ZFS Not Borked
A while back I posted an article where I thought the gigantic data volume on our Sun x4500 server was corrupted or lost because of a faulty disk. I tried a bunch of troubleshooting to no avail. I couldn’t get the system to boot up and mount the 20 TB zfs data volume cleanly. I was at the point where I had written off the data on that volume.
I left this problem alone for a long while, having bigger fish to fry in the mean time. Yesterday I got back to it, and decided to just nuke the system and start from fresh. I installed the latest build of Solaris 10 on it, which because it’s Sun hardware, went very smoothly. Then, once I had the boot environment configured I tried importing the zfs volume that I thought was screwed, and it imported cleanly. The data is all there, and the volume mounts automatically when I reboot it. I’m doing a pool scrub to validate the data, but it seems to be all available.
I decided to try this because I had a similar problem on another machine with a much smaller zpool. The zpool stopped mounting and started acting really weird after I destroyed a zvol that was shared via iscsi. Doing a clean install of Solaris fixed the problem and the zpool is still ok.
I’m relieved to get the data back, because there was about 9 TB of useful stuff on there (mostly backups of production data from the last year or so).