Shared loader directory, RAC and WebLogic Clustering

A friend wrote:

I was reading your paper on OTN http://www.oracle.com/technetwork/articles/havewala-gridcontrol-088685.html Its a very interesting paper and well written. I am doing a similar implemention using OEM 11g on a RAC environment and had a few questions..

1. In the paper you reference about using an IP configuration to load balance between a RAC environment. In this option, uses will round robin between the two management servers running on the two instances. In this case, what happens to the feeds that come from the managed agents to the EM cluster. the feed from the agent is only received by one instance and the certain data such as the log details are available only on one instance.. how does the data get over to the other instances. Can i use ACFS to locate these files so that they are visible to both servers.

2. Have you tried to cluster the weblogic servers so that they make the OMS highly available? the Weblogic server that comes with the GC download is not complete and many of the features are not available. however if the organization has weblogic can this be used, what are the implications with use this option, specifically with respect to future OEM upgrades etc.Is weblogic clustering supported for OEM implementation?

My reply:

Thanks for your kind words. RAC is optional. Even the load balancer is optional. The feed from the agents also goes to the load balancer on a different port and it is routed to the available management server. In normal case, this is ok. If there is a failure, that is where the concept of the shared loader directory is used - so other management servers can be used to pick up the load. This concept is explained in the advanced EM manual. You should be able to use ACFS for the shared loader directory, however it is not certified or supported for use for this purpose. See the manual for the additional steps to set up the shared loader directory.

WebLogic Clustering will not be used by EM GC 11g Release 1.

Good Luck for the implementation. My tip is, keep it as simple as possible. My idea would be to first put this on a single database with Data Guard, and then upgrade to RAC after EM GC has settled down.

Regards,

Porus.