Disk Configuration on cluster with SAN

  • Given a choice which is the best place for the SQL server binaries?on local w2k server drive or SAN drive.I have allocated RAID1 array on SAN drive.

    Hitesh

  • We keep ours local, data on fibre. Nice and simple, plus good separation if you're having a problem with your storage.

    Andy

  • I would agree.

    What is the benefit of having the binaries on the SAN where that space could be used in the true shared nature. You should never have to copy the binaries across the SAN , or back them up to a central point which I see as the true benefit of the SAN.

    The essential thing is to have them on a RAID 1 config.


    Simon Sabin
    SQL Server MVP

    http://sqlblogcasts.com/blogs/simons

  • I'd keep them local. That way in the event of SAN failure, you could still start SQL server and load the dbs elsewhere, reattach and go.

    Steve Jones

    steve@dkranch.net

  • If you are doing a cluster as the subject suggests, you have to put your binaries on a shared disk.

    Good, Fast, Cheap, pick two.


    Good, Fast, Cheap, pick two.

  • I'll have to show my lack of knowledge here. I thought each copy of the cluster kept its own binaries, then you just had a shared quorum disk?

    Andy

  • - Each node keeps its own binaries on its own local disks.

    - Quorum goes on a shared disk.

    - Any data that is to be failed-over, e.g. Database data files, go on shared disks.

    That's the standard setup. Quorum disk must be shared, else you'll never be able to do failover. Data files must be shared, else applications won't start up properly during failover (SQL Server won't be able to start up databases, for example, if it can't access the data files). It's pointless to place Binaries on a shared disk, since a node will only use the binaries that its local installation of the application points to, even during failover.

    And just to clarify, "shared" disk does not mean that both nodes can access the disk simultaneously; Windows clusters work on a "shared nothing" principle: only one node can access a disk at a time. Shared simply means that the disk is visible to multiple nodes in the cluster, and should a failover occur, another node can acquire access to the disk.

    Matthew Burr

    Edited by - mdburr on 08/21/2001 11:13:56 AM

  • In a cluster, the binaries have to go on the local drive. Even when not in a cluster, I would place the binaries on the local drive. Then, your data is reading out of your fiber card while the binaries are being used by SCSI.

    Brian Knight

    bknight@sqlservercentral.com

    http://www.sqlservercentral.com/columnists/bknight

Viewing 8 posts - 1 through 7 (of 7 total)

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