mdadm woes

[dk06]

No problem

I don't know why I thought creating this in 'general help' would have been better, but this is the snippets from my nfs_repair output.

FYI, I tried both the xfsprogs out in the repositories (v2 I think) and the latest and greatest (v3.0.1).

The guys over in the linux forums are pretty sharp when it comes to server/raid issues, maybe give them a shot: http://www.linuxforums.org/

& maybe post in Debian as well as Ubuntu,
I think there are more Debian server admins in general than Ubuntu so they may be able to offer more help
Good Luck

[datarhythm]

OK. So I think I know what happened (to the XFS that is).

Remember when I said:

"this is probably going to be one of those moments in my life that I think back and say 'what the hell was I thinking?'"

Well, I pretty sure I was spot on on my prediction. You see, when you build a RAID5 array, and you specify the order of the devices in which you build the array (/dev/sdb1, /dev/sdc1, etc), the ORDER is VERY IMPORTANT. So when I had did mdadm -C (blah blah), I should have put the exact order in which my original output was. If you can scroll back on that specific command I submitted, you'll see that it wasn't in the correct order.

So it wasn't that XFS got corrupted, I had just re-arranged the bits around such that XFS said "WTF?"

Story of my life.

The moral of the story, kids: Copy down the order in which you create a RAID5 array. At least I'm pretty sure that's what happened.

Needless to say, my data is gone. I can't 'rearrange' again to the proper order--already too late. And I don't have $2,500 lying around to get it repaired.

[tgrimley]

I should have put the exact order in which my original output was. If you can scroll back on that specific command I submitted, you'll see that it wasn't in the correct order.

Will just confirm this. When you assemble it doesn't matter (there is a superblock that tells mdadm what order) but when you recreate it must be in the same order as you originally created the array.

see http://osdir.com/ml/linux-raid/2009-06/msg00165.html for additional confirmation. glad your data is safe. (will remember this for when my raid array dies)