Leave feedback
  • Discussion

    StreamServe in what cluster group? (Microsoft Cluster Services)

Enter a new topic
  • Anders Ekstrom Anders Ekstrom
    0 likes 3346 views

    When configuring clusters I have seen two different setups regarding the cluster groups:

    1)StreamServe Repository Server and other resources needed by StreamServe is in the Clusters clustergroup (total 3 ip-adresses)

    2)In the Clusters Clustergroup only cluster IP, quorum disk and clustername is placed. Then the StreamServe Repository Server and other resources needed by StreamServe is placed in its own cluster group containing a StreamServe cluster IP, repository network name and all fileshares etc needed by StreamServe (total 4 ip numbers).

    Which setup is to recommend (and why)?

    Monday 18 January, 2010
  • David Shih David Shih StreamServe Employee
    0 likes

    Both scenarios work, but technically, Scenario 2 is recommended. I'm not sure about other vendors (e.g. IBM HA CMP, Veritas, Solaris, Super Dome, Red Hat...), but Microsoft's Best Practices suggest having the "Cluster Resources" (e.g. the Quorum disk, the Cluster IP and network name) in a separate cluster virtual machine ("cluster group"), with NO other resource (services, applications, print queues, physical disks, shares, etc.) in that cluster group. Ideally, the Quorum disk resource only needs to be a tiny (e.g. 10-50 MB) LUN. The special Cluster Group is *supposed to be* a standalone group, with nothing else in it. The rationale, is that the Cluster Resources should be able to move independently from other applications. If the primary Cluster Resource group has to fail over, or goes down, then it _could_ take every other cluster group (i.e. *EVERY* clustered resource). In practice, this usually doesn't happen. But it could. Additionally, adding resources to the cluster resource group would make it slower to come back online in the event of a failover. Best practices also suggest not putting *anything* else on the cluster quorum disk. If the quorum disk filled up and ran out of space, the entire cluster could go down. So the problem with Scenario 1, is that it doesn't follow best practices. You tend to see Scenario 1 more often. Because most of the time, it works fine. It's the most obvious configuration. Most Windows sysadmins have very limited experience with MSCS, and haven't bothered studying the official documentation. Even the admins who have taken the MCSE certification probably don't remember the right way to it. Scenario 2 is a little less obvious. And by creating new Resource Groups, you introduce new Network Names... and have to make sure that your StreamServe instances point to the correct Network Name of the correct Resource Group (i.e. NOT that of the Cluster Group). Attached are some old notes about disk configurations with MSCS...

    Intro to MSCS for StreamServe.doc

    Monday 18 January, 2010
  • Anders Ekstrom Anders Ekstrom
    0 likes

    Thanks David!

    I have seen both setups and knew that recommendations from MS was to use a separate clustergroup for Strs – now I have some more info from you to support this kind of setup.

    //A

    Monday 18 January, 2010