Formula una domandaFormula una domanda
 

Con rispostaHyper Cluster

  • giovedì 2 luglio 2009 1.17nrdfeeling Medaglie utenteMedaglie utenteMedaglie utenteMedaglie utenteMedaglie utente
     
    Hello Everyone,

    I've a Cluster in a 2 node Hyper-V Server Core.
    I created the cluster and add it to our SCVMM.

    The cluster is connected using HBAs. The problem is when assigning the LUNs I see each LUN 4 times (only one of those is accessible). I know that each LUN has 4 different paths, but the hosts should only see one!!!! I have PowerPath installed and after some testing I end up installing the Windows MultipathIo and did a reboot. Now I see each LUN 1 time (as supposed).  My question is: Is supposed to have PowerPath and MultipathIO installed in order to work (I mean, see the LUNs correclty).

    I have a feeling that this doesn't sound correct, I mean, having 2 different services (PowerPath "EMC" and MultipathIO "Windows") doinf the same thing?

    What is your opinion according with your expirience?

    Another question is, Do I need a to a 2 node cluster configuration (I'm asking because I has told that now, we don't have the need to have a quorum in a shared drive!!!)?

    Thanyou all.

Risposte

  • giovedì 2 luglio 2009 8.36ND82 Medaglie utenteMedaglie utenteMedaglie utenteMedaglie utenteMedaglie utente
     Con risposta

    Hi!

    Usually it should be enough to install one of them. They could even influence each other so that none of them works. What is the recommended way of EMC in this case? You should always try to install the storage the way the storage vendor tells you to.

    The wittness disk is used for the following purpose: The cluster always needs a majority (quorum) to run. There are different ways to reach this majority: node majority, node and disk ... (you can configure the qurom model in the failover cluster mmc - right click on the cluster - more actions - configure cluster quorum settings) In your case you have two players atm, namely the two nodes. If one node goes down you have 1 of 2 left, means 50% which is not enough (majority means more than 50%), so the whole cluster would go down. If you add a wittness disk you have 3 players, so if one node goes down you have 2/3 left which is more than 50% and the cluster stays up and running. BUT you always have to take care of your ressources, because each node has to be able to take all the workload of the whole cluster. Otherwise some of the services will not be able to come up again on the other node if one node fails.

    Hope that helps,
    Andreas

  • giovedì 2 luglio 2009 11.59ND82 Medaglie utenteMedaglie utenteMedaglie utenteMedaglie utenteMedaglie utente
     Con risposta
    Hi!

    The same way as I described earlier:
    failover cluster mmc - right click on the cluster - more actions - configure cluster quorum settings
    There you have the two options for disk and file share wittness

    kr,
    Andreas

Tutte le risposte

  • giovedì 2 luglio 2009 8.36ND82 Medaglie utenteMedaglie utenteMedaglie utenteMedaglie utenteMedaglie utente
     Con risposta

    Hi!

    Usually it should be enough to install one of them. They could even influence each other so that none of them works. What is the recommended way of EMC in this case? You should always try to install the storage the way the storage vendor tells you to.

    The wittness disk is used for the following purpose: The cluster always needs a majority (quorum) to run. There are different ways to reach this majority: node majority, node and disk ... (you can configure the qurom model in the failover cluster mmc - right click on the cluster - more actions - configure cluster quorum settings) In your case you have two players atm, namely the two nodes. If one node goes down you have 1 of 2 left, means 50% which is not enough (majority means more than 50%), so the whole cluster would go down. If you add a wittness disk you have 3 players, so if one node goes down you have 2/3 left which is more than 50% and the cluster stays up and running. BUT you always have to take care of your ressources, because each node has to be able to take all the workload of the whole cluster. Otherwise some of the services will not be able to come up again on the other node if one node fails.

    Hope that helps,
    Andreas

  • giovedì 2 luglio 2009 9.51Edwin vMierloMVP, ModeratoreMedaglie utenteMedaglie utenteMedaglie utenteMedaglie utenteMedaglie utente
     
    In addition what Andreas just said: you can also configure a fileshare-witness in stead of a witness disk, the fileshare will play the witness role for your two node cluster

    rgds,
    Edwin.
  • giovedì 2 luglio 2009 11.28nrdfeeling Medaglie utenteMedaglie utenteMedaglie utenteMedaglie utenteMedaglie utente
     
    Ok,
    Ok, ut how do I configure the wittness disk or fileshare-wittness? If I understand you guys correctly this is a local file, right?
  • giovedì 2 luglio 2009 11.59ND82 Medaglie utenteMedaglie utenteMedaglie utenteMedaglie utenteMedaglie utente
     Con risposta
    Hi!

    The same way as I described earlier:
    failover cluster mmc - right click on the cluster - more actions - configure cluster quorum settings
    There you have the two options for disk and file share wittness

    kr,
    Andreas