Disaster recovery - two sites

[ljoobo]

Hi all,
what is the best way to achieve two-way disaster recovery between two TSM servers in two sites? One idea is electronic vaulting of TSM DB and some critical data over thin IP link and offsite tapes for the rest.
One way is using virtual volumes. What about mirroring DB & log over distance?
Can you help me with these and maybe some other options?
Thanx for any suggestions/advice

[Eldoraan]

If you're talking about internal TSM db/log mirroring - don't do that over a WAN type link. Too much lag will be tough to keep copies in sync.

We use the unsupported config of SRDF on EMC storage to replicate our db/logs to remote site. It works pretty well but IBM will not support us if we have trouble.

[heada]

Remote vaulting of the TSM DB Backup with server-to-server over the IP link is your easiest and most supported option. That gets the database to the remote location, you still need to do the copy pool volumes. You can do that also with remote vaulting but make sure your pipe if big enough to handle that (your nightly backup volume over the WAN might be too much)

-Aaron

[GregE]

We use the unsupported config of SRDF on EMC storage to replicate our db/logs to remote site. It works pretty well but IBM will not support us if we have trouble.

I'm looking at the same things sometime in the next few months. What does IBM support if not that?

Right now we have a single TSM server. What we will probably have (options are still not definite) is a hot standby server at the DR data center. This server will not normally take any backups and will be the same name as the primary in the event of DR or DR testing, so server-to-server is out of the question.

I like the option of replicating not only the DB, but also the logs, with SRDF. Getting the logs will allow a great recovery point objective. Have you tested recovery at your DR site with that setup?

[Eldoraan]

As for what IBM will support, in the 5.x and earlier versions, no method of replicating db/log as far as I know. We've basically designed our DR environment to fire up the offline replicas after an SRDF split. If they don't come up, we use DR Plan file and do tape restore. So far, all of our DR tests of the SRDF replicated db/logs has been successful. We are running in syncronous mode, not sure what async would do to replication since you could get a snap while transactions are being committed. We've been running this way for about 5 years and so far it's worked through 5.1 to 5.4. Haven't moved our AIX servers past that since new direction is NBU and TSM will eventually become legacy platform.... unless NBU blows up and we have to reverse!

[Jeff_Jeske]

We are also an EMC shop. We use SRDF to replicate the data to the DR site storage frame via private fibre.

Our setup is this... we backup only production servers at the primary site. We backup our qual and test servers at the DR site. The prod and test TSM servers are identical. In the event of a disaster we bring up production at the DR site on the test server. This is all scripted and has been tested to be functional.

The thing that gets tricky is the zoning. It took quite a bit of planning to sort out all the details but it works flawlessly.