Description of problem: Bug #1898578 details a failure to attach a volume to an instance where n-cpu raises MessageUndeliverable after creating a BDM record. This results in n-api eventually raising MessagingTimeout and the overall request failing. This leaves a BDM record in the database suggesting that the volume is attached to the instance when in fact it is not. n-api should rollback and remove any BDM records created by n-cpu when it encounters MessagingTimeout. This issue was previously raised in bug #1752734 but assumed resolved by other fixes to oslo.messaging that landed later in Rocky. Version-Release number of selected component (if applicable): 17 / master How reproducible: Always Steps to Reproduce: 1. Cause MessagingTimeout to be raised from calls to reserve_block_device_name Actual results: BDM records created by n-cpu are not cleaned up by n-api. Expected results: BDM records created by n-cpu are cleaned up by n-api. Additional info: