In <31A0D4AF3516D0118C810001FA442FC9692C0D@ARES>, on 02/04/98 at 01:13
PM,
Michael Bartl <michael.bartl AT ZENTRALE.ADAC DOT DE> said:
>Not being able to provide identical hardware after a disaster case is
>an unsolved problem we have, too.
>I think we do not need the possibility of merging registry databases
>or doing other tricks, as this all may lead to inconsistent system
>configurations.
>What we really need is a possibility to mark certain settings in the
>registry as "changed" or "invalid" in a way that makes Windows NT
>install the new components from the CDROM during next bootup.
>Providing this in my opinion is not an ADSM developers task, this
>feature should come from Microsoft, as it is also needed when single
>components have to get changed without the possibility of previously
>uninstalling them.
>> We are currently running ADSM V3 on NT 4.0 servers. The machines are
>> Digital Prioris HX 6000 series. We have about 20 of this machines all
>> provided by the same supplier. However, we have noticed none of the
>> machine's hardware setup are IDENTICAL. Some have slightly different
>> versions of the same network cards, different screen drivers,
>> different firmware settings etc. etc.
>> The problem is the registry data containing information about the
>> hardware components of the machine because how can we be sure the
>> machine will come back up if some of the hardware settings are no
>> longer valid.
I have found that their is NEVER a need to do a "Disaster Recovery" of
an NT Server, ever (well, barring a disaster wiping out ALL the
servers in a domain, that is).
If you configure, as a matter of course, EVERY server to be a domain
controller, and after a disaster you have at least one domain
controller surviving, there is *NO*NEED* to disaster recover other NT
servers. Here is why.
Simply promote the surviving NT server to Primary Domain Controller
and re-install the other servers from scratch as Backup controllers.
Allow the DBCD replication to occur, then restore the *data* backups.
This means, of course, that you should have a separate system
partition (C) with only "re-installable" software on it. All data
should live on the DATA drive (which is backed up by ADSM).
The only case in which an actual disaster recovery is required against
an NT server is where ALL domain security information is destroyed,
and the most cost effective way to prevent that is to backup the PDC
installation partition to standard tape using NTBACKUP, or, if it is
really important, having a BDC located at a different site (offsite
but network connected).
I have not yet found a Server which cannot be more effectively rebuilt
by any other method than pristine re-install, replicate, and restore
data.
It does mean, however, that you need to have "installation media"
handy for recovery (the NT CD, SQL Server CD, SNA Server, etc) and
know how to do an initial install of all the software on the server.
The only thing that you have to re-create is the network shares.
Assuming that you have used a "rational" method of laying out shared
directories, this can be easily automated and done in minutes, even
for thousands of home directories.
I had originally planned to use ADSM to do disaster recovery of NT
servers, but re-wrote the Disaster Recovery Plan to completely avoid
this necessity, thus simplifying the entire process. The same method
can be (and is) used to "disaster recover" AIX machines by using an
initial NIM install onto the new hardware, then restoring the "data"
and selected configuration information.
This is a much more effective method than trying to generate the AIX
equivalent of MKSYSB images for NT.
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Keith Medcalf (416) 410-5791 http://www.dessus.com/ IBM
OS/2, LAN Server, DB2, TCP/IP, DOS, Windows 95, Windows NT
Finger or email kmedcalf-pgp AT dessus DOT com for my PGP Public Key
|
