I found the best way is to build the OS to match the original, then recover all data including system files via NetBackup. I do have BMR configured, but the sequential order for restore will take much longer then OS, and then full systems restore on top. I don't do BMR for Solaris and Linux systems at all, just a full system backups and then OS build with data restores. It works well on Windows with full systems restore including the regsitry and system state, no really a problem with diffirent hardware, there are some tricks anywhere. You would need de-select a few system files, use w2koption per the tech note, and you will be fine.
We tested this and you are right that you will likely run into problems doing an entire system restore via netbackup. Even with the best will in the world, there is bound to be some kind of configuration file which will mess things up on the running system during the restore. Will the servers at site 2 be the same architecture / patch level / NICS etc ? Unless you have everything 100% the same you'll run into snags. Also... make sure that when you lay the data down you don't lay down the Netbackup files or you'll hose your restore (as we found out !!).
If the servers will have identical names and IP's etc... why not just build them as if they were the servers in your home data center but with duplicate (but empty) file systems. In a DR situation all you'd need to restore would be your data files into those empty filesystems. The OS stuff would be as if they were your servers in your home data center.
Mark Glazerman
Desk: 314-889-8282
Cell: 618-520-3401
? please don't print this e-mail unless you really need to
Most servers get really ticked off if you try to overwrite the running OS files - assuming they'll let you do it at all. Probably because an overwrite is effectively a delete and then create. So great, what happens when you restore that critical library file that Solaris was using to run ? Or heck, when you restore bpbkar ? Or inetd ?
If the CSA guy refuses to back down though, no sweat, ask for a proof of concept test. "Let's see what really happens when we do it this way." If only because it ought to be fun to see exactly how messed up the destination server gets. =) And don't sign off on it as the full DR method until you get one.
Guys i'm after a bit of backup as am perhaps doubting myself now.
I'm having a bit of a row with a CSA (solutions architect) at our company.
A backup design has been done where 2 sites have Solaris clients
configured with the same name and ip, and one site is just disconnected
from the network.
There is a Netbackup Server 6.5.3 (Windows) which backs up the
connected clients from site 1.
What they want for a DR test is this :-
1. Disconnect the clients from site1 on the network.
2. Enable the network connections of clients on site2 (with same name
and ip of site1 clients)
3. Restore to the running Solaris server through netbackup of a client
image taken on site 1.
My understanding was you wouldn't ever try and restore a whole system
from file system based backups to a running solaris OS. Is that correct
still?
I also can see all sorts of problems having the client names the same
and same ips. Arp tables etc. I'd personally have the client names
referenced differently in Netbackup regardless of the hostnames which
could be the same?