My original posting:
For three very similar Sparc 10 Solaris 2.4 NFS servers each running
ODS 4.0 mirroring on eight external drives, I have tried to build a
backup system disk to be used in case any one of the system disks
fails. All metadevice state database replicas exist on the eight
external drives so that, I thought, as long as /etc/nodename,
/etc/hosts, /etc/hostname.fddi0, /etc/system, /kernel/drv/md.conf,
/etc/opt/SUNWmd and /dev/md on the backup disk were the same as on
a failed system drive, the backup disk should run in place of the
failed one.
In testing, however, it booted complaining "metainit: venus: there
are no existing databases."
Many thanks go to
Joe Dietz <jdietz@mis.uswest.com>
lorenzd@gcm.com (Daniel Lorenzini)
AdamBracewell <bracea1@gatekeeper.research.natpower.co.uk>
rtrzaska@uk.mdis.com (Ray Trzaska)
Bruce Parker <bruce@sammy.pls.ov.com>
Philip.Kao@artecon.com (Philip Kao)
It was Daniel Lorenzini who suggested to do a metadb operation again
just to see what other files would get updated. In doing so I found
that the device files /dev/rdsk/c[12]t[0-3]d0s0, where my 8 db replicas
reside, got written and the minor device numbers of those devices were
different from machine to machine. Unfortunately the backup disk still
would not work after I copied those device files using cpio and a reboot.
My guess is that metadb refuses when it sees wrong checksums in mddb.cf.
I can change the minor device numbers in /etc/system and mddb.cf but
can do nothing to the checksums generated by "metadb -a -f." My
solution, therefore, is limited to running "metadb -a -f" and
"metainit -a" on the replacement drive when the time comes. It is not
ideal because it causes recyncing of the mirrored drives but it works.
Anchi
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:10:32 CDT