Cluster Autoscaler in a pending status

[v-petukhov]

Hi there!

I've got my cluster and wanted to make it autoscaling. But, as I've discovered cluster-autoscaler pod did not run and stuck in pending state and I can't find reasons why.

Maybe someone had such situation??

P.S. Last time I've using this great package in a 7 months ago and cluster-autoscaler works well, but as I can see there is new version of it (cluster-autoscaler) using in package.

Here some of my configs and descs:

part of kube.tf:

control_plane_nodepools = [
    {
      name        = "control-plane-fsn1",
      server_type = "cpx11",
      location    = "fsn1",
      labels      = [],
      taints      = [],
      count       = 1
    },
    {
      name        = "control-plane-nbg1",
      server_type = "cpx11",
      location    = "nbg1",
      labels      = [],
      taints      = [],
      count       = 1
    },
    {
      name        = "control-plane-hel1",
      server_type = "cpx11",
      location    = "hel1",
      labels      = [],
      taints      = [],
      count       = 1
    }
  ]

  agent_nodepools = [
    {
      name        = "agent-fsn1-cpx41",
      server_type = "cpx41",
      location    = "fsn1",
      labels = [
        "node.kubernetes.io/server-usage=odysseus"
      ],
      taints               = [],
      longhorn_volume_size = 0,
      count                = 2
    },
    {
      name        = "agent-nbg1-cpx41",
      server_type = "cpx41",
      location    = "nbg1",
      labels = [
        "node.kubernetes.io/server-usage=odysseus"
      ],
      taints               = [],
      longhorn_volume_size = 0,
      count                = 2
    },
    {
      name        = "agent-cx41",
      server_type = "cx41",
      location    = "nbg1",
      labels = [
        "node.kubernetes.io/server-usage=cluster-monitoring"
      ],
      taints               = [],
      longhorn_volume_size = 0,
      count                = 1,
      backups              = true
    }
  ]

  autoscaler_nodepools = [
    {
      name        = "autoscaler-cpx41"
      server_type = "cpx41" # must be same or better than the control_plane server type (regarding disk size)!
      location    = "fsn1"
      min_nodes   = 1
      max_nodes   = 7,
      labels = {
        "node.kubernetes.io/server-usage" : "odysseus"
      }
    }
  ]

Screenshot of cluster-autoscaler pod describe:

[v-petukhov]

Heh! I've realised, what it was. Too small control plane nodes. So, I'll close discussion :)

[mabilinab]

hi @v-petukhov!
I see that you have configured taint tolerations in your cluster-autoscaler deployment. How did you do that?
I have added a taint to my master node, and now, the cluster-autoscaler pods can no longer be scheduled since they do not have the right taint toleration.