cancel
Showing results for 
Search instead for 
Did you mean: 

proof of no-loss sync scenario

Adrian_Ogden_2
Level 2
I'm trying to setup VVR in a fully sync (no override) configuration with another site (in this case with Oracle 10g installed on Windows 2003).

If I create a script that is continuously updating a table which along with oracle control files and logs etc is being replicated to another site I get a mismatch in number of rows present after breaking the LAN and performing the following:

Primary: Remove volume from RVG and restart Oracle on it.
Secondary: Stop replication, remove volume from RVG, assign a drive letter and startup Oracle.

There are outstanding entries in the SRL on the primary side (as one would expect) but do these get committed to the actual data volume on the primary side when its removed from the RVG thus making it more recent than the secondary?

Or how else can I ensure both end have the same data on the disks upon a WAN failure?

Thanks
1 REPLY 1

TomerG
Level 6
Partner Employee Accredited Certified
USING NETWORK FAILURES: 

If you chose "synchronous: fail", then when the link is broken, the Oracle application on the Primary site would get I/O errors and stop. Now following your procedure you should have the same data on both sides.

If you chose "synchronous: override", then when the link is broken, the Oracle application on the Primary converts to asynchronous mode, and will flush data and be more "up to date" than the Secondary, and you are correct, they will be different.

If you chose "synchronous: off" then of course the Primary will always have newer data then the Secondary (since this is asynchronous mode).

USING SITE FAILURES:

Note that if you had "synchronous: override" (or "synchronous: fail") and the Primary site CRASHES (not a network issue), then the secondary would have no data loss, and the data on the Secondary would be the same as what the Primary had when it failed.