vitobotta/hetzner-k3s v2.3.2 on GitHub

Improvements

More robust cluster deletion: when deleting the cluster, we first attempt to fetch the list of the nodes to delete via kubectl, and if that fails or it's taking longer than 10 seconds, we fall back to an Hetzner API call. This update also more reliably detects autoscaled nodes for deletion. (by @KristijanL)
Cluster Autoscaler now has a dedicated root level config section allowing you fine tune some timing settings. (by @KristijanL)
Added missing volumeattachments permission to ClusteRole for the Cluster Autoscaler. (by @KristijanL)
We still let you set up masters in multiple places, but we don’t force you to put each master in a different location anymore. Until now, when you set up a cluster with multiple locations in Europe, you had to place one master in Falkenstein, one in Helsinki, and one in Nuremberg, in that order (alphabetical). But because of the higher latency between the German locations and Helsinki, you might prefer to put all the masters in just the German locations or all in the Finnish location. It’s up to you!
Published documentation with mkdocs on Github Pages (by @vvarp)
You can now pass extra arguments to the cluster autoscaler. (by @Mokto)
You can now set up Cilium egress with a floating IP. This helps make sure that all outgoing traffic goes through a node that has the floating IP attached, which can be helpful in certain situations. (by @Mokto)

Fixed a problem with SSH access on Debian that was caused by a bug in how config files were processed.
Fixed handling labels and taints that forward slashes or dots in the key.
Now you can configure the tool to run some commands before k3s is set up, not only after. This change fixes a problem that showed up when the cluster has no public IP. In that case, the network needs a bit of extra configuration before k3s starts, so we let you run those steps early. (by @KristijanL)

If you have a cluster with each master in a different place, make sure the order in the config file is fns1, hel1, nbg1. But if you want to move the Helsinki master to a German location to reduce latency, you can do that now. First, you need to drain the Helsinki master (master2). Then, delete it with kubectl delete node <cluster name>-master2. Next, remove the relevant instance from the Hetzner Console. After that, change the location for the second master from hel1 to either fsn1 or nbg1 in the configuration file. Finally, rerun hetzner-k3s create.
Replace post_create_commands with additional_post_k3s_commands. Optionally configure a additional_pre_k3s_commands section for commands that must be run before the k3s setup.