Is it possible to enable autoscaling on existing cluster?

[otavio]

Considering an existing cluster is deployed, is it possible to add autoscaling support?

[mysticaltech]

Yes, normally you just upgrade the module and apply.

[Manikanta-Reddy-Pasala]

i have only enabled metrics_server. no other stack(Prometheus, grafana) is installed.

[ifeulner]

The status output does not really fit to what you describe.

There are four nodes in the cluster Health: Healthy (ready=4 unready=0 (resourceUnready=0) notStarted=0 longNotStarted=0 registered=4 longUnregistered=0)

And cluster-autoscaler sees no need to do anything - Name: k3s-autoscaler Health: Healthy (ready=0 unready=0 (resourceUnready=0) notStarted=0 longNotStarted=0 registered=0 longUnregistered=0 cloudProviderTarget=0 (minSize=0, maxSize=3))

So the load does not need any scaling... you can set min_nodesto 1 to force autoscaler to create at least 1 node.

[captnCC]

Like @ifeulner said: from the logs, the autoscaler seems to be configured correctly.
So I think it's more a problem of the pod resource allocation. Do you have resource requirements set for them?
This is required as its the metric by which the cluster-autoscaler works (not the real machine load!)

I have tested a similar case, just instead of starting with 1.5, I started with a 1.6 cluster and no configured autoscaler, then added the autoscaler pool.

[Manikanta-Reddy-Pasala]

Thank you so much for the reply. i will try adding resource limits and share you the results soon.

[Manikanta-Reddy-Pasala]

Hi @captnCC ,

i have added resource limits for my deployments.

resources:
            limits:
              cpu: '1'
              memory: 1500Mi
            requests:
              cpu: 200m
              memory: 512Mi

do I need to apply HPA also or just a resource limit is fine for auto-scaling to work?