Adaptec HostRAID ddf1 on Supermicro X7DBR-i+
Andrew Peabody <apeabody <at> systemsbiology.org>
2007-12-18 20:13:57 GMT
First I want to say that dmraid was been a wonderful tool for use. We use
it with many Sun X2100s, and other motherboards with great success. The
only feature we really miss is an option like mdmonitor to notify us if
there is a problem with an array.
However, I have several identical Supermicro X7DBR-i+ (BIOS 2.0)
motherboards that seem to have issues with dmraid. I'm not sure if it's a
problem with the board, dmraid, or the combination. The motherboard
provides with option for either an Intel or Adaptec code base. We had
exceptional troubles (corruption, or drives just disappearing, normally
unable to complete install) withe Intel code base and decided to try the
Adaptec option instead. It appeared to work, but on closer inspection
after install it's still not working right. With 1.0.0.rc13 the array
appears to be recognized by dmraid, but never activated through the device
mapper, instead the operating system was mounting the underlying physical
device based on it's label. I noticed that it was using the ddf1 "driver"
for the array and looking through the changelog and noticed that there was
a fix for some ddf1/Adaptec problems in rc14 so I upgraded to that and
rebuild initrd, etc. However, while I appear to be able to successfully
manually start the array, it appears to not work within the sysinit/initrd
default procedure. While "/sbin/dmraid.static -ay -i -p -t" grep, etc
does return the array name eg: ddf1_Volume0 the later command to activates
fails with a seg fault eg: /sbin/dmraid.static -ay -i -p "ddf1_Volume0".
However without specifying an arrayname it appears to activate properly.
It looks like the initrd was written for dmraid using the "-p" arrayname
to utilize kpartx to create the partition device maps with the "p"
deliniator. I would guess that if one could specify the partition
deliniator then kpartx and the might not be needed? Any thoughts on how