Post reply

Failing over a SQL Cluster Instance

  • August 6, 2014 at 7:09 pm

    #289591

    I'm curious as to how other people here are handling planned fail-overs within their SQL Cluster environment.

    When you do a Failover do you stop any application related services first, or do you fail-over with the application still running?

  • August 7, 2014 at 1:11 am

    #1735852

    With the apps still running.

    That way its a true representation of a unplanned fail over.

    If the apps are not cluster aware then stopping them for a planed fail over isn't the same as an unplanned fail over.

  • August 7, 2014 at 2:25 am

    #1735867

    Thanks for your comment 🙂

    Hope to get some more contributing as well.

  • August 7, 2014 at 2:41 am

    #1735874

    If you're doing the failover as part of a real-life test, then Anthony is right - leave the applications running. If you're failing over for some other reason, for example to apply a SQL Service pack, you'll want to consider how well the application handles an interruption in the availability of the database. If it's a fairly robust application, again, there's probably no need to stop it, unless you have informed users that it will be unavailable due to maintenance.

    John

  • August 7, 2014 at 2:51 am

    #1735877

    Thanks for your thoughts John.

  • August 7, 2014 at 2:58 am

    #1735881

    If you plan a failover, always stop your application services first. Why intentionally rip the database from under the app when you could have a clean shutdown and startup. The end user experience will also be maintained too.

    Any unplanned failovers should have been done during your operational acceptance testing phase.

    -----------------------------------------------------------------------------------------------------------

    "Ya can't make an omelette without breaking just a few eggs" 😉

  • August 7, 2014 at 3:09 am

    #1735886

    Perry Whittle (8/7/2014)

    If you plan a failover, always stop your application services first. Why intentionally rip the database from under the app when you could have a clean shutdown and startup. The end user experience will also be maintained too.

    Any unplanned failovers should have been done during your operational acceptance testing phase.

    Thanks for your comment Perry.

    A bit of background, I did up a document of how each of the applications responded to a cluster failover, and middle management was happy with that given the time of the failover being done. It's done early morning when there is little or no use of the applications so OS patching and other similar activities can be done during the day.

    However the management level above that are now asking why we aren't stopping the application services. That was what prompted me to ask what others are doing.

  • November 21, 2014 at 12:14 pm

    #1760252

    I know you said you were doing this in the early morning with little activity. Just keep in mind that if a massive update, insert, or delete is running without commits in them if you fail over SQL Server you could have a very large rollback on your hands and the database will be unavailable during this time. We once experienced that and it took 30 minutes for the rollback to complete.

  • November 21, 2014 at 12:55 pm

    #1760261

    PhilipC (8/7/2014)

    Perry Whittle (8/7/2014)

    If you plan a failover, always stop your application services first. Why intentionally rip the database from under the app when you could have a clean shutdown and startup. The end user experience will also be maintained too.

    Any unplanned failovers should have been done during your operational acceptance testing phase.

    Thanks for your comment Perry.

    A bit of background, I did up a document of how each of the applications responded to a cluster failover, and middle management was happy with that given the time of the failover being done. It's done early morning when there is little or no use of the applications so OS patching and other similar activities can be done during the day.

    However the management level above that are now asking why we aren't stopping the application services. That was what prompted me to ask what others are doing.

    Where possible, you should always take the applications and databases down cleanly.

    -----------------------------------------------------------------------------------------------------------

    "Ya can't make an omelette without breaking just a few eggs" 😉

Viewing 9 posts - 1 through 8 (of 8 total)

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