search

wbuchwalter / Kubernetes-acs-engine-autoscaler

[Deprecated] Node-level autoscaler for Kubernetes clusters created with acs-engine.
Other
71 stars 22 forks source link

All nodes reported as under-utilized-undrainable #86

Open EPinci opened 6 years ago

EPinci commented 6 years ago

Hi, I managed to deploy the autoscaler to my cluster with RBAC and all post ACS-Engine v12.0 settings. I can see it is running but the logs always report:

2018-03-15 17:31:49,126 - autoscaler.cluster - INFO - ++++ Maintenance Begins ++++++
2018-03-15 17:31:49,127 - autoscaler.engine_scaler - INFO - ++++ Maintaining Nodes ++++++
2018-03-15 17:31:49,127 - autoscaler.engine_scaler - INFO - node: k8s-nodepool1-93283987-0                                                    state: under-utilized-undrainable
2018-03-15 17:31:49,127 - autoscaler.engine_scaler - INFO - node: k8s-nodepool1-93283987-1                                                    state: under-utilized-undrainable
2018-03-15 17:31:49,128 - autoscaler.engine_scaler - INFO - node: k8s-nodepool1-93283987-2                                                    state: under-utilized-undrainable
2018-03-15 17:31:49,128 - autoscaler.cluster - INFO - ++++ Maintenance Ends ++++++

As much as my nodes are well under utilized, how can I know what is actually blocking the scale down?

Thank you.

wbuchwalter commented 6 years ago

Most likely this is blocked by system pods. You can check by looking at kubectl get pods -n kube-system. kube-proxy and kube-svc-redirect pods are ignored, but your kube-dns, heapster, tunnelfront and dashboard pods might be spread among the 3 nodes.

EPinci commented 6 years ago

Indeed I have kube-system pods everywhere: heapster, kube-dns, kube-proxy, dashboard, tiller Is there a way to ignore these as well? Also I have a "non kube-system" daemonset, is it going to be an issue?

EPinci commented 6 years ago

Also looks like that the autoscaler pod itself is a blocker to scale down. There a way to bypass this one as well? Is this autoscaler PDB aware?