yahoo-eng-team team mailing list archive
-
yahoo-eng-team team
-
Mailing list archive
-
Message #15370
[Bug 1327497] [NEW] live-migration fails when FC multipath is used
Public bug reported:
I tried live-migration against VM with multipath access to FC bootable volume and FC data volume.
After checking the code, I found the reason is that
1. /dev/dm-<NUM> is used, which is subject to change in the destination Compute Node since it is not unique across nodes
2. multipath_id in connnection_info is not maintained properly and may be lost during connection refreshing
The fix would be
1. Like iSCSI multipath, use /dev/mapper/<multipath_id> instead of /dev/dm-<NUM>
2. Since multipath_id is unique for a volume no matter where it is attached, add logic to preserve this information.
** Affects: nova
Importance: Undecided
Assignee: Jeegn Chen (jeegn-chen)
Status: In Progress
** Changed in: nova
Assignee: (unassigned) => Jeegn Chen (jeegn-chen)
** Changed in: nova
Status: New => In Progress
** Description changed:
I tried live-migration against VM with multipath access to FC bootable volume and FC data volume.
After checking the code, I found the reason is that
1. /dev/dm-<NUM> is used, which is subject to change in the destination Compute Node since it is not unique across nodes
2. multipath_id in connnection_info is not maintained properly and may be lost during connection refreshing
The fix would be
1. Like iSCSI multipath, use /dev/mapper/<multipath_id> instead of /dev/dm-<NUM>
- 2. Since multipath_id is unique for a volume no matter where is is attached, add logic to preserve this information.
+ 2. Since multipath_id is unique for a volume no matter where it is attached, add logic to preserve this information.
--
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to OpenStack Compute (nova).
https://bugs.launchpad.net/bugs/1327497
Title:
live-migration fails when FC multipath is used
Status in OpenStack Compute (Nova):
In Progress
Bug description:
I tried live-migration against VM with multipath access to FC bootable volume and FC data volume.
After checking the code, I found the reason is that
1. /dev/dm-<NUM> is used, which is subject to change in the destination Compute Node since it is not unique across nodes
2. multipath_id in connnection_info is not maintained properly and may be lost during connection refreshing
The fix would be
1. Like iSCSI multipath, use /dev/mapper/<multipath_id> instead of /dev/dm-<NUM>
2. Since multipath_id is unique for a volume no matter where it is attached, add logic to preserve this information.
To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1327497/+subscriptions
Follow ups
References