On Mon, Mar 28, 2011 at 1:17 PM, Mark Morley <[email protected]> wrote: > Hi there, > > I have a small backup server (8.2-STABLE). It boots from ufs and has a zfs > pool for backups that consists of 8 drives configured as 4 mirrored devices, > totally around 2.5 TB. > > Been working great, no issues, until the past few days when remote rsyncs to > it have started to get very slow (it's only at around %50 capacity). > Rebooting it helps for a while, then it gets slow again. But this isn't the > problem now... > > After the last reboot, it froze while booting right at the point where the > file system gets mounted. No errors, it just doesn't proceed past the ZFS > version message. > > I rebooted single user and tried to access it with "zpool status", and the > command hangs in the same way. Any attempt to access it ("zfs list", for > example) does the same thing. > > The disks themselves seem fine. They are all connected to a pair of Adaptec > RAID controllers (configured as individual drives, with mirroring handled by > zfs) and the controller software shows them all to be intact. > > I disabled zfs in rc.conf and was able to boot, but I can't access the pool. > > Any ideas on how to diagnose and hopefully repair this? >
Your going to need to download a recent -CURRENT ISO that contans zfs v28, then you can try to recover the pool as outlined in this post http://opensolaris.org/jive/message.jspa?messageID=445269 zpool import -nfF -R /mnt rpool If it reports that it can get back to good pool state, then do actual import with zpool import -fF -R /mnt rpool In case first command cannot rewind to older state, try to add -X option: zpool import -nfFX -R /mnt rpool and if it says that it can recover your pool with some data loss and you are ok with it, then do actual import zpool import -fFX -R /mnt rpool Note: I haven't given this a try on my system yet, let me know how it goes ;-). Scot _______________________________________________ [email protected] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to "[email protected]"
