Getting a consistent copy

We're running Oracle 10.2.0.4 on RedHat. We're supporting an application that requires completely consistent copies when it sets up what you can think of as additional cluster nodes.

Our current process is:

1) Stop the source application
2) expdp the database (about 100G)
3) restart the source application
4) load the dump into the target server and let the app "catch up" on

This has a rather significant problem in that it requires we stop the source application (which is no-bueno).

We tried using exp with the consistent flag, which dumped ok, but when we times the loads back into the system they taking > 1 day (whereas the data pump import takes about 2 hours).

One thing we're considering trying here to achieve the same thing though is to use SAN snapshotting, something like this:

1. Leave the source running
2. SNAP /u01 on the san
3. Mount the SNAP on a different database server in read/write mode
4. Restart Oracle over there
5. Let it go through a recovery (since we did just crash it)
6. Run a data pump export out of there.

Is anybody doing something like this? Does anybody know of any subtleties that I should be aware of? In theory the above should work fine, but as one of my colleagues is fond of pointing out, in theory you shouldn't need practise. Received on Mon Jul 20 2009 - 12:09:38 CDT