Skip to Content

Introduction

This past weekend we implemented a Microsoft Cluster for High Availability of our ABAPJava Netweaver04 system. Now when we first started this project about a month ago, I would never had thought it would make the subject of a weblog. I figured this process was already well documented and would a walk in the park to implement. After all we were hardly the first people to implement SAP in a Cluster environment. It turns out this was quite true until about half way through the process when we hit the J2EE Add-in installation. From there things got quite a bit more complicated and more than a little fuzzy. Therefore I thought I would pull together our notes to help out anyone else who might be considering this process. Even if you aren’t running your system on the Microsoft platform or you aren’t considering a HA (High Availability) solution, you might want to read only. If nothing else you will get an overview of the many parts and pieces that can be combined in just an ABAPJava Netweaver04 system.

Please note that during our go-live weekend, we had a relatively small downtime window. Because of the rush, we took decent notes, but didn’t get that many screen shots. Therefore my screen shots are combination of ones from the development system (which we also clustered), the actual go-live, and after the fact captures. Therefore don’t be confused by the mixture of host names and virtual names that you might see.

The Layout

We opted to go with a two node cluster layout. In normal operation, our MS SQL database will run on one physical node. The SAP central instance will run on the other physical node. However the hardware of either node is big enough to run all the software in a fail over situation. Each node was a 3.2 GHz Dual Processor with 5 Gigs of memory. The following shows the physical and virtual services that we will have in the cluster.

image

To report this post you need to login first.

11 Comments

You must be Logged on to comment or reply to a post.

    1. Thomas Jung
      Actually I discussed that OSS note and the SAPINST_USE_HOSTNAME.  It did work when we were applying later support packages (SP12) to our installation.  However I could never get it to work with the NetWeaver04 SR1 (SP9) installation media.  Even with this switch – in my testing I still had one entry (just one field with Message Server Host) in the configuration that still had hostname instead of virtual host that I had to update manually. So although the tools do seem to now support this, I will probably still manually double check everything for the next couple of support packages until I feel comfortable. 

      It is good to hear that the IGS problem is fixed in the future. Although the solution of the standalone IGS running as a clustered service, is actually working fine right now. 

      (0) 
  1. Former Member
    do you maintain user sessions when moving the cluster group? I got this issue with a SAP R/3 4.7 installation (same architecture of your).
    (0) 
    1. Thomas Jung Post author
      No you can not maintain user sessions when moving the cluster group. The R/3 instance has to be shut down and restarted.
      (0) 
        1. Thomas Jung Post author
          I believe that this is similar to other applications running within the a cluster.  There is a short disruption in operation as processes are moved from one node to another.

          From a technical standpoint, SAP ABAP systems use large areas of shared memory to hold user sessions.  When you switch from one cluster node to another, you would loose access to your memory context.  Therefore a clean shutdown and restart during a node switch assures that this process happens without risk to transactions that are in process.

          (0) 
    1. Thomas Jung Post author
      I do work for SAP Now. This weblog was written a little over a year ago when I still worked at a customer – Kimall International.
      (0) 
  2. Former Member
    Is it possible to cluster two SAP instances in the same cluster? We have been trying to accomplish this but have not succeeded. Any inputs from your side will be greatly appreciated.

    Thanks,
    Fahad

    (0) 
  3. Former Member
    For our clustered SAP install, when a user logged in through App Server rather than CI, he is being disconnected and sometimes we get error Export_Table_Update_Conflict and DBIF_REPO_SQL_ERROR. I believe this should not happen when using App Server to login and the user should only see an hour glass and then continue with his transactions.
    Do we have to set any additional profile parameters on the App Server or CI side to overcome this issue?

    Thanks,
    Fahad

    (0) 
    1. Thomas Jung Post author
      When I wrote this blog some time ago, I held a different job.  I worked for a customer and I was recording my experiences during a project.  Since then I have taken a new job (with SAP) and haven’t been involved any longer in any MSCS setups.  I’m afraid that I have been away from the topic too long and I don’t have any helpful hints for you. But to your error – I can definetely tell you that this is NOT normal.
      (0) 

Leave a Reply