bare-metal restore to diff h/w and h/w db

From: Mike Broderick (broderic@MIT.EDU)
Date: Tue Feb 04 2003 - 23:52:53 EST


I need to develop a bare metal restore procedure for disaster recovery.
The new h/w may not be identical to the old. I boot the install CD and
restore / and /usr from tape then boot the restored O/S filesets in
single user mode and with genvmunix with plans to touch up device names
in fstab and fdmns and rebuild the kernel. But when the system boots, I
cannot access much of the devices. I get negative id values for many
devices in the h/w db...

# hwmgr v d
 HWID: Device Name Mfg Model Location
 ------------------------------------------------------------------------------
    3: (unknown)
    4: (unknown) (unknown) (unknown)
    5: (unknown)
   29: (unknown) 3.5in floppy fdi0-unit-0
  -73: (unknown) QUANTUM DLT7000 bus-2-targ-4-lun-5
  -71: (unknown) QUANTUM DLT7000 bus-2-targ-4-lun-2
  -69: (unknown) QUANTUM DLT7000 bus-2-targ-3-lun-5
  -67: (unknown) QUANTUM DLT7000 bus-2-targ-3-lun-2
  -65: (unknown) QUANTUM DLT7000 bus-2-targ-2-lun-5
  -63: (unknown) QUANTUM DLT7000 bus-2-targ-2-lun-2
  -61: (unknown) QUANTUM DLT7000 bus-2-targ-1-lun-5
  -59: (unknown) QUANTUM DLT7000 bus-2-targ-1-lun-2
  -57: (unknown) QUANTUM DLT7000 bus-2-targ-0-lun-5
  -55: (unknown) QUANTUM DLT7000 bus-2-targ-0-lun-2
  -53: (unknown) QUANTUM DLT8000 bus-1-targ-5-lun-0
  -51: (unknown) COMPAQ CD-224E bus-3-targ-0-lun-0
  -49: (unknown) COMPAQ BD009635C3 bus-0-targ-0-lun-0
  -26: (unknown) SWXCR ctlr-0-unit-2
  -24: (unknown) SWXCR ctlr-0-unit-1
  -22: (unknown) SWXCR ctlr-0-unit-0
  -20: (unknown) SWXCR xcr0

I can get devices created for the first few devices w/ positive ids
using '/sbin/dn_setup -init' and/or 'dsfmgr -K' but not for the negative
ids. I then tried to clean things up w/ this...

# hwmgr refresh comp
hwmgr: Error (13) Permission denied.
# hwmgr refresh scsi
hwmgr: Refresh SCSI database operation was successful
#

But I still have the same negative id values and unknown devices.

Is there some brute force way to wipe out the entire h/w db and recreate
it once I've booted the retored O/S filesets

                                                   _Mike

P.S. - FWIW, My ultimate task is to devise a disaster recovery
bare-metal restore process for Netbackup backed-up systems. We have the
O/S backed up as a single stream (3 big NB tar (which is a modified GNU
tar) tarballs (/,/usr,/var) on one tape. When I try to just boot the
O/S CD, mount the disks on /var and restore these tarballs to the
mounted filesets all the hard links don't restore as they try to restore
to absolute paths which crosses devices .(E.g., a hard link from 'a' and
'b' both in /etc retores as a hark link from /var/mnt/root/etc/a to
/etc/b which fails because /etc/b is on the CD.) So I'm trying to
devise another not too lenghty process to accomplish this. I tried doing
a simple clean O/S install and [v]dump that to tape to be [v]restored
whereever. But I get the above problems (negative ids) presumably since
the O/S install was [intentionally] done on different h/w. I wonder if
I may just end up having to do a O/S install from CD on the destination
system at recovery time, then restore the NB tarballs on top of that.
 Anyone else dealt with this scenario before and can offer tips?



This archive was generated by hypermail 2.1.7 : Sat Apr 12 2008 - 10:49:06 EDT