Backups over backup interface.

[Sujata_Mutatk2]

Hello,
I need some help on how to design a backup network such that backup and restore all take place over backup interface but for very few servers , we should be able to use production network interface.
At present we are having media servers talk to client over backup interface and clients talk to master server over production interface, that is fine as long as Restores work fine. But they are not. When you issue a restore , it will be in Queue for long (bprd log shows , it cannot connect to client , because it is trying to connect to client from master on production network and it is not reachable that route and eventually it goes through backup interface and restore takes place).
If we use REQUIRED_INTERFACE in bp.conf , certain servers break as they cannot talk to master on backup interface but they are handful.
Any tech note on this???
Thanks

[Lance_Hoskins]

Sujata,

Make sure you have the backup interface in DNS. I would not depend on WINS resolution for backups. Also, make sure you have it defined as the perferred interface in the client properties. Let me know if you have other questions. I also believe that the master shouldn't be talking to the client unless it was also the media server that performed the backup. It should be the master talking to the media server and then the media server issuing the command off to the client.

Lance

I stand corrected on my commend about the master not needing to talk to the clients. Indeed it does need to talk to intiate communication. Must have had a little too much wine before my post last night! :)

Message was edited by:
Lance Hoskins

[Raghava_Chary]

Sujatha,

The simple solution is to use hosts file if your network is below 500 servers.
We use the same concept, just make an entry in the clients hosts file with master and media servers hostnames and BACKUP IP information ...

#Hosts file
hostname_b 172.10.128.46 (this should be your client)

#Netbackup master and media information
master_b 172.10.1.1
media_b 172.10.1.2

Similarly in the master and media hosts files you have the following entry
hostname_b 172.10.128.46 (this should be your client)
master_b 172.10.1.1
media_b 172.10.1.2

You dont have to do any other modifications like selecting an interface etc .. etc. This is working for 1000+ clients in our environment.

Just check for the bpcd log what error it is giving, check whether bpcd is listening on backup network or production.

Raghava

[Sujata_Mutatk2]

Lance,
Thanks for the helpful tips. The Backup interface is in the DNS. Master is not a media server. bprd log on master was showing that master was trying to use production interface and eventually erroring out with message "cannot connect to client" then after retries eventually restore was getting actve. Let me do some more tests and look into log.
Where is the setting for Prefered interface , I think that will be useful.
Thanks

[Lance_Hoskins]

It's in the "Client Properties-->Universal Settings-->Use specified network interface" from the NBU console. If you're intending on it using the backup interface for all communications, this is a must.

Also, make sure the order of the NIC's is set correctly. Lastly, the backup NIC (all dependant on the environment of course) only needs an IP address and subnet mask.

Let me know what you find,
Lance

[Sujata_Mutatk2]

Lance,
Prefered interface is same as "REQUIRED_INTERFACE" entry in bp.conf file???

Secondly I did some test restores in our environment today , they are working as they should be but the day we were doing DR on Fri ,we ran into situation , the one I explained above in my question.
It was a Solaris host with two disks. One disk with bare minimum solaris, netbackup client s/w with patches. I was trying to restore root and main OS on alternate disk. Restore will not start , it will sit in queue and may be after some timeout it will get active and then work. I am trying to see what was missing so that we can put that in DR procedure.

Were we missing routes?, I had put hosts entry in /etc/hosts on master/media/client.

Any way I will let you know.
Thanks

[Chia_Tan_Beng]

Sujata,

Looks like your master server can reach your client via production as well as backup LAN. If that's the case, technically you can configure both master & client to communicate only via backup OR production LAN only. If there's no other constraint, try to configure your master/media/client for DR environment to communicate via backup LAN or production LAN ONLY, that would save you lots of troubleshooting time.

[Sujata_Mutatk2]

I need to test restore under Disaster situation. Otherwise restores are working fine.