I wouldn't worry too much about this, but I understand your point. I had 2 DMX3's on 2 sites a while ago and managed them through 2 agents on each site (2 x 2). I found out exactly the same thing.
I asked EMC the same thing back then and there was no solution they said.

The only solution I found was disable an agent and let the failover take place from this one agent to the other. when you bring the agent back up the Symm stays where it is and the load is better balanced.

Perhaps that with newer ECC's than the 5.2SP5 which I had, this behavior is enhanced and is working better.

Anyone ?

I used to call my personal and friendly CE for things like this.

We have around 50 Symm's across the sites so it makes quite a difference to us when the DCP's are added into the equation.

This seems kind of odd to me. We had this problem with CC 5.2sp5 but since we upgraded to 6.1 (I can't remember what it was doing at 6.0) I have always seen the primary agent be local to the Symm.

Mind you we have a much simpler config than you. I only have 3 DMX3 arrays at 3 sites with an SRDF/A link between two of them. I have Symm/SDM agents deployed on local hosts at all three sites.

I guess the only thing you didn't specifically mention checking is that the local hosts for each site have access to gatekeepers from each and every array at that site. If you have an agent on a host at site A but that host can't "see" gatekeepers from an array at that site you could end up managing the array through an RDF link from an agent at another site.

Allen

Yes we do have GK's (8 from each array) allocated to the local management servers.