Add RAM to 2K5 Cluster

  • The decision has been made by management to add 8GB of RAM to each of our 2 node SQL 2005 Std Ed. cluster with 8GB already. OS is Windows 2003 Enterprise 32bit.

    Can someone tell me at what point in the process do I need to change the SQL Maximum server memory setting?

    Our plan so far is to:

    1. upgrade the RAM (+8GB) on the inactive node. (Node 1) startup etc ...

    2. failover (using Cluster manager) to node 1 with the 16GB RAM (can we set max sql ram here).

    3. take node 2 offline - shut down node & Add the RAM (+8GB). Start up, bring on-line

    4. failover again to node 2.

    (at this point both server are are 16GB.

    When can we set the maximum server ram to use the added 8GB on each server.

    Currently 2 GB is for OS and 6144 is for SQL. New setting will be 2GB for OS and 14336 for SQL.

    Do we need to failover again to use/see the added memory after changing these settings OR can we change the setting on the active node with 16GB ram before we failover to the 2nd server upgrade?

    Just need to obtain the steps to add ram to the 2 nodes on cluster and have SQL see it.

    Thank you.

  • Hi Dana,

    I think the best time to set the max memory would be between step 1 and 2.

    Then as the instance is failing over to node 1, the updated configuration that you have applied will come into effect on the hardware on which you have already installed the additional RAM.

    Then you can add the RAM to your second node and fail the instance back without needing to reconfigure again - only needs to be done once for an active-passive cluster.

    If you are going to allow SQL Server to consume the whole 14GB of RAM though, then I don't think that the max memory configuration is really necessary.

    I think you would typically set the max memory configuration when you don't want SQL Server to use all the RAM in a server. For example if the server had 2 instances of SQL Server, you might constrain each instance to use a maximum of half the available RAM.

    You might however like to consider the min memory setting. This will allow SQL Server to immediately grab a large chunk of memory as soon as it starts or as soon as the instance fails over from one node to another.

    This I think is considered a good idea because then the instance already has a nice hefty amount of RAM reserved from the outset, and it doesn't need to keep making multiple requests to Windows to give it more RAM until it reaches a decent level.

    Just a word of caution, be careful with the min memory setting because if the amount of memory that you have specified as a minimum is not available, then SQL Server will probably not start.

    No such danger with the max memory setting.

    Good Luck.

  • Thanks Andy for the good explination.

    I will check with the vendor to see if min of 4096 and a MAX 8192 will be ok for SQL. That will add another 2GB for the OS for a total of 4096.

    Thanks again.

  • Since your system is a 32bit environment and you already have 8 GB per node, your already hitting the max memory limitations of the OS. Is there any reason why you need to add more RAM to your system

  • I am trying to get that answer for a week now. All I know is the System Admin is trying to schedule a time to add it. I was not involved in the process.

    I'm Just the DBA!

  • Dana,

    So what reason did the Sysadmin give in saying that adding more RAM SQL cluster? Do you have any performance data on your environment. Maybe if you have a 64bit environment the situation would be different.

  • We are at WIN 2003 Enterprise. Cluster is only under enterprise edition.

    Still trying to get an answer to why we are doing this.

  • The answer to why we are doing this is to be proactive. We will be adding 100's of users and database interfaces in the next few months.

    Steps to follow again are:

    1. Upgrade Ram on inactive node (Node 1) (+8GB) (total will be 16GB) - restart.

    2. Change SQL Max mem to 11264 in management studio server properties, memory. Want to keep at least 4GB for OS.

    3. failover the cluster to node 1 with the 16GB RAM.

    4. take node 2 offline - shutdown and add additional 8GB ram. start back up bring on-line.

    5. failover to node 2 - to validate failover and configuration change.

    that leaves 5GB to windows.

    thanks in advance for additional replys.

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

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