cinder shows wrong device name for attached volume

[Isaac Hailperin]

Using grizzly on centos 6.4, attached volumes are shown with wrong device names in cinder. E.g. I have this volume wich I attach, cinder tells me its attached as /dev/vdb:
 | attachments | [{u'device': u'/dev/vdb', u'server_id': u'3318b373-8792-4109-b65c-ee138dcd525f', u'id': u'c424719d-1031-4e29-8086-8a62b385d99a', u'volume_id': u'c424719d-1031-4e29-8086-8a62b385d99a'}] |

 but on the machine (ubuntu 12.04 cloud image) the device is /dev/vdc:

root@u15:~# cat /proc/partitions
major minor #blocks name

 253 0 2097152 vda
 253 1 2088450 vda1
 253 16 400 vdb

-> attaching volume to instance:
root@u15:~# cat /proc/partitions
major minor #blocks name

 253 0 2097152 vda
 253 1 2088450 vda1
 253 16 400 vdb
 253 32 1048576 vdc
 253 33 1047552 vdc1

Maybe this is not cinders fault, as it the volume seems to be the second attached storage device? could be due to ubuntus image, but still, the device name is wrong.

[John Griffith (john-griffith)]

So this comes up a fair bit the post from the mail list [1] describes it pretty well.

This is more of an issue with how the hypervisors perform the mount than anything else.

http://lists.openstack.org/pipermail/openstack-dev/2013-April/007208.html

[Isaac Hailperin (isaac-hailperin)]

I see.
Thanks for looking into this.

My humble suggestion would be to remove the device name from cinder, since it seems not possible to reliably predict the device name in the guest os.

[Andres Buraschi (andres-buraschi)]

Would you think that adding a comment in horizon and a warning when invoking the CLI is a valid first step to address this issue? This leads at least to check the attached device name instead of trusting the given name...

[Isaac Hailperin (isaac-hailperin)]

I think that would help.

[Andres Buraschi (andres-buraschi)]

Related bug:
https://bugs.launchpad.net/horizon/+bug/1255136

[Verónica Musso (veronica-a-musso)]

Partial fix for Horizon approved: https://review.openstack.org/#/c/60261/
Partial fix for Nova waiting for approval: https://review.openstack.org/#/c/60257/

[Sean Dague (sdague)]

This is not fixable for libvirt.