The Hetzner Cloud controller manager integrates your Kubernetes cluster with the Hetzner Cloud API.
- instances interface: adds the server type to the
node.kubernetes.io/instance-typelabel, sets the external ipv4 and ipv6 addresses and deletes nodes from Kubernetes that were deleted from the Hetzner Cloud. - zones interface: makes Kubernetes aware of the failure domain of the server by setting the
topology.kubernetes.io/regionandtopology.kubernetes.io/zonelabels on the node. - Private Networks: allows to use Hetzner Cloud Private Networks for your pods traffic.
- Load Balancers: allows to use Hetzner Cloud Load Balancers with Kubernetes Services
Read more about cloud controllers in the Kubernetes documentation.
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: ExternalIPThis 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.
-
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=externalflag to thekubeletbefore initializing the control plane withkubeadm 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-providerflag is deprecated since K8S 1.19. You will see a log message regarding this. For now (v1.26) it is still required. -
Now the control plane can be initialized:
sudo kubeadm init --pod-network-cidr=10.244.0.0/16
-
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
-
Deploy the flannel CNI plugin:
kubectl apply -f https://raw.githubusercontent.com/coreos/flannel/v0.9.1/Documentation/kube-flannel.yml
-
Patch the flannel deployment to tolerate the
uninitializedtaint: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"}}]' -
Create a secret containing your Hetzner Cloud API token.
kubectl -n kube-system create secret generic hcloud --from-literal=token=<hcloud API token>
-
Deploy
hcloud-cloud-controller-managerUsing Helm (recommended):
helm repo add hcloud https://charts.hetzner.cloud helm repo update hcloud helm install hccm hcloud/hcloud-cloud-controller-manager -n kube-systemSee 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
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>
.
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
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.
To run unit tests locally, execute
go test $(go list ./... | grep -v e2e) -vCheck that your go version is up to date, tests might fail if it is not.
If in doubt, check which go version the test:unit section in .gitlab-ci.yml
has set in the image: golang:$VERSION.
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
- 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.- Run the tests
go test $(go list ./... | grep e2e) -v -timeout 60mThe tests will now run and cleanup themselves afterwards. 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"This repository provides skaffold to easily deploy / debug this controller on demand
- Install hcloud-cli
- Install k3sup
- Install cilium
- Install docker
You will also need to set a HCLOUD_TOKEN in your shell session
- 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
- Create a server
hcloud server create --name ccm-test-server --image ubuntu-20.04 --ssh-key ssh-key-ccm-test --type cx11
- Setup k3s on this server
k3sup install --ip $(hcloud server ip ccm-test-server) --local-path=/tmp/kubeconfig --cluster --k3s-channel=v1.23 --k3s-extra-args='--no-flannel --no-deploy=servicelb --no-deploy=traefik --disable-cloud-controller --disable-network-policy --kubelet-arg=cloud-provider=external'
- The kubeconfig will be created under
/tmp/kubeconfig - Kubernetes version can be configured via
--k3s-channel
- Switch your kubeconfig to the test cluster. Very important: exporting this like
export KUBECONFIG=/tmp/kubeconfig
- Install cilium + test your cluster
cilium install
- Add your secret to the cluster
kubectl -n kube-system create secret generic hcloud --from-literal="token=$HCLOUD_TOKEN"
- Deploy the hcloud-cloud-controller-manager
SKAFFOLD_DEFAULT_REPO=your_docker_hub_username skaffold dev
docker loginrequired- 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!=
Apache License, Version 2.0