AKS cluster-autoscaler stucks in "Initializing".

[monaka]

On AKS (not with VMSS) 1.14.3, cluster-autoscaler stucks under "Initializing" state.

Nodes information on my node is:

NAME                       STATUS   ROLES   AGE     VERSION                             INTERNAL-IP   EXTERNAL-IP   OS-IMAGE             KERNEL-VERSION      CONTAINER-RUNTIME
aks-agentpool-39680657-1   Ready    agent   8d      v1.14.3                             10.240.0.35   <none>        Ubuntu 16.04.6 LTS   4.15.0-1052-azure   docker://3.0.6
aks-agentpool-39680657-2   Ready    agent   8d      v1.14.3                             10.240.0.4    <none>        Ubuntu 16.04.6 LTS   4.15.0-1052-azure   docker://3.0.6
virtual-node-aci-linux     Ready    agent   2d21h   v1.13.1-vk-v0.9.0-1-g7b92d1ee-dev   10.240.0.48   <none>        <unknown>            <unknown>           <unknown>

After inspecting code, I convinced it is caused only in case Virtual Kubelet was deployed.
I'll send the PR for fixing this soon.

[monaka]

Here is the root cause of this issue.

$ kubectl get nodes aks-agentpool-39680657-1 -o=jsonpath='{.spec.providerID}'
azure:///subscriptions/{{subscription-id}}/resourceGroups/mc_camino_camino-wus-2_westus2/providers/Microsoft.Compute/virtualMachines/aks-agentpool-39680657-1
$ kubectl get nodes virtual-node-aci-linux -o=jsonpath='{.spec.providerID}'
$  #no output

The controller doesn't set spec.providerID to Virtual Kubelet node.
At the same, there is no care whether spec.providerID is empty.

autoscaler/cluster-autoscaler/cloudprovider/azure/azure_cloud_provider.go

Lines 94 to 99 in 93298c0

	klog.V(6).Infof("Searching for node group for the node: %s\n", node.Spec.ProviderID)
	ref := &azureRef{
	Name: node.Spec.ProviderID,
	}
	return azure.azureManager.GetAsgForInstance(ref)

So every RunOnce() calls are exited with error.