locked
Doubt about clusters RRS feed

  • Question

  • Hi there,

     

    I´m with few doubts about a cluster enviroment, if anyone can help me will be very usefull:

    1 - A cluster is from windows and no sql server right?

    2 - I know that sql server do not divide the processos between the server connect on the storage, so, if i have 2 servers connected, 1 very powerfull and another not so much, can i spicify in witch server i wanna run my query?

    3 - Can a storage be "the storage" of 2 clusters enviroments?

    4 - The max servsr connectad on stardard version is 2 right?

     

    I belive thats it...thanks!


    ------------------------------------------------------------- Oracle OCA11g
    Friday, March 25, 2011 2:49 PM

Answers

  • exactly - this is the classic "active/passive" model which i would say, is the most common.

    One node has the storage. If it fails, the other node starts, which starts the SQL service on that node after first bringing the storage online.

    You can set up an active/active cluster if you can divide you applications amongst two SQL instances. In this model you would have two clusters using the same set of nodes but with different storage. In this setup you would have to allow enough performance on each node that it could cope with the possibility of both instances running off the same node.

    Ben

     

    Friday, March 25, 2011 3:27 PM

All replies

  • Hi Fabrizzio,

     

    I'll try to answer your questions.

    1) Clustering is set up at the Windows level first, and when that is done you can install SQL Server on to that cluster (which essentially provides a virtual computer name and IP address). When you install SQL you get the option for installation standalone or cluster. You would choose cluster after you had already set up your Windows cluster.

    2) The important point here is that only one instance of SQL Server runs on any one cluster node (i.e. machine) at any time. You can have different hardware specifications in different nodes but it is not recommended because during failover, SQL is started on the other server and having less OS resources it will probably not perform as fast as it did on the other node. 

    3) Storage for clustered must be a shareable resource, like a SAN for example. The storage is only ever presented to one node at a time, so only the active node can 'see' the disks on which the database/logs live. When there is a failover, the storage subsystem is 're-pointed' to the other node.

    4) You can have up to 16 nodes in a Windows Server 2008 R2 cluster with SQL Server 2008 +. Different OS/SQL versions offer different maximum amounts of node per cluster

     

    Ben

    Friday, March 25, 2011 3:19 PM
  • Ben,

     

    Just one more to go!

    First of all, thanks for the reply, very helpfull!!!!

    Ok about the failover, but, if i have a cluster eviroment, lets say, 2 nodes working with 1 storage, a simple cluster.

    I will be using just one node? the other one will only be there, waiting for fail?


    ------------------------------------------------------------- Oracle OCA11g
    Friday, March 25, 2011 3:23 PM
  • exactly - this is the classic "active/passive" model which i would say, is the most common.

    One node has the storage. If it fails, the other node starts, which starts the SQL service on that node after first bringing the storage online.

    You can set up an active/active cluster if you can divide you applications amongst two SQL instances. In this model you would have two clusters using the same set of nodes but with different storage. In this setup you would have to allow enough performance on each node that it could cope with the possibility of both instances running off the same node.

    Ben

     

    Friday, March 25, 2011 3:27 PM
  • Thanks Ben!
    ------------------------------------------------------------- Oracle OCA11g
    Friday, March 25, 2011 3:33 PM