Today if you enter an invalid (i.e. unparseable as a Duration) value for hiveConfig.spec.

• syncSetReapplyInterval
• machinePoolPollInterval
• clusterVersionPollInterval

...the respective controller pods will crash.

I feel like we can do better – validate earlier.

First, let's try adding pattern validation to these fields, like we have for e.g. ClusterDeployment.Spec.HibernateAfter.

If for some reason this doesn't work, we should add prevalidation (simply ParseDuration() and check for non-nill error) to hive-operator and fail there instead.