Re: asynchronous replication combined with manual insert

>my question generally is, if there can be a collision between MY
insert/update and the insert/update of the replication.

You can disable replication for a transaction so you do not need to worry about a particular insert you are making at all sites, being made by asynchronous replication a second time and causing conflict.

This is what I think you were asking?

Of course it is possible for your insert to succeed on one master, and fail on another. You would have to ensure that the application and table design prevented this.

e.g.

transaction 1 happens on master A
Non replicated transaction 2 happens at master B Non replicated transaction fails at master A due to changes made by transaction 1
--Asyncronous replication pushes transaction 1 from A to B, which may now also fail depending on its type, and resolution rules--

Your design sounds unusual - it may be that you could place the time critical tables in there own replication group, and replicate that more frequently (Some sites use replication at very high frequencies). Then you could place a fail over entry in the tnsnames.ora, so if the local master fails, the time critical transactions are automatically made on another master. This would achieve a similar effect (time critical transactions replicated quickly everywhere, but with fail-over to other masters), but still allow you to benefit from the Oracle replication architecture (i.e. confict resolution). Received on Thu Feb 03 2000 - 04:07:18 CST