2 node clustering recommendations for SQL 2008 (with windows server 2008)

  • Hi,

    Apologies if this has been asked before I did try looking around for a similar thread.

    At the present moment my company has propsed that we need to create a 2 node (active / passive) clustering solution for our SQL server databases. Although I will object to it, it is likley that other applications than SQL server will be installed on the servers (money is tight with us, so we have to make do with what we have), otherwise I would be happier with mirroring.

    Speaking with our head IT guy, he does not want to use a quorum shared drive for the database files as he is concerned with the single point of failure.

    This (from memory) immediately got me thinking about a majority node implementation. However, I'm fairly sure that this will not work I believe that the failed node tolerence is 0 for a 2 node cluster.

    Has anyone come accross this situation before. Would a shared drive be the best implementation for this scenario despite the single point of failure? Can majority node work?

    This will be my first time setting up clustering so any advice is appreciated. I have played around with clustering before, but that was on a course, and most of it was already set up. The rest of my knowledge comes from my exams (which was a little while back).

    Cheers,

    Sam

  • Just to add to this, reasearching this I have found node and file share majority for 2008.

    Node and File Share Majority is a quorum model in Windows Server 2008 Failover Clustering. The cluster is configured with two or more nodes with the quorum configuration stored on a file share. In this quorum model, a cluster remains active until half of the nodes and its witness share is available. In case, the witness share is offline, a cluster requires a majority of nodes to be up and running in order to successfully run the cluster. This quorum model is similar to the Node and Disk Majority model, but the quorum is stored on a network share rather than on a witness disk. This model is used in scenarios where a cluster is configured with an even number of nodes that do not utilize shared storage.

    Has anyone done this before? I'm curious in how the data is replicated across the nodes. Does this have to be done manually?

    Apologies if i'm asking stupid questions!

  • Sam,

    just trying to understanding how you are setting this up.

    Are you going to create one lun for Quorum and one lun for DB files?

    Or are you trying to create A Shared drive on a place other then a SAN?

    Sory just reread the post looks like you are not using a SAN...sorry cant be much help...I have always used a SAN.

    But I Believe Microsoft recommendation is not to use the Shared resource as of 2008 ( due to single point of failure) , i tried to find th the article but I was not able to locate it....

    good luck

  • dont know if you found the same article

    http://technet.microsoft.com/en-us/library/cc731739.aspx

    Understanding Quorum Configurations in a Failover Cluster

    No Majority: Disk Only (not recommended)

    Can sustain failures of all nodes except one (if the disk is online). However, this configuration is not recommended because the disk might be a single point of failure.

  • Sam -480699 (8/9/2010)


    Hi,

    Apologies if this has been asked before I did try looking around for a similar thread.

    At the present moment my company has propsed that we need to create a 2 node (active / passive) clustering solution for our SQL server databases. Although I will object to it, it is likley that other applications than SQL server will be installed on the servers (money is tight with us, so we have to make do with what we have), otherwise I would be happier with mirroring.

    Do you actually need to go to the trouble and expense of implementing a cluster, who will support it??? Ask all these questions and more!

    If you are deploying a SQL Server cluster ideally you should not be installing other applications onto these nodes

    Sam -480699 (8/9/2010)


    Speaking with our head IT guy, he does not want to use a quorum shared drive for the database files as he is concerned with the single point of failure.

    The quorum does not hold the database files, it holds the Windows cluster configuration and logs

    Sam -480699 (8/9/2010)


    This (from memory) immediately got me thinking about a majority node implementation. However, I'm fairly sure that this will not work I believe that the failed node tolerence is 0 for a 2 node cluster.

    Correct, an MNS based quorum requires 3 or more nodes. However, Windows 2003 SP1 and Windows 2008 offer a new feature which allows the deployment of a 2 node cluster using a 3rd machine as a file share witness.

    Sam -480699 (8/9/2010)


    Has anyone come accross this situation before. Would a shared drive be the best implementation for this scenario despite the single point of failure? Can majority node work?

    This will be my first time setting up clustering so any advice is appreciated. I have played around with clustering before, but that was on a course, and most of it was already set up. The rest of my knowledge comes from my exams (which was a little while back).

    Cheers,

    Sam

    MNS can work using the clustering feature mentioned above. If this is your first stab at clustering, check my tutorial at the following link. Go through this in the virtual environment, to get it down to a tee

    deploying a Windows\SQL Server cluster

    Good luck

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

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

  • Thanks for that, very useful!!

    I have more questions now (sorry).

    Is it possible to create a 2 node (active / passive) cluster where the SQL databases are not using shared storage. I have read a bit about using Node and File Share Majority? Is this possible, and if so, how is the data replicated across to the passive node? If there is a high volume of traffic, would this work?

    Also, can this be done on standard addition (I'm assuming yes as it's two nodes).

    Apologies, bit new to clustering so want to know more about my options (and whether we can avoid single point of failure with storage).

    Cheers,

    Sam

  • Is it possible to create a 2 node (active / passive) cluster where the SQL databases are not using shared storage. I have read a bit about using Node and File Share Majority? Is this possible, and if so, how is the data replicated across to the passive node? If there is a high volume of traffic, would this work?

    No this not possible, again Cluster is designed for high availability. U need to set up your cluster environment then after it is working, set up Log shipping.

    Log shipping is your DR solution, The cluster has many points of failure if it not on a SAN and if it is on a SAN then your point of Failure is the SAN itself

    So after you set up you cluster then its simple to set up logshipping....if money is the issue take a PC and turn into a server and do logshipping to it...better then nothing. I have had to do that in the past. you just put it in an offsite location and set up Database mail to let you know that the Database is up to date...

  • One last thought...maybe you need to look at Database mirroring.

    You can this will allow you both high availability and the log shipping that you are looking for.

    How to: Configure a Database Mirroring Session (SQL Server Management Studio)

    http://technet.microsoft.com/en-us/library/ms188712.aspx

  • Thanks for your replies. Mirroring is definately an option now as I have convinced the guy to have these servers purely SQL. However there will be some moaning with the developers on changing connection when principal fails. Ideally we want a common IP to use. But I have read you can use network load balancing to do this (purely having common IP and thats it).

    Furthermore, one of the databases we will need to mirror has a high amount of traffic. So network performance would be something to consider. The more I investigate, the more a simple 2 node clustering solution with shared storage appeals. Yes, there is single point of failure, but it looks like the best option in terms of mixing high availability with performance.

  • sounds like you are heading the right directions.

    In order to keep the dev at bay heres what we do..

    set up Cluster 99.999% uptime.....

    set up log shipping for DR to small box----

    Then set up a couple of VMs (with the correct IP) that are offline but are pointed to the Log shipping DB, If cluster goes down or you need point to the DR site say due maintenance then you turn on the VMs and away you go.

    Failback to the Cluster then turn of the VM's..

    GOOD LUCK!

  • Sam -480699 (8/10/2010)


    Is it possible to create a 2 node (active / passive) cluster where the SQL databases are not using shared storage.

    Thats the whole point of a cluster it requires shared storage. However, you now have the option of geographically dispersed clusters you may want to look into these a little more!

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

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

Viewing 11 posts - 1 through 10 (of 10 total)

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