Hi, in my multi server setup i have to mail servers that are mirrors of each other. We use 2 MX records with different priority for incoming mail which results in the "main" server getting most of the mails. That means that the bayes/fuzzy/neural filter on this server is trained much more as the server has more traffic. Note that these two rspamd instances are "clustered" with the rspam neighbors function. Is there a simple way to transfer this data to the other server without using a redis master/slave configuration? Or is this even necessary?
Could you maybe try something like this? https://stackoverflow.com/questions...e-a-redis-database-from-one-server-to-another I guess only Bayes data should differ? So if you can live with that, its likely not nescessary.
Well there are a few things that differ (see: https://rspamd.com/doc/tutorials/redis_replication.html) but i'm only interested in bayes as this helps alot in fighting spam in our setup. We in the process of implementing a spamtrap aswell which will feed rspamd as training material so it would be nice if both servers have at least a similar dataset for bayes. I'll have a look into it and see if it is possible to just move them over, or if there is another way to accieve this.
