Today I will explain how to setup very easily the anti-spam SpamAssassin and make it work with the OpenSMTPD mail server (OpenBSD default mail server). I will suppose you are already familiar with mail servers.
We will need two packages to install: opensmtpd-filter-spamassassin and p5-Mail-SpamAssassin. The first one is a "filter" for OpenSMTPD, it's a special meaning in smtpd context, it will run spamassassin on incoming emails and the latter is the spamassassin daemon itself.
As explained in the pkg-readme file from the filter package /usr/local/share/doc/pkg-readmes/opensmtpd-filter-spamassassin , a few changes must be done to the smtpd.conf file. Mostly a new line to define the filter and add "filter "spamassassin"" to lines starting by "listen".
Website of the filter author who made other filters
SpamAssassin works perfectly fine out of the box, "rcctl enable spamassassin" and "rcctl start spamassassin" is enough to make it work.
Official SpamAssassin project website
It should really work out of the box, but you can train SpamAssassin what are good mails (called "ham") and what are spam by running the command "sa-learn --ham" or "sa-learn --spam" on directories containing that kind of mail, this will make spamassassin more efficient at filtering by content. Be careful, this command should be run as the same user as the daemon used by SpamAssassin.
In /var/log/maillog, spamassassin will give information about scoring, up to 5.0 (default), a mail is rejected. For legitimate mails, headers are added by spamassassin.
I use a crontab to run once a day sa-learn on my "Archives" directory holding all my good mails and "Junk" directory which has Spam.
0 2 * * * find /home/solene/maildir/.Junk/cur/ -mtime -1 -type f -exec sa-learn --spam {} + 5 2 * * * find /home/solene/maildir/.Archives/cur/ -mtime -1 -type f -exec sa-learn --ham {} +
SpamAssassin is quite slow but can be speeded up by using redis (a key/value database in memory) for storing tokens that help analyzing content of emails. With redis, you would not have to care anymore about which user is running sa-learn.
You can install and run redis by using "pkg_add redis" and "rcctl enable redis" and "rcctl start redis", make sure that your port TCP/6379 is blocked from outside. You can add authentication to your redis server &if you feel it's necessary. I only have one user on my email server and it's me.
You then have to add some content to /etc/mail/spamassassin/local.cf , you may want to adapt to your redis configuration if you changed something.
bayes_store_module Mail::SpamAssassin::BayesStore::Redis bayes_sql_dsn server=127.0.0.1:6379;database=4 bayes_token_ttl 300d bayes_seen_ttl 8d bayes_auto_expire 1
Configure a Bayes backend (like redis or SQL)
Restart spamassassin after this change and enjoy. SpamAssassin has many options, I only shared the most simple way to setup it with opensmtpd.