I want to implement Always on from 1 primary server to a DR site on maby DBs, the AG will be one for each DB, and the synchronization mode for most of the AG will be synchronized. Failover will be done manually, and will always include all the DBs. we will never have some DBs active in one site, and others in the other site.
If a disaster will happen, in order to return to the primary, we are not expecting to use the auto recover of the mirror, but recreate the DB.
We have some cross-DBs transactions. My question is what will be if a disaster will happen in a middle of a cross DB transaction, will the transaction end or will rollback in both DBs, or could I find the DBs in a partial data lost situation. (meaning in one DB the command which was part of the transaction will end, and in the other will not exists).
I saw this article http://msdn.microsoft.com/en-us/library/ms366279(v=sql.110).aspx which worries me.
I have 2 scenarios:
1. Always on with synchronize mode on 2 seperate AG, both on the same server, replicas are to the same server, no automatic failover, when failover one AG, we fail over the other one as well. Each AG on one DB(i.e. DB A is under AG_A, DB B is under AG_B), I have a cross-database transaction. Can I be sure that the transaction will be on the same state on both replicas in the DR site, if a disaster happens in the middle of the cross DB transaction?
2. As above, but asynchronize mode. Will it keep the same state of the transaction on both replicas? will it help if I will have one AG to cover both DBs (when the mode is asynchronized)?