Question

A CloudOps engineer applies a stack update to a production CloudFormation stack that provisions an RDS database and application tier. During the update, one resource fails to create and CloudFormation begins rolling back. The rollback also fails because a security group cannot be deleted (it is still referenced by a resource created outside the template). The stack is now in UPDATE_ROLLBACK_FAILED state. What is the correct way to recover the stack so future updates can proceed?

Accepted Answer

Manually resolve the underlying issue (remove the external reference to the security group), then call ContinueUpdateRollback, optionally skipping the resource that cannot roll back.

Answer

Delete the stack entirely and redeploy from the template, accepting recreation of the RDS database.

Answer

Run drift detection on the stack to force CloudFormation to reconcile the failed resource automatically.

Answer

Create a change set against the failed stack to preview and apply the corrective update.