From: Martin F. Foster (martin_foster@pacific.net.au)
Date: Fri Aug 06 2004 - 05:42:39 EDT
Hello List,
There have been various postings in the past regarding strange problems with
Jumpstart + SVM mirror installs.
Typically, these installs where characterized by success through the
Jumpstart installation, but panics when rebooting with messages along the
lines of:
Cannot mount root on /pseudo/md@0:0,0,blk fstype ufs
...
with the system entering a continuous boot -> error -> reboot cycle.
I think that I have a solution to this problem. It is being shared with the
list for archival reasons, and in the hopes that someone could point out a
problem with the Jumpstart configuration that causes this. The problem is
present up Solaris 9 09/04.
Taking a closer look at an installed but failing system, I noticed that
/etc/lvm/mddb.cf and /kernel/drv/md.conf were devoid of metadb information.
The system requires this to boot SVM volumes properly, resulting in the
endless loop of panics & reboots that you saw.
Reading up on the Solaris docs, mddb.cf and md.conf are populated by the
metadb command that you run at the start of a manual install & configuration
of SVM. My solution was to add the following to my finish script:
<snippet of finish_script>
BASE=/a
METADB=c0t1d0s6
# Fix raid configuration. System doesn't properly put device info in
/etc/system, /etc/lvm/mddb.cf, /kernel/drv/md.conf
# -> ensure that METADB points to a database replica, created
# from the jumpstart profile
# -> it's important that the profile create at least two databases on
# two separate slices
# ex:
#metadb c0t0d0s6 size 8192 count 2
#metadb c0t1d0s6 size 8192 count 2
# -> for the example, use METADB=c0t1d0s6
/usr/sbin/chroot ${BASE} /usr/sbin/metadb -d /dev/dsk/${METADB}
/usr/sbin/chroot ${BASE} /usr/sbin/metadb -a -k /etc/system -c 2 ${METADB}
/usr/sbin/chroot ${BASE} /usr/sbin/metadb -p -k /etc/system /etc/lvm/mddb.cf
</snippet>
with corresponding Jumpstart profile...
<listing profile>
# profile keywords profile values
# ---------------- --------------
install_type initial_install
system_type standalone
metadb c0t0d0s6 size 8192 count 2
metadb c0t1d0s6 size 8192 count 2
partitioning explicit
filesys mirror:d10 c0t1d0s0 c0t0d0s0 6144 / logging
filesys mirror:d20 c0t1d0s1 c0t0d0s1 2048 swap
cluster SUNWCprog
</listing>
With metadb run as above, the system boots normally after installation.
Note that the 09/04 release has cleared the problem with using the 'free'
keyword for filesys mirror partions, which is nice.
There is still a very real possibility that my jumpstart configuration is
causing this problem. I welcome any solution that avoids a finish script
hack.
Cheers,
Martin Foster
Systems Engineer
Pacific Internet (Australia) Pty Ltd
Direct: +61 3 9674 7659
Mobile: +61 416 084 325
Fax: +61 3 9698 4959
Email: martin_foster@pacific.net.au
www.pacific.net.au
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 23:29:13 EDT