none
Causes of consistency checks on SQL database RRS feed

  • Question

  • DPM 2010

    I'm trying to determine what exactly triggers a consistency check for a database.  I have a clustered environment with multiple instances on the cluster.  When server patching occurs and the servers reboot(unexpected shutdown?) this seems to be one cause, but it doesn't seem to always apply to all the protected databases on the cluster.

    I've run the following tests on a new cluster group that is not actively used as of yet:

    1.Offline/online the sql instance - no inconsistency

    2.Moved the group to the other node in the cluster - no inconsistency

    3.Rebooted the node causing a failover - no inconsistency.

    Does a database have to have an open transaction for an inconsistency to be detected? Or just some database file differences due to changes made to the protected database but not yet reflected in a Full Express or incremental sync.

    I'm asking all this in an attempt to avoid consistency checks on some multi TB databases.

    thanks

    Thursday, October 18, 2012 1:32 PM

All replies

  • Hi,

    You should check this link (Mike Jacquet answer) :

    "Inconsistencies occur for two reasons.  The filter driver on the protected server loses track of block level changes (maybe a system crash, or unexpected cluster failover) or a Sync failed in the middle and we could not update the replica."

    http://social.technet.microsoft.com/Forums/en-US/dataprotectionmanager/thread/5dd8dcdb-056b-41bf-974c-af6773f82a67

    Stephane


    Please remember to click “Mark as Answer” on the post that helps you. This posting is provided "AS IS" with no warranties. knowledge is valid only if it is shared by All.

    My DPM blog Yet Another DPM Blog

    Thursday, October 18, 2012 3:21 PM
  • Thanks for the quick reply.

    So in my scenario with a 2 node active/passive cluster - say N1 and N2 - where N1 is the active node, should all cluster groups be moved to N2 before patching and rebooting and then moved back after the server is back up.  This would not be a system crash, or unexpected cluster failover so would the filter driver on the protected server lose track of block level changes?  Can I expect consistency or does it depend on the lightweight comparison of file metadata?  I can see how this scenario would cause an inconsistency if a sync is running.

    I'd be interested in hearing how others work around planned outages.

    Thursday, October 18, 2012 4:21 PM