx11 
  1. Setup k3s on this server
k3sup install --ip $(hcloud server ip ccm-test-server) --local-path=/tmp/kubeconfig --cluster --k3s-channel=v1.23 --k3s-extra-args='--no-flannel --no-deploy=servicelb --no-deploy=traefik --disable-cloud-controller --disable-network-policy --kubelet-arg=cloud-provider=external'
  • The kubeconfig will be created under /tmp/kubeconfig
  • Kubernetes version can be configured via --k3s-channel
  1. Switch your kubeconfig to the test cluster. Very important: exporting this like
export KUBECONFIG=/tmp/kubeconfig
  1. Install cilium + test your cluster
cilium install
  1. Add your secret to the cluster
kubectl -n kube-system create secret generic hcloud --from-literal="token=$HCLOUD_TOKEN"
  1. Deploy the hcloud-cloud-controller-manager
SKAFFOLD_DEFAULT_REPO=your_docker_hub_username skaffold dev
  • docker login required
  • Skaffold is using your own Docker Hub repo to push the HCCM image.
  • After the first run, you might need to set the image to "public" on hub.docker.com

On code change, Skaffold will repack the image & deploy it to your test cluster again. It will also stream logs from the hccm Deployment.

After setting this up, only the command from step 7 is required!=

License

Apache License, Version 2.0