tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: RAIDframe nested autoconfiguration



On Mon, 23 Mar 2009, Jed Davis wrote:

Hmm.  raid0 = sd0e, sd1e, missing; raid1 = sd2e, sd3e, raid0e; can
raid1e then fill out the missing member of raid0?

No, my changes don't try to fill in missing components of a RAID set
that's already been brought up; instead, I put off configuring any
degraded RAID sets until bringing up complete RAID sets can make no more
progress.

This, incidentally, means that in cases like the above (assuming the two
sets are both RAID-5), both raid0 and raid1 will be brought up in
degraded mode, and the administrator will have to intervene once the
system is up; but, if the "missing" disk in raid0 were present, then
everything would work.  (I considered having the kernel try to guess
which set to configure first in cases like that, but it seemed a bad
idea.)  This is also the cause of the mirror-splitting problem I've
described in a separate message.

        Is the order in which 'presumed degraded' disks are configured
        deterministic? For example, in ascending id order?

--
                David/absolute       -- www.NetBSD.org: No hype required --


Home | Main Index | Thread Index | Old Index