From: Gloria Chung (gchung@somanetworks.com)
Date: Tue Aug 01 2006 - 15:04:24 EDT
Sorry for the delay in posting the summary. I've been busy testing out
the solution. Thank you all who have responded (and esp. those who have
responded a second time after I wrote back for more questions=).
My original post is attached at the end of the message. My problem, as
many have told me correctly, is that I shouldn't have ufs running on the
same slice where the metadb lives. Once I stopped mounting ufs on the
metadb slice (slice3 in this case), breaking the mirror by metadetach
and then rebooting doesn't give me the "bad super block" error anymore.
Some people have asked me why I'm editing vfstab to reference real
device names. My real project is to upgrade Solaris8 with disksuite
running. According to the Disksuite4.2.1 Reference on the Sun website
(AppendixB), I have to first break the mirrors and then mount the slices
with real device name. Actually I have followed the instructions and
got the upgrade working fine, *when I have two disks inside the server
at all times*. The problem surfaced when I removed the second disk
during the upgrade (in order to preserve it as a backup).
On another note, the DiskSuite Reference also states that it is possible
to store metadb on the same slice as a metadevice. I would assume a
metadevice usually requires ufs to be running on the slice. Which
implies that ufs can be running on the slice where metadb is stored. I
think it can be done (since this is the way our system is configured
before the Solaris upgrade, and it has been working well for a long
time). But probably be best to be done by someone who really knows
Disksuite.
Thanks again for all the help.
Gloria
Here's my original post:
I am trying to accomplish the following, provided I start with two disks
in sync:
1. detach the submirrors on the second disk
2. remove the second disk physically
3. boot up the system again with only first disk functioning
4. later insert disk2 and resync
For step 1, I ran metadetach on all submirrors on disk2, ran metaroot to
reference the disk1 real device name (not mirror name), edited the
vfstab to reference real device names. But when I boot up (even without
removing disk2), I always get the "bad super block" error complaining
about the slice on disk1 where the database replica resides.
I have database replicas on both disk1 and disk2. Should I delete the
disk2 database replicas before booting up disk1? If so, should I do
that before or after I detach the submirrors?
_______________________________________________
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 - 23:40:29 EDT