To be clear, the issue here is that the domain is still active on the compute node and accessing the rbd volume so any attempt to delete the rbd volume itself fails.
I can't see how that's anything other than a failure by the operator to ensure that the original compute node has been fenced before attempting a local delete of an active instance.
To be clear, the issue here is that the domain is still active on the compute node and accessing the rbd volume so any attempt to delete the rbd volume itself fails.
I can't see how that's anything other than a failure by the operator to ensure that the original compute node has been fenced before attempting a local delete of an active instance.