Server to server backup

[Darrell Campbell]

I have two HPUX rp4440 servers with each using 11iv23. One of the servers is a production server that has all internal and external drives mirrored. I would like to break the mirrors and backup the data to the other server. From there I will put it to tape. There will be a total of about 150GB of data to deal with which is spread across multiple volumes.
Does anybody have any suggestions on the best way to get the data from the production sever to the backup server?

[TwoProc]

Well, I'm assuming you're breaking the mirror to pull the drive - so to get the data to the other server - you'd plug it into the server???

Why are you breaking the mirror?

Why not just use a "dd" command on the primary server and copy it to tape right there?

Or, use tar?

Or, use ignite?

Please explain the issues you're trying to overcome - from what i can see - breaking the mirror and putting the drive in the chain on the destination server "gets the data to the other server". From there you could put it on tape, or you could just put it on tape from the primary server - without breaking the mirror... same thing, much less work...

[Darrell Campbell]

No, I wasn't planning on pulling drives. I want to kill off existing users, shut down our database and break the mirrors so the data being backed up is from a quiet set of disks and then backup the data to the other off-site server's hard drives. From there I would go to tape at our leisure (probably using fbackup).
I think the real question is: after breaking the mirrors, I was wondering which soultion would provide the fastest server to server data transfer: "dd", "tar", "fpt","nfs" or any other solution. I sort of wanted to stay with a sloution/process that is scriptable.

[James R. Ferguson]

Hi Darrell:

I'd consider making JFS snapshots and using 'fbackup' for the backups. In this way you don't need to break mirrors and thus leave ample protection for you production server.

The use of 'fbackup' means that you have the ability to intrinsically handle both sparse and large files.

The use of the JFS snapshot provides the time to do a quiet backup on a static filesystem.

A nice guide to JFS snapshots can be found here:

http://www.docs.hp.com/en/5992-3385/ch03s05.html#bajcggcd

Regards!

...JRF...

[Bill Hassell]

Something to consider: fbackup is useless to replace a defective boot disk. Not having a local tape drive is very, very costly in terms of down time and recovery time. And remember that 150GB is a massively large amount of data for a typical 100 Mbit LAN connection. You'll get just a bit more than 5 MBytes/sec throughput over a 100 Mbit connection which means about 8 hours will be required to transfer the data. If you have a 1000 Mbit LAN connection, the time drops to an hour. This all assumes a typical fbackup-to-network backup design.

But if you need to restore all the data, you've got a very long recoverfy process. First, you have to install HP-UX on the server. This absolutely requires accurate knowledge of the volume groups and logical volumes as well as a copy of fstab, bdf and any other special features such as striping and mirroring. Once you build the new system and get everything mounted, now you can restore the fbackup data using frecover over the network. Figure at least a day or two for recovery.

Or you setup an Ignite network server and use Ignite to backup vg00. Then use fbackup to create full and incremental backup tapes. The network backup server can then provide a bootable image in case there is catastrophic damage to the boot disks, or you need to resize /stand, / or swap.

Or you could buy a tape drive...