[Yaird-devel] Bug#350861: yaird: Degraded RAID 1 array fails to boot

[Shaun Jackman]

Package: yaird
Version: 0.0.11-12
Severity: important

If yaird is run while the root partition, a RAID 1 array, is in a
degraded state, the system will not boot if the array is ever restored
to a complete state.

I installed a new kernel (linux-image-2.6.15-1-k7). I did not notice
at the time that /dev/md0, a RAID 1 array and my root partition, was
degraded; 1 of 2 disks were online. When yaird was run in the kernel
postinst, it generated an initrd that added only the one online disk
to the array. Some time later I noticed that the array was degraded
and add added the missing disk back to the array. Now 2 of 2 disk are
online. I rebooted the system, but the initrd only tried to add the
one disk to the array, and mdadm refused to start the array, since it
had only 1 disk, and apparently it really wanted 2. It pointed out
that one may pass the --run option to force it to start, but of course
I had no console or shell with which to pass this option.

So, it'd be better if the initrd built by yaird included *all* the
disks for the array, including failed disks. Perhaps it could check
this information against /etc/mdadm/mdadm.conf if it exists. Second,
it'd be much better if the initrd would start the array even with a
failed disk, which suggests passing the --run option to mdadm. This
latter point probably requires some discussion first.

Cheers,
Shaun