locked
ILM 2007 clustering

    Question

  • Greetings,

     

       As we all know, ILM 2007 (MIIS 2003 SP2) does not support clusterring, because it is not a cluster aware application. how can i benefit from sql server clustering to implement clustering for ILM 2007. or if anybody knows how can i implement clustering in ILM 2007? I would appreciate any help.

     

    I need badly any information that is useful

     

    Thanks in Advance

     

    Regards

    Wednesday, December 26, 2007 12:44 PM

Answers

  • Ahh yes, thanks for clarifying.  Let me see if I can explain - WLBS doesn't apply to ILM implementations as WLBS is an inbound service used to balance requests over a specific range of protocols and ports.  You would only load balance services for outside clients to talk to your ILM server - which should not be happening as ILM is not "advertising" any services to be consumed.

     

    If your intention is to use information in the metaverse for an enterprise application to consume you absolutely need to provision this information into an ADAM (or some other LDAP or SQL repository) and load balance the ADAM instance.  The only clients attaching to the ILM server should be Administrators accessing the system remotely.  While I am a big proponent of leveraging things like SQL Reporting Services to mine data in the metaverse, consuming for anything other than ad-hoc reporting is a really bad idea.

     

    Wednesday, December 26, 2007 3:36 PM
  • You cannot implement an Active/Active cluster with ILM because you can't have two installations of ILM talking to the same instance of the MicrosoftIdentityIntegrationServer db.  What you can do is have ILM installed on both nodes to one of the shared LUN's so the binaries and MAData folder is on one of the disks that is in the Cluster Virtual Server group.  This allows the ILM data to follow the database across nodes without the need to replicate the MAData folder on its own.  When using the Generic Service resource, one node will have the Active instance of ILM and the second node will have ILM in a stopped state.  During a failover, the cluster service will move the group (including the disk resources for ILM and the DB) to the other node and start the service.  As long as you didn't install anything dependent on one of the local non-shared drives everything should spin up the same way.

     

    NOTE - whatever process you are using to automate the Run Profile execution must also be able to either follow the cluster resource between nodes or you must replicate it through some other means.  There isn't a way that I'm aware of to replicate Task Scheduler jobs between nodes of the cluster.

     

    Wednesday, December 26, 2007 4:43 PM

All replies

  • Hi,

     

    With clustering there are two problems to solve:

    1. Providing fault tolerance for the database
    2. Providing fault tolerance for the application

    This is, of course, after you've eliminated any other single points of failure in the hardware and network.  In solving the first problem the use of SQL 200x clustering is fully supported.  This in itself can be configured in one of two configurations, one with the application on a single remote system communicating with a remote SQL cluster, or with the application on one or more of the nodes of the SQL cluster (questionably supported).


    In solving the second one, there is no official support for clustering of the application; however it does work if you use the Generic Service for the application resource.  AlexT has the definitive writeup on this "not officially supported" configuration.

     

    http://blogs.msdn.com/alextch/archive/2005/12/17/clusteredmiis.aspx

     

    Wednesday, December 26, 2007 1:29 PM
  • Thanks alot for your help,

     

    i'm grateful for your assistance, but sorry i have not been specific about this issue i'm facing, i would like to know if i can implement a Network load balancing cluster for MIIS, does microsoft provide clustering support in ILM2 Beta 2 (ILM 2008). i would be grateful if you help me with this issue.

     

    Thanks in Advance

     

    Regards

     

    Wednesday, December 26, 2007 3:17 PM
  • Ahh yes, thanks for clarifying.  Let me see if I can explain - WLBS doesn't apply to ILM implementations as WLBS is an inbound service used to balance requests over a specific range of protocols and ports.  You would only load balance services for outside clients to talk to your ILM server - which should not be happening as ILM is not "advertising" any services to be consumed.

     

    If your intention is to use information in the metaverse for an enterprise application to consume you absolutely need to provision this information into an ADAM (or some other LDAP or SQL repository) and load balance the ADAM instance.  The only clients attaching to the ILM server should be Administrators accessing the system remotely.  While I am a big proponent of leveraging things like SQL Reporting Services to mine data in the metaverse, consuming for anything other than ad-hoc reporting is a really bad idea.

     

    Wednesday, December 26, 2007 3:36 PM
  • so you say that i can failover MIIS by failing over the SQL service. and whenever any change is made to the metaverse, it is going to be replicated to the other node. thanks for your help. another question if i'm not bothering you. can you tell me if i can implement Active/Active failover cluster with MIIS? i'm very grateful Mr. Turner. thanks alot again

     

    Regards

    Wednesday, December 26, 2007 3:44 PM
  • You cannot implement an Active/Active cluster with ILM because you can't have two installations of ILM talking to the same instance of the MicrosoftIdentityIntegrationServer db.  What you can do is have ILM installed on both nodes to one of the shared LUN's so the binaries and MAData folder is on one of the disks that is in the Cluster Virtual Server group.  This allows the ILM data to follow the database across nodes without the need to replicate the MAData folder on its own.  When using the Generic Service resource, one node will have the Active instance of ILM and the second node will have ILM in a stopped state.  During a failover, the cluster service will move the group (including the disk resources for ILM and the DB) to the other node and start the service.  As long as you didn't install anything dependent on one of the local non-shared drives everything should spin up the same way.

     

    NOTE - whatever process you are using to automate the Run Profile execution must also be able to either follow the cluster resource between nodes or you must replicate it through some other means.  There isn't a way that I'm aware of to replicate Task Scheduler jobs between nodes of the cluster.

     

    Wednesday, December 26, 2007 4:43 PM
  • Sorry - one other addition regarding the automation of Run Profiles.  Given you requirements it's advised that you use the SQL Agent or a custom service (also using the Generic Service resource) to accomplish this.  Both of these methods should be able to failover across the nodes as long as the service is installed on the shared drive and not locally; I can't emphasize that enough!

     

    Wednesday, December 26, 2007 4:48 PM
  • Dear Brad,

     

    Thanks alot for your enormous efforts in helping me with this issue, i'm so grateful for your generousity in giving me some of your valuable time to help me.

     

    Thanks Again

     

    Regards 

    Thursday, December 27, 2007 8:16 AM
  • Brad, as always, excellent post.

     

    I did want to clarify one thing regarding the SQL Clustering.

     

    While ILM can't be setup as an active/active cluster (or multi-instance), there is no real reason why the sql database for MIIS/ILM could not be part of a multi-instance cluster (active/active), where the other SQL instances are supported different applications.

     

    SQL Server Instance 1 -- has MicrosoftIdentityIntegrationServer database

    SQL Server Instance 2 -- has database for MOM

    SQL Server Instance 3 -- has database for someother app

     

    These instances could run on 2-8 nodes. With 4 nodes you have an n+1 config and with anymore nodes beyond that an n+m (where m is the number of spares). Some folks are finding this as a more economical way to cluster SQL Servers, that way they don't have to have half the hardware doing nothing. In a 3+1 config only 1/4 of the hardware is doing nothing and you don't suffer performance degradation if one node goes down, but you would if two go down. In a 3+2 you still save, and can withstand two nodes failing without performance degradation.

     

    Brad, BTW, your suggestion about using the SQL Agent to schedule the run profiles is my preferred approach and will work in a clustered environment because the msdb database will failover to the other node.  However, if you are running DTS packages or SSIS packages you might encounter issues since DTS and SSIS are not really cluster aware. In fact SSIS would need to be installed on each node.

     

    Thursday, January 10, 2008 8:49 AM
  • Hi,

    I am writing to enquire whether some body can help us in implementing a fail over cluster for the ILM application server as mentioned in the below link... 
    http://blogs.msdn.com/alextch/archive/2005/12/17/clusteredmiis.aspx

    We are willing to pay for the services...

    Kind Regards
    Sree
    Monday, March 16, 2009 8:43 AM
  • Sree,

    We can help you with that. In fact I just did this on Friday night on Windows Server 2008 
    http://www.ilmbestpractices.com/blog/2009/03/ilmmiis-sync-engine-clustering-windows.html

    DLundell
    at
    ensynch 
    dot
    com



    David Lundell www.ilmBestPractices.com
    Monday, March 16, 2009 5:32 PM