exec ./cluster-autoscaler: argument list too long

[laurentpellegrino]

Using a kubernetes.tf that defines 22 cluster_autoscaler_nodepools, cluster-autoscaler pods crash in loop:

$ kubectl get pods --all-namespaces | grep autoscaler
kube-system cluster-autoscaler-hetzner-cluster-autoscaler-695f679f4b-55wwn 0/1 CrashLoopBackOff 7 (37s ago) 11m
kube-system cluster-autoscaler-hetzner-cluster-autoscaler-695f679f4b-9t5kw 0/1 CrashLoopBackOff 7 (3m31s ago) 14m
kube-system cluster-autoscaler-hetzner-cluster-autoscaler-84bdf5d844-v2znd 1/1 Running 10 (14h ago) 38d

Here are logs from a failing pod:

$ kubectl logs cluster-autoscaler-hetzner-cluster-autoscaler-695f679f4b-55wwn -n kube-system
exec ./cluster-autoscaler: argument list too long

The issue seems to come from the way nodepool configurations are passed. Is there a way to increase the argument list limit or pass data differently so that autoscaler starts smoothly?

[mlinares1998]

Hi @laurentpellegrino
Maybe it has to be with the fact the module saves the nodePools as a ENV string rather than a mounted JSON config file as stated in:

https://github.com/kubernetes/autoscaler/blob/master/cluster-autoscaler%2Fcloudprovider%2Fhetzner%2FREADME.md

This would require changes to the module to create a secret with the JSON file and mount it instead.

Right now I don't know about the current status of the project, @M4t7e has been absent for a while.