Post reply

Mirroring failover - long time to re-synch databases

  • November 15, 2011 at 10:17 am

    #247813

    Hi,

    We have mirroring setup between two SQL 2005 servers.

    They are set to asynch (high performance) due to WAN geographic locations of each server.

    Our failover procedures are manual. No witness involved.

    e.g.

    Stream A (London) -- Primary

    Stream B (Glasgow) -- Mirror

    When we failover, we switch to synch mode (high safety) and then failover and then switch

    back to asynch (high performance) mode.

    This past weekend we had to shut down stream A for data room maintenance.

    We did not failover, simply shut stream A down for 6 hours.

    Then we powered the A servers back up. Once it was up and running, the mirror monitor showed that it would

    take up to 9 hours to re-synch out mirrored database. The application was available that connects

    to stream A and users would of been able to connect, however our concern is that stream B (DR) would

    be up to 9 ours out of synch?!

    Any ideas? If stream A was down and Stream B had no activity what could it be trying to synch that

    would take up to 9 hours?

    The weekend before we had a different scenario, where we failed over to stream B (users did some testing

    on stream B for 10min) and then we switched back to stream A. The mirror monitor once again said

    it would be 8-9 hours before it was in synch? How could there be that much data to synch

    and for that lengh of time? In this case (had it been closer to Monday morning) this would of impacted

    users or we would of had to point our application to stream B while A was catching up.

    Any ideas/explanation? Apologies for the long question/synopsis.

    Thanks in advance,

    HM.

  • November 15, 2011 at 10:26 am

    #1408051

    If failover occurs during a transaction in which an index or table is created and then changed, failover might take longer than usual. For example, failing over during the following series of operations might increase failover time: BEGIN TRANSACTION, CREATE INDEX on a table, and SELECT INTO the table. The possibility of increased failover time during such a transaction remains until it is completed with either a COMMIT TRANSACTION or ROLLBACK TRANSACTION statement.

    Estimating the Interruption of Service During Role Switching

    http://msdn.microsoft.com/en-us/library/ms187465.aspx

Viewing 2 posts - 1 through 1 (of 1 total)

You must be logged in to reply to this topic. Login to reply