More hardware fun

In a move that would give the honourable Will Woods paroxysms, today I upgraded my HTPC from Mandriva 2009 to Mandriva 2009 Spring. Live, with urpmi. Over an ssh link. While the media center it runs (Freevo) was still running in the foreground. ...playing a video. Well, the upgrade happened, the video kept playing, and Freevo continued to work happily after the upgrade was done and the video had finished. Try THAT with Windows. (I _did_ lose the remote control until I rebooted...inexcusable! Who's the lirc maintainer again? Oh, er, right - me.) On reboot, though, my RAID array had disappeared. With help from the awesome zcat in Fedora IRC, we determined that it was something to do with RAID-related changes in udev - it had somehow created one array device for each drive that was part of the real array, and these arrays were in some sense running (though obviously they didn't work for anything). mdadm -S /dev/md_d12[5-7] - to stop those odd arrays - then re-starting the real one turned out to be the trick. Never would have figured that one out myself in a month of Sundays. However, the real array came up degraded, with one disk being reconstructed from scratch. I'd already noticed some suspicious errors relating to that disk in the logs, so it got the old heave-ho, I ran down to Crystal Mall to pick up a new one (500GB for $65...sheesh, we're living in the future. Hell, I bought a 16GB micro SD card for $60 while I was there. 16GB. It's smaller than my frickin' thumbnail...), threw it in there, and now the array's building that drive into itself. That'll take all night. Ah, well - at least I didn't lose my data, and that was the point of the whole RAID-5 array exercise in the first place. So that's done its job. (the failed drive was a Seagate 7200.11, for anyone keeping track. Yes, of the OEM variety. And yes, I bought it on May 3rd 2008, or just over one year ago, or just after the one year shop warranty on OEM drives expired. Sigh. Replaced it with a WD, since that's what the closest shop had on hand.)

Comments

jkeller wrote on 2009-06-01 13:21:
Not that it would help much now, but Seagate drives have a 5-year manufacturer's warranty on them. You might still be able to come out with an extra drive. And don't RAID arrays come up as degraded on every reboot? I seem to remember reading that you have to manually tell it to save meta data on the array itself (by default, it's in memory?) or some such. I'm interested myself, since I'm interested in making a RAID 5 array...
steletch wrote on 2009-06-04 14:33:
Hello Adam, happy to see you had the same troubles as me :-) Could you detail a little bit more what you did, because i presume this is not only a mdadm command (and all my drives are on the RAID, i assume yours are not). JKeller is also right, you should probably get a refund for your drive on seagate web site (even if the vendor is not saying it, the drive *is* still guaranteed, but only from the website). Could you post the results at https://qa.mandriva.com/show_bug.cgi?id=50709 please Last, we need to register now?
adamw wrote on 2009-06-04 18:07:
you always had to register for my blog... really removing the spurious, udev-generated 'arrays' was all I had to do to get everything to a base state, then just re-starting the real array as normal worked fine. Do you have suspiciously named nodes like I did - /dev/md(something) ?