Pat Wilson wrote:
> Hmm.  It's been my naive understanding all this time that if
> I slapped an OS and the ADSM client on a unix box, I could
> restore the whole thing - now it looks like that's not the
> case (or at least it's not that simple).  ADSM won't restore
> over open files, apparently (not that I really blame it), among
> other things.
>
> So - how _is_ everyone providing DRM on Unix machines?  The straightforward
> approach would be to dump / and /usr locally, but I'm trying to get
> _away_ from having to do local dumps and having local tape drives - that's
> the whole point of investing in ADSM.  This is a heterogenous shop, so
> one vendor's solution (NIM and mksysb for AIX, for example) isn't going to
> work for everything...

I am currently dealing with setting up bare metal recovery plans for a mixed
population of AIX and HP-UX systems (with some Sun systems likely to be added
in the future). I am considering the following approach for HP-UX:

1.Install HP-UX and ADSM client from CD-ROM.
2.Use pvcreate -B to create a potentially bootable volume.
3.Create a new volume group on it.
4.Run mkboot to install boot utilities.
5.Create logical volumes with the same size as selected file systems on the
  destroyed system, taking care to get the ones corresponding to /stand, the
  primary swap area, and / contiguous and in the right order.
6.Mount the new logical volumes corresponding to file systems with names
  other than the normal ones.
7.Use ADSM to restore files to the new logical volumes.
8.Execute lvlnroot commands to record the locations of the logical volumes
  intended for /stand, the primary swap area, and /.
9.Replace the contents of the new logical volumes' counterpart of /dev with
  the contents of /dev.
10.Do the same for the LVM configuration backups in /etc/lvmconf.
11.Copy /etc/lvmtab to the corresponding location in the new logical volumes.
12.Edit the new logical volumes' counterpart of /etc/fstab to call for
   mounting the restored file systems with their normal names.
13.Reboot from the disk containing the new volume group.
14.Delete the volume group used at the start of the process.
15.Recreate any remaining file systems and restore their contents.

I suspect that this basic approach will be transferable to other Unix
implementations, although I expect the details to be noticably different. Has
anyone done this sort of thing?