Support for Auto scaling

[lpellegr]

Your project looks very promising. Unfortunately, I was looking for a way to add auto scaling agent nodes. Any plan to add such support?

[lpellegr]

OMG, I did not notice the Advanced section. It seems you already support auto scaling! I will give a try immediately.

[M4t7e]

Yes, it is supported. Unfortunately there is still the kubernetes/autoscaler#7210 bug. Let me fix that quickly...

Right now this project is under heavy development and there might come a breaking change in 0.* releases. But I'm almost done and just working on Longhorn integration as last component on my to do list. After that, it'll be hopefully stable with only backwards compatible changes.

[lpellegr]

Thank you so much for all the hard work on this. The use of Talos looks fantastic.

Up until now, I’d been using https://github.com/kube-hetzner/terraform-hcloud-kube-hetzner, but the current version is completely broken, and unfortunately, my live cluster went down today as well. I'm giving your project a try right now.

[lpellegr]

@M4t7e Is there a way to pass the autoscaler some options. Something like:

  cluster_autoscaler_extra_args = [
    "--scale-down-delay-after-add=45m",
    "--scale-down-delay-after-delete=4m",
    "--scale-down-unneeded-time=5m"
  ]

[M4t7e]

The bug should be fixed in 0.2.1. I just tested it and the Autoscaler works now without errors.

Up until now, I’d been using https://github.com/kube-hetzner/terraform-hcloud-kube-hetzner, but the current version is completely broken, and unfortunately, my live cluster went down today as well. I'm giving your project a try right now.

I'm familiar with that project as I am part of the core team and have mainly contributed to the networking and security aspects. It is a very nice and popular project and can totally recommend that in general.

For our business, we require a cluster that is completely reliable and reproducible, with no automatic upgrades or other surprises running in the background. Additionally, I have not found any Kubernetes project for Hcloud that adequately manages the versioning of all its components. This project will not be maintenance free, but hopefully very stable and predictable for production workloads.

[lpellegr]

Thanks. I just did a try with version 0.2.1 but without success (#5). To be honest, I don't have experience with terraform and just try to follow the docs in the README.

[M4t7e]

@M4t7e Is there a way to pass the autoscaler some options. Something like:

  cluster_autoscaler_extra_args = [
    "--scale-down-delay-after-add=45m",
    "--scale-down-delay-after-delete=4m",
    "--scale-down-unneeded-time=5m"
  ]

I've just added support for this in version 0.3.0. You can find the updated documentation in the README. Please note that the variable name has changed from autoscaler_nodepools to cluster_autoscaler_nodepools.

[lpellegr]

Incredible! I will give it a try. Thanks a lot.

[M4t7e]

The variable cluster_autoscaler_extra_args has been renamed to cluster_autoscaler_helm_values. This naming and functionality have been extended to all Helm-based components, allowing for the merging and overwriting of existing values with the provided ones, similar to using multiple Helm values files via the Helm CLI.