2012 Cluster misconfigurations

  • All,

    I have a SQL Cluster that I manage, but did not build and was not with the company when it was built. The specs are below:

    2012 Server Standard x64

    SQL Server 2012 Standard x64

    Active/Passive Cluster set up

    The whole environment is Virtual with a Dell Equallogics SAN

    The big problem, the SAN is running out of space.

    The two solutions:Add more disk space (which is being frowned upon) or somehow reclaim space (the method they want).

    Specifically the drive that houses the data files (.mdf) is set at 800GB for a 30GB database, that has slow growth!! The quorom has 500GB allocated to it and it is only 2GB. The drive that houses the log files 400GB and never comes close that.

    So my question is, how can I reclaim some of this space? Attempting to Google this, I am not having it much luck as I am not sure of the key words or process to look for!!

    If anyone has any suggestions, articles/URLs I can read, or any piece of advice on this, please let me know!!

    Thank You

  • If you can, I would create new drives which are sized correctly and add them to the server. Shutdown the instance. Copy the data files and log files to the new drives in exactly the same folder structure. Rename the drive letters so that the new drives replace the old ones. You should then be able to start up SQL and it will find the database files as normal. You can then set the old drives to offline for a short time in the SAN and if you're happy everything is working as expected, then delete the old drives.

    I had to do something similar when we had to increase the size of the log drive for a sql server db whilst a huge maintenance job was run over the weekend. The log grew to well over 300GB when it normally stayed around 10GB. We couldn't shrink the drive after the maintenance task and the log was shrunk back down to 10GB, so had 290GB of wasted space. We simple added a new drive, moved the log, switched the drive letters from old to new and then dropped the large drive from the SAN.

  • GBeezy (12/3/2014)


    All,

    I have a SQL Cluster that I manage, but did not build and was not with the company when it was built.

    I feel your pain i've been there before. In each case i totally rebuilt the cluster with only minimal outages during resource failover.

    GBeezy (12/3/2014)


    somehow reclaim space (the method they want).

    I've gotta say this would be my first step, rationalise what's there. Most people who hear the phrase SAN assume it means unlimited available storage :laugh:

    GBeezy (12/3/2014)


    Specifically the drive that houses the data files (.mdf) is set at 800GB for a 30GB database, that has slow growth!!

    Is it expected to grow anywhere near that? I probably wouldn't remove this for now as you may only have to add it back later.

    GBeezy (12/3/2014)


    The quorom has 500GB allocated to it and it is only 2GB.

    This is ridiculous :crazy: quite apart from the fact there's no real need to even use disk witnesses in 2008\2012 windows clusters!!

    The maximum size for a quorum drive is 512MB, looks like someone got this completely wrong huh?

    Run the quorum config and remove the current drive by setting the mode to node majority set only.

    You may then remove this drive totally from the cluster, before assigning a new disk witness, confirm the following

    • How many nodes in the cluster?
    • Are they all on same site?

    GBeezy (12/3/2014)


    The drive that houses the log files 400GB and never comes close that.

    That's a definite candidate for resizing at a later point. Get the quorum sorted first.

    -----------------------------------------------------------------------------------------------------------

    "Ya can't make an omelette without breaking just a few eggs" 😉

  • Thank You Maddave!!!

    Does this type process work for the quorom? We need to reclaim alot of that space to be used else where.

    Thanks Again.

  • Thanks Perry, I am going to need to Google this quorum stuff that you mentioned!!

    Thanks Again!!

  • GBeezy (12/3/2014)


    Does this type process work for the quorom?

    No, see my post above for details of the quorum

    -----------------------------------------------------------------------------------------------------------

    "Ya can't make an omelette without breaking just a few eggs" 😉

  • With the quorum, I would say you could just remove this from the cluster and then delete the disks, then re add them and set the quorum up again. It shouldn't have any impact on the SQL Cluster whilst you're doing this. (although I have never done it, but I don't think it should!).

    As above, we only set out quorum to 512 MB.

    Edit: Just read Perry's remarks on this, and I would follow his advice on removing an setting up the quorum properly.

  • Perry Whittle (12/3/2014)


    • How many nodes in the cluster?
    • Are they all on same site?

    -----------------------------------------------------------------------------------------------------------

    "Ya can't make an omelette without breaking just a few eggs" 😉

  • Perry,

    This a 2 node active/passive cluster and they are both at the same site.

    Thank You

  • don't bother with a disk witness and use a fileshare witness, this can be placed on a well connected peer server or even a domain controller.

    -----------------------------------------------------------------------------------------------------------

    "Ya can't make an omelette without breaking just a few eggs" 😉

  • Perry Whittle (12/3/2014)


    don't bother with a disk witness and use a fileshare witness, this can be placed on a well connected peer server or even a domain controller.

    Just curious why you would rather go with a fileshare vs disk witness. From my understanding, when using shared storage disk witness is the better choice and file share should be used with multi-site clusters.

    Bob
    -----------------------------------------------------------------------------
    How to post to get the best help[/url]

  • fileshare witnesses are mainly for geographically dispersed clusters, i find them flexible enough even for same site clusters. Of course where possible i want to use MNS only

    -----------------------------------------------------------------------------------------------------------

    "Ya can't make an omelette without breaking just a few eggs" 😉

Viewing 12 posts - 1 through 11 (of 11 total)

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