Manually failover a DAG... RRS feed

  • Question

  • I have something of a unique scenario. 

    I have a DAG setup between two servers at different sites. The witness server is at site 1. My understanding is that should the server at site 1 go down, the witness server at site 1 will communicate that to the DAG member at site 2 and enable the failover. However, if the entire site 1 goes offline, how can the DAG member at site 2 be made aware? Should I/can I enable some kind of manual failover? I read in this article ( how to disable automatic failover, but I do not know how to enable the member manually when I need to. Forgive the newbie questions. DAG is all very new to me. What am I missing?

    Tuesday, December 4, 2012 12:38 AM

All replies

  • In your scenario you have the following - is this correct?

    • Site 1 - Witness share
    • Site 1 - Mailbox server 1
    • Site 2 - Mailbox server 2
    • DAG between the sites

    I actually built a lab last week testing different scenarios and this is what happens with your set up:

    • If the witness server at site 1 dies the DAG stays up (needs 2 nodes to stay online) so long as network connectivity between sites exists
    • If the witness server AND the mailbox server at site 1 die then the whole DAG goes offline and manual intervention is needed
    • If the mailbox server in site 2 goes offline, the passive databases in site 1 come online automatically

    So what if you lose site 1?

    • As above, the whole DAG goes offline and you need to manually bring the DAG back online on the single node in site 2
    • There are two ways to do this - either the 2008 R2 cluster tools or using the Exchange cmdlets if you have DAC enabled
    • Having only two mailbox servers split across two sites (if it is a WAN) isn't high availability due to the fact if the WAN goes down for a few minutes, site 2 will think it is offline and active its passive databases in site 1 which will mean when the WAN comes back up again, your users are connecting across the WAN to site 1 - depends how much high availability you need

    How I bring site 2 online?

    The post I made last week about this asing the same question is here -

    If you have DAC enabled (as per thread above) the following document can be followed (from thread above) and this document takes you through it step by step - VERY useful!:

    • Edited by Douggly Tuesday, December 4, 2012 12:55 AM
    Tuesday, December 4, 2012 12:53 AM
  • In my scenario,  site 1 is where are the users are. Site 2 is a virtual site which only contains an Exchange server and a DC. It's sole purpose is to give the users a place to connect to if the primary Exchange server goes down. If the Internet connection at site 1 goes down, then the server at site 2 will think that both systems are offline. Then what? How do I bring this online in case of an emergency?

    Tuesday, December 4, 2012 6:31 AM
  • So, I ran some testing in which  brought the mailbox server at site 1 down. The mailbox at site 2 mounts, but the clients do not "failover." I look at the Failover Cluster Manager and it says that the whole cluster is down and I can not bring up site 2. However, if I bring the mailbox server at site 1 up again, I can bring the cluster back online and the users will then see the mailbox server at site 2 (until I activate the mailbox at site 1 again). Have I done something wrong?  I just need to make sure that, in the event of a server outage at site 1, users can continue using the mail server at site 2.

    Forgive my obviously "newbie" questions.

    Sunday, December 16, 2012 3:40 AM