Cinder volume delete error with rbd backend in concurrent mode

asked 2014-12-01 19:07:44 -0500

sanjeo gravatar image

Hi all, We use Openstack + Ceph, and find that cinder-volume worker thread will be killed by the signal SIGABRT when executing the librbd rbd_delete fuction when deleting volume in concurent mode. Then the cinder-volume parent progress will re-spawn the worker thread, but the rbd driver will return ImageBusy exception and the volume's status will be turned from deleting to available.

I just wonder where the SIGABRT signal from and why?

thanks. -Jun

edit retag flag offensive close merge delete