Hi,
I’ve been using mailcow quite a while on a Hetzner VPS and migrated half a year ago almost all services (except monitoring) back to my homelab, since I’ve managed to get some good uplink and infrastructure set up at my basement.
Since then, I am using a VPS running OPNsense as my “external entry point” for my services. (haproxy, postfix, rspamd, clamav, etc.) connected via wireguard s2s back home. The public facing MX is the Postfix plugin from OPNsense, attached to it rspamd as a milter. This rspamd instance on the OPNsense is connected read only to the bayesian classifier at the local mailcow. So if I move a message to the Junk locally, I’ve got learns for SPAM (or other way around HAM) at the local rspamd and the external one is using those informations too.
Unfortenately, I am not quite happy with the filtering of spam and also the connection between those two rpsamd services, because I see more spam overall reaching my inboxes now with this particular setup.
I would like to ask if some of you have a similar setup and how did you manage to “marry up” those two rspamd instances?
Did you make a cluster out of those two and who is the “primary” in your cluster?
I am fully aware that this is not a supported setup of mailcow and I hope that I am allowed to ask this. 🙂
Regards,
minifisch