Hi,
I am working with management in my organization with regard to our DAG fail over settings. By default, the fail over setting is set to 'Good Availability' (missing 6 logs or less per DB). My organization did not feel comfortable with data loss so we changed it to 'Lossless'.
Of course, we had a SAN failure and we lost a DAG member and nothing failed over to the surviving DAG member. Even eventID 2092 reported the same log sequence for many databases yet the surviving DAG member did not mount. Example:
Database xxxmdb28\EX-SRV1 won't be mounted because the number of lost logs was greater than the amount specified by the AutoDatabaseMountDial.
* The log file generated before the switchover or failover was: 311894
* The log file successfully replicated to this server was: 311894
Only after the SAN was restored and the surviving server came back did another 2092 EventID get logged stating the log sequence was not 311895 (duh! - because the database got mounted again. )We opened a support case with Microsoft and they suggested no databases mounted because the surviving DAG member could not communicate with the non-surviving member (which is crazy to me because isn't that THE POINT of the DAG??). Maybe there is always a log file in memory (?) so AutoDatabaseMountDial set to lossless will never automatically mount any database? Who knows.
In any case, we are trying to talk ourselves back into setting it back to 'Good Availability' setting. Here is where we are at now:
2 member DAG hosting 3000 mailboxes on about 36 databases (~18 active on each node) in same AD site (different physical buildings) w/witness server
AutoDatabaseMountDial set to lossless
Transport dumpster is set to 1.5x maximum message size
What level of confidence can we expect we will not have data loss with the properly configured transport dumpster and a setting of 'Good Availability?' I am also open to other suggestions such as changing our active-active DAG to active-passive by keeping all active copies on one server.
Also, has anyone experienced any data loss with 'Good Availability' and a properly configured transport dumpster?
Thanks for the guidance.