Having DCOM using a specific network route when multiple are avail

Hi,

We are running WS 2003 on several servers, some are clusters. For security
and redundancy reasons, we have 4 networks (and 4 NICs) per server. Each
network is for a spcific purpose: One is used for administration, one for
backups, one for DCOM between servers, one for DCOM between servers and a DMZ
server.

Question: Is there a way to configure DCOM in a way that all DCOM traffic
goes only over a specific network. Currently, we are under the impression
that DCOM establish a connection using our 4 networks and choose one randomly
to do the job.

Thanks in advance for your help.
Phil.

(I use this old thread because this was closest to the problem that I was troubleshooting)

This appears to be a hard task. Googling for "[HKEY_LOCAL_MACHINE\System\CurrentControlSet\Servic es\Rpc\Linkage]" might be useful.

However, I have not tried it. I had a different but related problem.

I have a DCOM client (OPC client) in a computer that has two NICs and two network addresses in different subnets (say, network1=10.10.1.0 and network2=10.10.2.0 with subnet mask 255.255.255.0). The OPC client connects to OPC server that has only one network connection, the network2. Also callbacks are used (as is usual in OPC communication). Now the callback connection for some reason uses the IP-address in network1. I would like it to use the IP-address in network2 to avoid extra routing of the packets.

I solved the problem by temporarily disabling Network1 in the OPC client computer. Now the callback connection used Network2 as was desired. After enabling Network1 again, I restarted the OPC client and (surprisingly?) the callback connection still used the correct route.


I believe that the reason was that Network2 was added to the OPC client computer afterwards and DCOM remembered that Network1 was the correct route for the callback connections. But I can only guess because I haven’t found any documentation how the DCOM callback selection works. While troubleshooting the problem, I disabled the unwanted route from the OPC server computer to the OPC client computer by adding a route definition (with command route add) that prevented connection by using the incorrect network. However, after this, the callback connection was not created at all. The Advise call returned error 0x80040202/No server specific error codes defined. EventLog did not contain any entries in either computers.

Because I have no control which network is used in the callbacks, I am afraid that after some reboots etc the communication can again switch to use the incorrect network.

Jari