It has come to my attention that the /Tcptrunkwidth-x setting in the MTA config file is ignored by the MTA. This apparently was done around the time of GroupWise 6, but it was never documented. Now the MTA handles TCP threads dynamically.
Here is what I was told:
The GWMTA algorithm for IP connection allocations is this:
The TCPTrunkwidth switch is ignored by the MTA. The TCP sender threads are started dynamically. On startup, the MTA will check its holding tanks, and any holding tank with a message in it will have 1 sender started. As messages are routed through the system, and those messages are dropped into a holding tank, the sender will check the number of messages that are backed up for that destination. If there are more than 15 messages backed up, and there are less than four senders for that destination, a new sender will be started.
When the MTA restarts, all the sender threads go away, and we start the process over again.
All this said,
- You should still increase the /tcpinbounds, as those default to 40.
- In the MTA eDir Agent object, you should still enable the 2nd Priority threads.
I am assuming that these have not changed and have not been made dynamic. For those that have heard me recommend this in the past – sorry, I just wish we had known sooner.