Fix crash when rancherd patches local cluster

[jakefhyde]

Issue: rancher/rancher#51980

Problem

#410 introduced data directory validation to the webhook, designed to prevent users from changing data directories once the cluster has been created. Harvester leverages rancherd, which is effectively a self-bootstrapping mechanism for creating the Rancher local cluster. Part of this bootstrapping involves patching the local cluster to set the rkeConfig: {}, which result in Rancher reconciling it like any other normal CAPI cluster.

Previously, the webhook did not come up quick enough in the rancherd process, but now the webhook and it's related validatingwebhookconfiguration exists before the cluster can be patched. The data directory validation causes the webhook to panic (and reject the request), as it was not expected that a cluster would go from an rkeConfig with a nil value, to one without one.

Solution

if oldCluster.Spec.RKEConfig == nil, accept the request.

CheckList

• Test
• Docs