ZFS fun and games

In changing anything you run the risk of running into unseen problems and that’s just what happened when I attempted to upgrade the power supply in my main system. I was upgrading to a higher capacity PSU in order to cope with the 7 disk drives, a tape device, a DVD writer, a quad core AMD processor and an Nvidia GeForce GTX 260 graphics card. This was because every time I tried to kick off a backup the old PSU would reset the system due to the unrated power supply and obviously I wanted to back up as much of the system as I could in case of issues. So, the initial PSU replacement to an unbranded 800 Watt PSU from Ebuyer only took 15 minutes to replace once I’d sorted out all the cables,plugs etc but the noted strangeness when OpenSolaris booted. It complained about OHCI usb errors (and stopped the keyboard/mouse from working) and ATA disk errors (which happened to be my mirrored ZFS boot device). Just in case it was a configuration error I tried booting the live OpenSolaris cdrom and got the same thing….strange. So thinking something was really wrong, I replaced the original power supply and after which I didn’t see the errors any more booting from live cdrom but had managed to corrupt my ZFS rpool….bugger. So searching on the net I noticed that lots of other folks have run into similar corruption issues and they’ve managed (by some assistance from some ZFS gurus) to get the pool back to read only mode and recover the data. There is also some mention of CR 6667683 – “need a way to rollback to an uberblock from a previous txg” which when implemented will allow the admin to rollback to a previous transaction….but not implemented until build 125 and I’m on build 111b…..bummer. After some shared shell access and some jiggery pokery by Victor and Jason they managed to get my rpool back to read only (Thanks Chaps!)….whew. Once I’d copied the data to my working pools, I re-installed OpenSolaris and then copied my root rpool data back, created a new beadm boot environment and managed to get back to where I was prior to the corruption….yay. Having spoken to another colleague at work, I’m going to try a well known PSU and see what happens…but not first without detaching a mirror or first boot from cdrom to confirm that all is well….. 🙂