I have contracted an integrator to install the following:
- Brocade 2800 Silkworm switches
- Emulex lp9002 HBAs
- Sun Eclass servers
- STK L700 w/9940 FC tape
- EMC 8830 disk array
Each Sun has two HBAs connecting into the brocade SAN. We are running EMC PowerPath and additional products to manage the zoning and SAN. We are using Veritas Netbackup to backup the locally installed media servers. We are starting to experience strange tape media errors with a full load on the tape drives. We have had STK, Veritas and the integrator trying to resolve the issue.
I am concerned that they may have overlooked the obvious, running tape and disk down the same HBAs. I have read several articles by you and others stating that tape should be on a separate controller. It has been recommended by EMC and Sun that this is a best practice. What do you think? I need an impartial opinion that isn't trying to push product or point fingers.
SAN-based backup can be done using various methods. Veritas Netbackup on Solaris supports "server free" backup by using the "extended copy command" (Xcopy) in the SAN. Using Xcopy can remove the production server (your server with two HBAs) from the backup equation. The issues for backup are:
1) Who's going to do it? The backup server or the production server
2) How are we going to mount the LUN to back it up?
3) Will LUN access be over the network or the SAN?
4) Do we have to back up offline or can we do it online?
Who is going to do it depends on whether you have a shared library in the SAN and whether the backup engine needs to be on both a backup server AND production server, or you just need an agent on the production server. LAN based backup with open file agents and online agents for databases negates the need for a third HBA for SAN backup.
Offline-based backup where the production server also has the backup engine installed and access to the LUN is through the production server, may not need a third HBA since the application is down anyway.
Online-based backup can be done in two ways. You quiesce the production app only long enough to snap off a BCV then bring production back up, and the production server backs up the BCV to a SAN shared library. This is a good candidate for a third HBA. The other online method is to break off a BCV and have the backup server mount that BCV through the SAN and back it up directly or with Xcopy through a data router. This scenario would not benefit from a third HBA.
Hope this helps.
Editor's note: Do you agree with this expert's response? If you have more to share, post it in our Storage Networking discussion forum.
This was first published in March 2002