I''ve got an LDOM that has raw disks exported through redundant service domains from a J4400. Actually, I''ve got seven of these LDOM''s. On Friday, we had to power the rack down for UPS maintenance. We gracefully shutdown all the Solaris instances, waited about 15 min, then powered down the storage. All of the LDOM''s came up except one. They each have a mirrored RPOOL on 2 of these drives. The one with the problem shows the rpool as a mirror, with both devices online but the pool''s unavailable. There''s a message at the bottom that says: "Additional devices are known to be part of this pool, though their exact configuration cannot be determined." Only two disks were EVER part of this LDOM, and it''s an rpool, so no raidz or stripe even possible. Anyone know how I can get past this? -- This message posted from opensolaris.org
Hi John, The message below is a ZFS message, but its not enough to figure out what is going on in an LDOM environment. I don''t know of any LDOMs experts that hang out on this list so you might post this on the ldoms-discuss list, if only to get some more troubleshooting data. I think you are saying that the LDOM that is not coming up has a mirrored root pool on 2 devices... You might be able to use the following command to see if the disk labels are coherent for each these devices for the broken LDOM''s root pool: # zdb -l /dev/dsk/xxxxx Maybe someone else has some better ideas ... Cindy On 01/19/10 07:48, John wrote:> I''ve got an LDOM that has raw disks exported through redundant service domains from a J4400. Actually, I''ve got seven of these LDOM''s. On Friday, we had to power the rack down for UPS maintenance. We gracefully shutdown all the Solaris instances, waited about 15 min, then powered down the storage. > > All of the LDOM''s came up except one. They each have a mirrored RPOOL on 2 of these drives. The one with the problem shows the rpool as a mirror, with both devices online but the pool''s unavailable. There''s a message at the bottom that says: > > "Additional devices are known to be part of this pool, though their > exact configuration cannot be determined." > > Only two disks were EVER part of this LDOM, and it''s an rpool, so no raidz or stripe even possible. Anyone know how I can get past this?
I was able to solve it, but it actually worried me more than anything. Basically, I had created the second pool using the mirror as a primary device. So three disks but two full disk root mirrors. Shouldn''t zpool have detected an active pool and prevented this? The other LDOM was claiming a corrupted device, which I was able to replace and clear easily. But the one pool I originally posted about looks to be permanently gone, since it believes there is another device, but doesn''t know where the device is or what it was ever called. If I could import it and re-do the mirror somehow, or something similar, it''d be great. Is there anyway to force it to realize it''s wrong? Obviously, I should''ve kept better track of the WWN''s - But I''ve made the mistake before and zpool always prevented it. -- This message posted from opensolaris.org
Hi John, In general, ZFS will warn you when you attempt to add a device that is already part of an existing pool. One exception is when the system is being re-installed. I''d like to see the set of steps that led to the notification failure. Thanks, Cindy On 01/19/10 20:58, John wrote:> I was able to solve it, but it actually worried me more than anything. > > Basically, I had created the second pool using the mirror as a primary device. So three disks but two full disk root mirrors. > > Shouldn''t zpool have detected an active pool and prevented this? The other LDOM was claiming a corrupted device, which I was able to replace and clear easily. But the one pool I originally posted about looks to be permanently gone, since it believes there is another device, but doesn''t know where the device is or what it was ever called. If I could import it and re-do the mirror somehow, or something similar, it''d be great. Is there anyway to force it to realize it''s wrong? > > Obviously, I should''ve kept better track of the WWN''s - But I''ve made the mistake before and zpool always prevented it.
John wrote:> I was able to solve it, but it actually worried me more than anything. > > Basically, I had created the second pool using the mirror as a primary device. So three disks but two full disk root mirrors. > > Shouldn''t zpool have detected an active pool and prevented this? The other LDOM was claiming a corrupted device, which I was able to replace and clear easily. But the one pool I originally posted about looks to be permanently gone, since it believes there is another device, but doesn''t know where the device is or what it was ever called. If I could import it and re-do the mirror somehow, or something similar, it''d be great. Is there anyway to force it to realize it''s wrong?You can try limiting access to one device at a time by removing one device from LDOM configuration, or creating separate directory like /tmp/dsk and copying symlink for the device you want to try there and trying to do zpool import (if device is removed at the LDOM level) zpool import -d /tmp/dsk (in case you prefer trick with symlinks) Posting label 0 (from zdb -l /dev/rdsk/... output) of both involved disks may provide more clues. regards, victor> > Obviously, I should''ve kept better track of the WWN''s - But I''ve made the mistake before and zpool always prevented it.