Fedora 8 to fedora 10 upgrade MDADM problems

[kenned3]

The other day i booted the F10 DVD and upgraded an existing 64 Bit fedora 8 installation.

after the reboot i fixed the NVIDIA drivers (expected) and tried to fix the MDADM problems (somewhat expected).

Here is the issue:
Every time i reboot the machine, the array randomly fails drives. Sometimes if fails 3 and refuses to start, other times it fails 1 and starts.

i can fix this by :

mdadm --stop /dev/md_d0
mdadm --stop /dev/md0
mdadm --create /dev/md0 --assume-clean --level=5 --raid-devices=5 /dev/sdb1 /dev/sdc1 /dev/sdd1 /dev/sde1 /dev/sdf1

vgscan
vgchange -ay

mount /dev/mapper/lvm--raid-multimedia /Multimedia/

Here is what i think the probem is.
sometimes it decides that one (or more) of the disks belongs to another raid unit with the same name.

IE, of the 5 drives, it says that 3 belong to "lvm-raid" while the other 2 belong to "lvm-raid" as well.

The missing drive which belongs in MD0 is what is in md_d0

Any thoughts?

[jvillain]

I just finished upgrading from FC9 to all the latest and greatest rawhide packages and got a similar problem. When booting it kept cratering when the volume manager tried to mount a logical volume located on a software raid 5 device. After I removed the partition from my fstab and rebooted I was able to determine that my raid 5 device was being created a /dev/md127 rather than /dev/md0. After stopping md127 and manually recreating it as md0 I could then mount my logical partition.

Doing some searching it seems that the initrd may be at fault. One post I read indicated that if your / partition can't be read for some reason then md will not be able to see /etc/mdadm.conf and read the proper device id out of it. That doesn't seem likely in my case as I can boot off of /.

I have also read some other stuff indication that your mdadm.conf should be in your initrd. Mine most certainly is not. You can check yours

mkdir junk
cd junk
zcat /boot/initrd-<version>.img | cpio -i *
ls etc


Then also check
cat init

And see if there are any references to your raid device in there. Again I don't have any. I tried doing a mkinitrd on my working system to see if that would help and it didn't. So there could be an issue with mkinitrd. It looks like the people using dmraid are having issues with mkinitrd. You can look at

https://bugzilla.redhat.com/show_bug.cgi?id=476818

for more info on that. It may also give you some clues of where to look for clues.

I am sorry I don't have a better answer for you.

[kenned3]

I too tried to build a new initrd, without success (same problem with the array).

One would assume (perhaps incorrectly) that mdadm would have been tested with F10, and the upgrade should have "just worked"?

be nice to get this fixed as its somewhat annoying having to restart it manually ever reboot.

[jvillain]

Have you opened a bug report? If not then I will.