Description of problem: Activating MD arrays has now been moved to userspace. The mkinitrd script inserts mdadm commands to activate arrays needed for boot: mdadm -As --auto=yes --run /dev/mdX On my test systems (MD RAID1 with two disks) this fails to start the array with the following error: mdadm: /dev/md1 not identified in config file. If I manually change this to either a UUID based assemble or a /proc/partitions assemble, e.g.: mdadm -Ac partitions -m dev --run /dev/md1 mdadm -A /dev/md1 --uuid=0162440e:3979df92:588758b2:ac2e4403 Then everything works fine & the array starts up. I'm not really sure mkinitrd is the right component to file this against; from what I could see in the mdadm docs, the assemble/scan/run combination should work but I was unable to start the arrays manually using these options in rescue mode either. Version-Release number of selected component (if applicable): mkinitrd-6.0.28-4.fc9.x86_64 How reproducible: 100% Steps to Reproduce: 1. Install a system with root on an MD device (I used LVM on top, but shouldn't matter) 2. Reboot Actual results: Boot fails with "mdadm: /dev/mdX not identified in config file." Expected results: Array starts. Boot succeeds. World domination achieved.
Edit summary for easier searching.
Same problem. "mdadm: /dev/md1 not identified in config file." Added md1 to the mdadm.conf in the initrd file and it booted fine. It seems Anaconda isn't building the mdadm.conf file properly. I say this because the mdadm.conf file says that it was "written out by anaconda".
I'd agree with comment #2 - this seems like more of an anaconda problem (it certainly appears like mkinitrd was written with the intent of copying over a working mdadm.conf from /etc, left there at install time). Switching component.
We do a better job in rawhide. Anaconda now uses the mdadm command to create the mdadm.conf file. It appends the "wrote by anaconda" to show that it was created at install time. reopen this bug if this is still an issue in f9.
Thanks Joel - I'll have some time this week or next to re-test on the box I originally hit this on. Will test with f9 final and re-open if there's still a problem.