RAID catch-22 on power5
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
partman-md (Ubuntu) |
Fix Released
|
Undecided
|
Colin Watson |
Bug Description
Binary package hint: partman-md
On a power5 system, I went partway through a 20070201.1 alternate install and built RAID on the drives. When I went back to the install with Herd 3 after experiencing problems with the daily, I found myself in a catch-22 in partman-md.
I was unable to use the md device, unable to remove it (although it prompted me to remove an raid5 volume that appeared to be the one I made earlier) as it complained of /dev/md/0 not existing. I was of course unable to make a new raid volume as it could not find any unused raid partitions.
What's worse, the retry/cancel cycle would silently die when I tried to cancel, leaving me with a progress bar at 47% and no interface to continue.
I tried deleting and recreating the partitions, rebooting in between, but had no luck.
This mostly fixes it. There's another long-established problem that still arises if you use LVM on RAID - you get an error message about not being able to inform the kernel of changes to /dev/md0p1 - but you can safely select "Ignore" at that error message and continue.
partman-md (30ubuntu1) feisty; urgency=low
* init.d/md-devices: Handle /dev/md* as well as /dev/md/* (LP: #84228).
* Set Maintainer to <email address hidden>.
-- Colin Watson <email address hidden> Tue, 27 Mar 2007 13:41:30 +0100