How to Avoid Rebasing databases.


HI All,
Gud mng...

Progress database version 11.7.4.
OS is AIX 7.1.
We have 2 servers Prod and DR with Open edge replication plus enabled.
Source DB size
db1 is 800gb
db2 is 50gb
db3 is 15gb
db4 is 10gb.

Target DB size
db1 is 800gb
db2 is 50gb
db3 is 15gb
db4 is 10gb

Activity : Management are planning to perform DR drill activity to make source as target and target as Source.

While changing roles from target to source and source as target. In DR db1( will become as source) take online backup -repltargetcreation flag which take 3 hrs. to complete and to restore in source db1( which is target) takes another 3 hrs. and this whole process is taking approximately 7hrs to complete.

Is their way to re-sync new source and target databases without taking backup and restore.



With old-style one source, one target setups it can, in theory, be done. I think it might have worked for me once many, many years ago. Mostly i ended up rebaselining, something pretty much always glitched. There are some kbases that talk about it, search for "failback".

It sounds like the biggest problem with your current procedure is that a backup takes 3 hours. That's pretty slow for just 800GB, have you considered backing up and restoring to internal flash drives? That would probably take closer to 40 minutes.

Depending on how long the DR test takes maybe you could do the restore prior to completing the DR test and then roll-forward AI files on the failback server to get the trailing transactions. That should be quite a bit faster than waiting for a backup & restore cycle after everything is done.

With "replication sets" it is a *lot* easier and works quite well. But that requires one source and two targets. The configuration is a bit tricky but once you have it going it is very smooth for these sorts of purposes.

The really nice thing about replication sets is that you don't have to wait for a backup or a restore. The transition can be as fast a a db start & restart. A customer has a 2TB database that we can "flip" in about 5 minutes. We could go faster if we wanted to more aggressively kill off connected users.

The other nice thing about replication sets is that there is never a point where you don't have at least two live copies of the database available.