SQL Server Performance

Heartbeat network in SQL 2008 clustering

Discussion in 'SQL Server 2008 Questions [Archive Only - Closed]' started by WingSzeto, Feb 1, 2009.

  1. WingSzeto Member

    We are doing active passive clustering. The way the BOL shows how to configure heartbeat network is to just let cluster use the network and not let the public use it. Is it possible to have two network ports set up like that? Will the cluster confuse which one is for the heartbeat? We have three network ports. One is for the public. The second is for the replication. And the third is heartbeat. The replication network is the one that public doesn't need to use just like the heartbeat network so that's why I am thinking to set it up just like the heartbeat network, am I thinking it correctly?
    wingman
  2. satya Moderator

    What is the operating system used here, Win 2003 or 2008?
    Why do you want to use 2 ports in this case?
  3. WingSzeto Member

    It is Windows 2008 64 bits and we are also using SQL 2008 standard 64 bit. Each node has a total of 4 network ports (two network cards, one onboard and one PCI network card) and we are using three out of four and the fourth one is diabled. So we are actually using three ports We do that to isolate the traffic so the heartbeat is not affected by anything.
    wingman
  4. WingSzeto Member

    After thinking about this more, maybe I do think it wrong. The heartbeat network should be the only one that allows cluster use. Maybe the replication network should have the same settings as the public network in the cluster, in which they both allow cluster and client to use. Let me show our current network settings in the cluster.
    Public: 10.10.0.1
    Replication: 10.10.1.1
    heartbeat: 10.10.2.1
    Here are my questions:
    1. In each of the Network cluster properties, it has a radio button that said "Do Not Allow the cluster to use this network". What will be the situation to use it? Since our replication network is used only between the publisher and subscriber, would it be the situation? Or should I still alow both Cluster and client to use it?
    2. If I want a failover happen when the replication network fails, how do I set up the dependancy? For the public network, its IP is tied to the SQL instance name so if the network fails, it will fail over because of the dependancy. For the Heartbeat network, it has the Disk and Majority voting mechanism to handle the failover. But for the replication network, there is no dependency between its ip to any application or instance name at my current setup because there is no application or service I can tie it to. Is there a way?
    Any pointers are very much appreciated.
    wingman

Share This Page