Fix version of cloud-controller-manager

[KockaAdmiralac]

A few hours ago, the Kubernetes team pushed this pull request to Google Cloud registries, which appears to have broken students' cluster setups. The cloud-controller-manager pod crashes during startup, ending up in a CrashLoopBackOff state. The output of kubectl logs -n kube-system cloud-controller-manager-[generated ID] is:

flag provided but not defined: -allocate-node-cidrs

Usage of /go-runner:

  -also-stdout

     useful with log-file, log to standard output as well as the log file

  -log-file string

     If non-empty, save stdout to this file

  -redirect-stderr

     treat stderr same as stdout (default true)

go-runner is the executable from the registry.k8s.io/build-image/go-runner image, and despite the team setting CMD ["/cloud-controller-manager"] it seems like when kops runs the cloud-controller-manager container it actually runs the go-runner executable, and not the actual cloud-controller-manager, which is why the cloud-controller-manager-specific flag is not recognized.

Either way, this pull request should fix the error by just downgrading to a previous version of the cloud-controller-manager image. As of yet, nobody has reported issues on the image's repository.

[frikai]

Tested the change and can confirm that it fixes the issue