cluster Fail when network switch fail

Discussion in 'Clustering' started by samir, Oct 11, 2006.

  1. samir

    samir Guest

    hi,
    I have a 2 node cluster Windows 2003 server EE.
    all is OK but when we reboot the network switch, the cluster fails and i
    have to reboot all the nodes to restart the cluster.
    Is there a way to let the cluster in a waiting state if the public NIC are
    off and it will be on when the NIC are ON.
    I heared about disabling the media sensing but I am not sure that will
    resolve my problem.
     
    samir, Oct 11, 2006
    #1
    1. Advertisements

  2. Samir

    1) disable Media sense for all network adapters on all nodes :
    http://support.microsoft.com/kb/239924
    2) ensure all your NIC's has the latest drives
    3) ensure your heartbeat network is using a cross-over cable (not affected
    by your network-switch reboot)

    and try again, cluster should stay up

    HTH,
    Edwin.
     
    Edwin vMierlo, Oct 11, 2006
    #2
    1. Advertisements

  3. samir

    samir Guest

    thanks Edwin but I cannot find the DisableClusSvcMediaSense key.
     
    samir, Oct 11, 2006
    #3
  4. samir

    samir Guest

    thanks Edwin but I cannot find the DisableClusSvcMediaSense key.
     
    samir, Oct 11, 2006
    #4
  5. Samir,

    As per KB article :

    Add the following registry value:
    Value Name: DisableDHCPMediaSense
    Data Type: REG_DWORD -Boolean
    Value Data Range: 0, 1 (False, True) Default: 0 (False)

    (you need to add the value)

    rgds,
    Edwin.
     
    Edwin vMierlo, Oct 11, 2006
    #5
  6. In Windows 2003 it is enabled by default and therefore will not show up in
    the registry.

    --
    Chuck Timon, Jr.
    Microsoft Corporation
    Longhorn Readiness Team
    This posting is provided "AS IS" with no
    warranties, and confers no rights.
     
    Chuck Timon [Microsoft], Oct 11, 2006
    #6
  7. samir

    samir Guest

    Edwin,

    the DisableDHCPMediaSense exist and set to 1, but I cannot find the
    DisableClusSvcMediaSense key on the HKEY_LOCAL_MACHINE\Cluster\Parameters.
    do I have to create it ?
     
    samir, Oct 11, 2006
    #7
  8. In W2K3 SP1 you do not have to create it


     
    Edwin vMierlo, Oct 11, 2006
    #8
  9. samir

    samir Guest

    thanks Chuck

    is there another solution?

     
    samir, Oct 11, 2006
    #9
  10. samir

    samir Guest

    Edwin,

    I confirm that the Media Sensing is disabled, I just made a test with one
    node, if I unplug the Wire, the NIC stay on.
    Is there another solution ?
     
    samir, Oct 11, 2006
    #10
  11. samir

    samir Guest

    Edwin,

    I confirm that the Media Sensing is disabled, I just made a test with one
    node, if I unplug the Wire, the NIC stay on.
    Is there another solution ?
     
    samir, Oct 11, 2006
    #11
  12. did you ensure you are on the latest drivers ?
    did you ensure your heartbeat is a cross over cable ?
     
    Edwin vMierlo, Oct 11, 2006
    #12
  13. If the private NICs are supported using a crossover cable or even a separate
    dumb hub, they will be fine...however, if the switch losses power, the IP
    addresses will go offline and this will cause a 'regroup' and an attempt to
    restart the resources on the same node...failing that, a failover will occur
    and online attempts will be made....several 'ping-pong' attempts later, the
    groups will eventually fail because of loss of the public networks. That is
    the way cluster works. If you like you can administratively take the groups
    offline, do your work on the switch and then bring the groups back online.
    Disabling Media sense, in effect by default, only prevents the TCP/IP stack
    from becoming 'unbound' from the NIC. It does nothing else for you in this
    scenario.


    --
    Chuck Timon, Jr.
    Microsoft Corporation
    Longhorn Readiness Team
    This posting is provided "AS IS" with no
    warranties, and confers no rights.
     
    Chuck Timon [Microsoft], Oct 11, 2006
    #13
  14. samir

    samir Guest

    the cluster is an HP MSA500 packaged cluster and the heartbeat is a cross
    over cable.
    the reboot of the switch was not programmed, it was a power failure and it
    happens 2 times this year.
     
    samir, Oct 11, 2006
    #14
  15. We can only do so much....cannot save you from that - sorry.

    --
    Chuck Timon, Jr.
    Microsoft Corporation
    Longhorn Readiness Team
    This posting is provided "AS IS" with no
    warranties, and confers no rights.
     
    Chuck Timon [Microsoft], Oct 12, 2006
    #15
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.