Bug#361408: Why it works on 2.6.8
Shawn Willden
shawn at willden.org
Mon May 22 05:19:53 UTC 2006
It appears the scripts in the initrd image for 2.6.8 don't attempt to start
the md devices. I'm not sure how 2.6.8 manages to mount /dev/md0 on root,
but it does, and then the rest of the md arrays remain unstarted until
the /etc/init.d/mdadm-raid script runs, which happens after the root
partition is mounted and the regular system init scripts are under way. That
script uses mdadm, not mdrun, and starts all of the arrays correctly.
On 2.6.15, my initrd contains scripts that attempt to start the arrays, which
fits with my observations from watching the boot process. What I see is that
the md devices are started much earlier with 2.6.15, and then I see the
output of running the mdadm-raid script later at the normal point in the
process. Well, that's what I see on the occasions that root can be mounted,
anyway.
Tomorrow I'll figure out how to remove the md startup from the initrd scripts
for 2.6.15. I suspect that will make the problem go away (though not really
fix it).
Shawn.
More information about the pkg-mdadm-devel
mailing list