[lug] RAID0 on root solved

D. Stimits stimits at comcast.net
Sun Nov 21 22:23:34 MST 2004

Well it turns out I was doing everything right for the RAID setup, 
though I did manage to lose about a week of email when I copied my 
backup week-old email to the original instead of the other way around :P 
Aside from the email blunder, all is up and running fine with a 
non-custom initrd and RAID0 on root, kernel 2.6.9, Fedora Core 2.

Apparently there is another chicken-and-egg problem with SCSI and RAID 
on root. I don't know if this problem has to exist or if it is just 
another SNAFU, I suspect it is a SNAFU. The bottom line is that to boot 
root with RAID0 SCSI drivers must be compiled in. I experimented with 
manually passing md0 parameters to the kernel via grub options, and it 
was complaining about not being able to find the individual partitions 
that made up the md0 device...despite having the grub itself run on SCSI 
and reading SCSI there just fine. Someone had mentioned the stage1 boot 
loaded was all that was running here at this stage of boot, and that it 
wasn't the kernel that was running the SCSI...this was correct. 
Unfortunately, RAID autodetect runs *prior* to loading initrd. Manually 
passing parameters caused it to generate useful messages that were not 
there without manual parameter passing. Now if RAID/MD autodetect code 
would occur *after* loading initrd, all would be fine for RAID root on 
SCSI, but because the order is backwards, it fails unless SCSI is 
compiled statically into the kernel.

Now I am still curious (I lost the email), someone had said they'd put 
raidstart manually into the initrd. It might still be useful for rescue 
purposes to be able to do this, and I am wondering where the 
statically-linked raidstart is from?

D. Stimits, stimits AT comcast DOT net

More information about the LUG mailing list