janosmiko / hetzner-k3s Goto Github PK

Currently, the servers created by the Cluster Autoscaler are not added to the firewall even when the HCLOUD_FIREWALL environment variable is set properly.

As a workaround set the firewall to be applied to servers using "hcloud/node-group" labels.

In a new and empty hcloud project I created a token and used the docker image and the example yaml-file (see below). Only change is the k3s version, for which I chose 1.25.5-k3s1. All resources are created as expected, but when hetzner-k3s reaches the "Deploying Hetzner CSI Driver..." step the application crashes:

Command output: Error from server (NotFound): error when deleting "STDIN": csidrivers.storage.k8s.io "csi.hetzner.cloud" not found
Exit code: exit status 1
Error: cannot create cluster: cannot deploy hetzner csi driver: shell command exited with error: 1, output: Error from server (NotFound): error when deleting "STDIN": csidrivers.storage.k8s.io "csi.hetzner.cloud" not found

Restarting the cluster-create command does not yield a different result.

In this line

I've manually added the CSIDriver with the yaml configuration (as defined for deletion)

I'm happy to provide more information if requried.

As mentioned above my yaml-file:

---
hetzner_token: <my-token>
cluster_name: test
kubeconfig_path: "/cluster/kubeconfig"
k3s_version: v1.25.5+k3s1
public_ssh_key_path: "/tmp/.ssh/id_rsa.pub"
private_ssh_key_path: "/tmp/.ssh/id_rsa"
verify_host_key: false
location: nbg1
schedule_workloads_on_masters: false
masters:
  instance_type: cpx21
  instance_count: 3
worker_node_pools:
  - name: small
    instance_type: cpx21
    instance_count: 4
  - name: big
    instance_type: cpx31
    instance_count: 2
autoscaling_node_pools:
  - name: small
    instance_type: cx41
    instance_min: 3
    instance_max: 5
  - name: big
    instance_type: cpx51
    instance_max: 10
post_create_commands:
  - apt update
  - apt upgrade -y
  - apt autoremove -y
  - shutdown -r now
enable_encryption: true

Command to run:

docker run --rm -it \
  -v ${PWD}:/cluster \
  -v ${HOME}/.ssh:/tmp/.ssh \
  janosmiko/hetzner-k3s:v0.1.5 \
  create-cluster \
  --config-file /cluster/test.yaml

If you can use the env variable (which did not work for me btw), the hetzner_token key should not be required in the yaml.

If the number of masters is 3, make it possible to create the masters in 3 different regions.
Eg:

• first in nbg1
• second in fsn1
• third in hel1