Le 15/02/2020 à 06:14, Julien Muchembled a écrit : > After a quick investigation, I understand that was because of bayesian > filters.
I think I misused the term "Bayesian". I used it in opposition to filters that analyze the form of the message or how it is transported. > If you think it's not possible with default configuration, > I can provide more information and you can stop reading here. I didn't see that the default logrotate settings are so aggressive and I don't have the logs anymore. > In fact, I don't expect much from this bug report. I am already quite unhappy > of rspamd for other reasons, mainly because it's too slow at learning ham/spam > (I configured dovecot to trigger that when I move mails from/to the Spam > folder). Should I have feed rspamd with my spam & ham folders when I set it up? Bogofilter is clearer about the fact it has to be done. I have "bayes_classify: skip classification as ham class has not enough learns: 89, 200 required" logs and that would explain why I found it slow at learning ham/spam. > With this incident, I am now decided to try bogofilter. Done 2 weeks ago (along with postfix-policyd-spf-python) and now I purge rspamd from my server. Julien