The issue occurs when some critical problems happened, for example multiple attach/detach/create volume requests to cinder-api and cinder-volume would leads to slower response of cinder service. To reproduce the problem I could suggest to stop cinder-volume service for couple of seconds during detaching. After nova-compute throws exception, the cinder-volume service could be started and finally the detaching procedure won't be retried by nova-compute.
Hello Vinay,
The issue occurs when some critical problems happened, for example multiple attach/ detach/ create volume requests to cinder-api and cinder-volume would leads to slower response of cinder service. To reproduce the problem I could suggest to stop cinder-volume service for couple of seconds during detaching. After nova-compute throws exception, the cinder-volume service could be started and finally the detaching procedure won't be retried by nova-compute.