On monday, returned home to find my computer wasn't responding, straight away I reckoned that something bad had happened. I was right - a hard drive had managed to get itself toaster. It was actuallly the partition table that had been lost. I am no stranger to partition table losses, it had happened to me before and I knew just what to do - unfortunately what worked last time around didn;'t this time.
Google was no help. The first result for 'linux corrupt partition table recovery' was my own page from last time around! Unfortunately gpart didn't work this time. It's a wonderful piece of software but it doesn't seem to be so good at detecting XFS partitions.
There were two partitions on the hard drive, one had reiserfs and the other had XFS. I could afford to lose the reiserfs partition - it was only a set of backups, the same couldn't be said about the xfs partition which was /var/ there I had my mysql databases, websites and subversion repositories.
Sure I had backups - but they were 12 days old. I did a comprehensive backup before moving back from Debian to Fedora and since then been negligent. Still I could afford to lose the subversion repository - all the upto date code is on another partition but it's a tedious process to checkout each branch, copy over the files from the working copy, commit it and then checkout again.
The local copies of the websites can simply be pulled down from the live site (and the repo updated again). I did lose about half a dozen unpublished blog entries but no harm done.
Another loss was the rpm database, I had to boot with the DVD and choose the upgrade option to recreate those databases. Even then rpm -qa doesn't show some of the packages that are in fact installed.
I used dd to copy the entire hard drive onto another and tried all sorts of things including parted and xfs_check no dice. My relationship with XFS lasted all of 12 days.