Domino SMTP connection round robin

Neil Langston
23 January, 2017

Domino has its own internal processing or ideas on how it implements TTL for DNS entries on SMTP Connection documents (where “Connection type:” is SMTP), which is frustrating as it doesnt always conform to what is out in the wild which you have diligently configured in DNS. Therefore if you are attempting to transfer messages over SMTP to a bunch of known internal Microsoft Exchange servers using a GSS/LB you can sometimes (obviously depending on your implementation) find that Domino will keep trying to deliver to one of the servers rather than trying a different one as part of your load balanced set. This can result in mail buildups whilst we wait for Domino internal TTL to reset itself if something happens to one of the servers and we happen to be connected to that one when it goes down. Because of the way domino calculates when it should try the connection next, you can find these delays get exponentially longer.

There are plenty of notes ini (SMTP_RELAY_HOST=, SMTP_RELAY_RENEWTIME=, RouterDNSErrorTTL=) and configuration document (Dynamic cost reset interval, Initial transfer retry interval) tweaks you can do to make this situation better, but these did not give the perfect result (for our very large implementation) to ensure message delivery for our environment. Because the coexistence environment was in a large global bank with 200,000+ employees we could not have a situation where there might have been mail delays…. therefore it had to be resolved….

What was found (under the guidance of IBM) is that we could configure a list of IPs in the “SMTP MTA relay host” part of the SMTP connection document to round robin the requests to different Exchange servers. The key here was to have double square initial brackets, then the list of IPs individually between square brackets, but do not close the initial square bracket at the end. For example:-

[[192.168.1.1],[192.168.1.2],[192.168.1.3],[192.168.1.4],[192.168.1.5]

Hey presto,.. no more mail delays when there were issues with the Exchange servers using a GSS/LB. 🙂