Issue

At present, If a user tries to delete an instance of operator's CRDs (eg: TektonChains) after the operator is uninstalled on a cluster, the deletion will be blocked as there will be no webhook endpoints to validate the delete operation.

This issue is caused by the fact that, OpenShift-pipelines operator handles the registration of its own mutatinga and validating webhook configurations explicitly. This had to be done as OLM management of webhooks does not fit certain assumptions in knative/pkg refer this

Possible Solution

Add graceful shutdown logic to tektoncd/operator webhooks to remove the webhook registrations added by the webhook on start up.

Acceptance Criteria

• deletions of instances of CRDs like TektoncdChains are not blocked by admission controller, especially on a cluster where the operator is uninstalled.
• a mechansim exists in operator code to make sure the webhooks are unregistered when the the webhook pods go down.