Cluster Autoscaler cannot parse Windows VMAS agent pool created by AKS-Engine

[ankrause]

CloudProvider: Azure
Kubernetes version: 1.13.1
Cluster Autoscaler version: 1.13.1
AKS-Engine version: 0.28.1 and 0.29.1

I deployed a hybrid cluster to Azure using AKS-Engine with VMAS-based agent pools and small initial node counts (2 and 2). The resulting VM names followed the format {4 digits}k8s{3 digits}, where the last digit was the instance number which could presumably grow beyond a single digit. When evaluating scale options, the Cluster Autoscaler attempted to parse those machine names using the regex:

k8sWindowsVMNamingFormat               = "^([a-fA-F0-9]{5})([0-9a-zA-Z]{3})([a-zA-Z0-9]{4,6})$"

Since the regex doesn't match against the generated VM name, it returned an error indicating the resource was missing an identifier:

vmNameParts := vmnameWindowsRegexp.FindStringSubmatch(vmName)
if len(vmNameParts) != 4 {
    return "", "", -1, -1, fmt.Errorf("resource name was missing from identifier")
}

Hopefully this is the right repository for this issue.

[aleksandra-malinowska]

cc @feiskyer

[feiskyer]

@ankrause Thanks of reporting the issue. It should be fixed with latest aks-engine pattern.

[feiskyer]

/assign