Thursday, 20 May 2010

Server manager and a clustered agent for enterprise server

I could not find much “how to” documentation on this subject.

Actually, I could not find any.  I guess there is a point document out there, but my search skills are not up to scratch.

I have a client that is 8.96, they are upgrading to 8.98 and have clustered enterprise servers.  This is fairly easy to configure, but then how is server manager going to work?

I decided to install SM on the active node and install the files on the same drive as E1, so that the config would follow it’s way with JDE.  I do NOT think that this is the right move, but cannot be too sure.

I added the generic service clustered resource to the batch server group.  I then changed the properties of this service not to restart and not cause the group to restart.

Moved the group to the failover node and then installed the SM service on this second node (InstallAgentService.bat).  I then created some dependencies, that this service would only come up once JDE was up and the network name was up etc.  All important stuff.

I thought I had it in the bag, until…  I failed over and then went to the console.  The console was registering the base machines, not the cluster name.  So for each clustered server, I had two physical machines listed one with an active agent and active JDE and the other with shutdown JDE and shutdown agent. Surely this could get better!

I did some reading and they recommend host file changes for the cluster, but I don’t like doing that – especially for the agents that live on the cluster.

The only time you notice a problem is when you are running on a failover node – which is not a huge problem.  I’m guessing that if you do the following, the statistics will be available anyway:

  1. stop agent
  2. stop console
  3. rename folder in targets dir on agent for the service that you want to see
  4. start console
  5. start agent

I think you’ll then be able to see the server and the stats.

No comments:

Extending JDE to generative AI