Do i require to freeze the i/o for spliting the BCV in DR

A BCV on a synchronous R2 is no different than if it was on the R1 side - If you think you should freeze the I/O on the R1 side, then you should do the same on the R2 side as well. The reason for this is that not all the BCV's are "split" at exactly the same time (even with a -instant) so unless you do something to hold the I/O on the R1 devices, you run the risk of getting your BCV devices at different points in time (just slightly) and not maintaining consistency across all of the devices.

To get a consistent BCV image from either the R1 or R2 side, you need to do at least one of the following:

1. use Timefinder Consistency groups and perform a -consistent split. This split will hold the writes to the R1 devices whilst the split occurs - in this way, you are getting a consistent, crash recoverable image on the BCV as no I/O is performed during the split process. The write holding is done by the symmetrix or by powerpath (if you have older microcode).

2. use some type of freeze / thaw process which essentially does the same as above but at the database layer.

3. Use a database backup technique like hot backup mode during the split process. In this way, you can make the database consistent by using the regular database roll-forward/backward on the BCV itself after the split has completed.

4. Use none of the above, but have your data solely on 1 hyper volume.


The bottom line is that just performing a regular split of a BCV without any additional "help" does not guarantee you a consistent image, but you might get "lucky" some of the time..

Glen.