With a Cloud Provider enabled, Windows worker nodes cannot join cluster due to DNS suffixes being passed to kubelet as commands

[rosskirkpat]

What kind of request is this (question/bug/enhancement/feature request):
bug

Steps to reproduce (least amount of steps as possible):

• setup multiple DNS suffixes in AWS DHCP Options Sets
• enable aws cloud provider in a new mixed cluster
• add new windows worker node

Result:

kubelet will be unable to start due to interpreting the DNS suffix options as commands.

For example:

--hostname-override=ip-X-X-X-X.XXXXX.com gad.XXXXX.com apa.gad.XXXXX.com eur.gad.XXXXX.com gmea.gad.XXXXX.com nam.XXXXX.com us-east-1.compute.internal --feature-gates=HyperVContainer=true,WindowsGMSA=true

They will see the following when adding windows nodes in the kubelet logs:

10328 server.go:163] unknown command: apa.gad.XXXXX.com

Other details that may be helpful:

Issue can be addressed in the following powershell script by truncating to the first hostname returned.

https://github.com/rancher/rke-tools/blob/v0.1.65/windows/cloud-provider.psm1#L229-L232

gz#12873

[sowmyav27]

Validated On 2.4.9-rc5

• Deploy windows cluster - AWS cloud provider enabled, on k8s 1.18,1.17,1.16 and 1.15 (1 etcd/control, 1 worker node - linux nodes and 3 windows worker nodes)
• Deploy PVC and LB
• PVC and LB deployed successfully on all clusters except for this issue

[sowmyav27]

Closing this issue as the forward port issue is tracked separately.