CAPI is taking too long removing taint node.cluster.x-k8s.io/uninitialized:NoSchedule from nodes

[njuettner]

What steps did you take and what happened?

From reading the docs bootstrap provider can optionally taint worker nodes at creation with node.cluster.x-k8s.io/uninitialized:NoSchedule. We noticed that CAPI is taking up to 5 minutes removing the taint from nodes.

Unfortunately this is marked as optional but there is no flag to omit it. From observing the code we cannot see any place where this is possible.

We don't think it is intended, especially when managing MachinePools externally by cluster-autoscaler.

Additionally there are other provider specific controller like aws-cloud-controller-manager who sets a similar taint node.cloudprovider.kubernetes.io/uninitialized waiting for nodes becoming ready.

What did you expect to happen?

Ensure taint configuration in bootstrap provider is optional.

Cluster API version

1.4.7

Kubernetes version

No response

Anything else you would like to add?

No response

Label(s) to be applied

/kind bug /area bootstrap