I'm trying to setup k8s autoscaler, I went through: https://hub.helm.sh/charts/stable/cluster-autoscaler and did a lot of experiment but cannot make it work. I have two choices but both not working:
- deploy autoscaler on master, I tried:
helm install my-release stable/cluster-autoscaler --set autoDiscovery.clusterName=hide-cluster,rbac.serviceAccount.name=default,tolerations[0].effect=NoSchedule,tolerations[0].operator="Equal",tolerations[0].value=/"true/",tolerations[0].key=node-role.kubernetes.io/master
but it always says the fieldtolerations[0].value=/"true/"
is not as expected, I tried several possible versions I can think of, like using 'true', "true", true etc. none of them works. - deploy on a non-master node: it asks for network plugin in this case... I think only the master node has network plugin.
There is very very limited document about how to setup autoscaler without using kops or EKS. I also wonder how many autoscaler solution we have on the market? It seems not many learning resources online for autoscaler.
Edit: I finally make the autoscaler run following this guide: https://docs.aws.amazon.com/eks/latest/userguide/cluster-autoscaler.html
But it does not include EC2 instances created by ASG to nodes. I checked the log it says the ASG is already found and registered, no error or other hint I can find why no nodes shows up in kubectl get nodes