Post reply

Three Node cluster - Active /Active /Passive

  • August 8, 2005 at 12:58 am

    #88425

    Hello everybody,

    We are trying to consolidate our servers, (already posted few questions in this forum) and as a requirement from the management we have been asked to move databases to instances (3 nodes) running under Active /Active /Passive mode.

    My doubts here are

    (1) Can we ever do this? I said "Yes", it may be possible but never tried. I thought I will get back with you guys, in here! Can anybody give your thoughts on this?

    (2) As I was doing my research on this, my teammate came up with this article

    http://www.microsoft.com/technet/prodtechnol/sql/2000/maintain/failclus.mspx?pf=true.

    Now, as I went thro' this and was specifically highlighted by my team member on a topic:

    Scenario One: Four-Node Multiple-Instance SQL Server 2000 Failover Cluster, Three Active Nodes, One Standby (N+1)

    Here is the small brief,

    With four-node support, Windows 2000 Datacenter Server provides more flexibility in terms of a cluster configuration. The recommended way of using a four-node Windows 2000 Datacenter Server cluster in a SQL Server environment is to have three of the nodes each owning an instance of SQL Server 2000 and have the fourth be the warm standby. This is not unlike a log shipping scenario, or a single-instance failover cluster in which at least one node is waiting for work. This scenario is known as N+1. Instead of configuring your failover cluster to allow the instances to fail first to a node with another instance of SQL Server 2000 running, the fourth node should be configured as the primary failover. This would reduce the issue of having too many instances starving the resources of one node. AWE memory should be enabled in this scenario to allow each instance of SQL Server to address more memory than the 1 GB currently available. This allows your applications to scale out rather than limiting them if they exceed the memory allocation for SQL Server.

    Can somebody please help me to understand the highlighted portion? And my teammate argument here is that, why do you want to run a 3rd instance on the 3rd node for configuration - Active /Active /Passive, when it is going to be a warm standby server and being a primary failover?  I am unable to convince with any explanation but my gut feeling says, the configuration is possible and we would require a 3rd instance running too. But why is it, "according the document", I don't get it.

    I really appreciate all your time and help on this.

    Thanks!

    Viking

  • August 8, 2005 at 5:06 am

    #580360

    Hi Viking, We're also looking at implementing this setup in the very near future, the highlighted bit in the text simply says that you should configure the passive node explicitly as being the primary failover node.

    This means that when one of the active nodes fails it automatically trys to recover on the passive node rather than doubling up on one of the other active nodes (causing a potential doubling of the workload on that node).  If the passive node was already in use (say 2 of the servers had died) then there would be no choice but to failover onto the active nodes.

    Hope this helps, i'm sure I could have explained it better but it's a monday and I can't be bothered to go into too much detail

  • August 8, 2005 at 7:31 am

    #580389

    Mike, Thanks a lot for a timely reply. I was totally out of hopes, that would I ever be able to understand and explain it. You just trigerred my thoughts flow!

     

     

    Thanks!

    Viking

  • August 8, 2005 at 7:32 am

    #580390

    No Problem Viking,

    I'm glad I could help

  • August 9, 2005 at 9:57 am

    #580793

    To further clarify...you do not want to run an instance of SQL on the third node...the word "passive" means that zero cluster resources are running on that node...an active node is a node with a cluster group on it (cluster groups would basically each be an independant instance).

    I hope this helps...but basically you end up with teo cluster groups with two instances of SQL on two active nodes...and one node waiting as the preferred failover node (this being the passive node) for those two cluster groups.

  • August 9, 2005 at 10:02 am

    #580796

    hmmm.... not quite sure you're right there...

    Yes the passive node would be exactly that 'passive' and as such you're right in saying that it wouldn't be 'running' a sql instance, however, it would still have to have sql installed (at the virtual server level) in order to provide failover support for the other nodes should one of them fail.

  • August 9, 2005 at 10:06 am

    #580798

    That is correct, SQL the software itself would be installed, however it is not an "instance" the word instance talks about a "unique" SQL installation. The total number of instances in this cluster as a whole would sit at 2, not 3...but the binaries themselves would be installed on all three servers.

Viewing 7 posts - 1 through 6 (of 6 total)

You must be logged in to reply to this topic. Login to reply