On Tue, May 15, 2012 at 01:50:40PM +0200, lee wrote:
> On 05/14/2012 05:34 PM, Akemi Yagi wrote:
> > On Mon, May 14, 2012 at 8:20 AM, Gareth de Vaux <[log in to unmask]> wrote:
> >> https://bugzilla.redhat.com/show_bug.cgi?id=772926


By misfortune, I had to look at the MD-RAID array setup
code in dracut, and I am sorry to report that it looks
like an unholy mess of shell scripts. Feels like looking at
historical SYSV or BSD code from the 1970-ies before the invention
of perl, etc. Cannot believe my eyes that somebody created that
stuff in the 2010-ies.

Instead of issuing "mdadm --scan --start --all" (or whatever
is the right incantation) and relying in mdadm.conf to be correct,
they try to be clever with parsing UUIDs from the kernel command
line and from mdadm.conf embedded in initramfs.

And as is the usual result, they get it right for the trivial
cases when a simple "mdadm start all" would have worked
and they get it wrong for the case when correct handling of UUIDs
actually does matter.

In my case, I have to boot from a raid array with a UUID different
from the UUID inside mdadm.conf inside the initramfs file. I specify
this other UUID on the kernel command line, but some contents
of mdadm.conf seem to confuse the dracut scripts and it fails to start
all raid arrays.


-- 
Konstantin Olchanski
Data Acquisition Systems: The Bytes Must Flow!
Email: olchansk-at-triumf-dot-ca
Snail mail: 4004 Wesbrook Mall, TRIUMF, Vancouver, B.C., V6T 2A3, Canada