Allow setting GPUs limits in cluster autoscaling

[parthmishra]

When configuring node auto-provisioning, it would be helpful to also be able to include GPU-specific resource limits. Currently, this module only allows setting CPU and Memory limits while the API allows for GPU-specific strings.

I imagine supporting this option may require a slight change to the cluster_autoscaling block since the type is any valid GPU string identifier. Example usage might look something like:

cluster_autoscaling = {
    enabled       = true
    min_cpu_cores = 0
    max_cpu_cores = 128
    min_memory_gb = 0
    max_memory_gb = 2000
    min_gpu       = 0
    max_gpu       = 16
    gpu_type      = "nvidia-tesla-k80"
  }

Is something like this currently possible? Apologies if this has been asked before. Thanks!

[morgante]

No, it's not currently possible, but definitely a good idea.

I don't currently have time to implement this, but we would be happy to review a PR adding it.

[max-sixty]

Is this closed by #944?

[morgante]

Yes, I think so.