[Rspamd-Users] Change of Bayes classifier and redis-DB

[Aste]

Hi!

Tuesday, October 15, 2019, 2:43:00 PM, you wrote:
WH> Hi,

WH> I have a problem with redis and rspamd. The memory usage of the redis DB grew
WH> linearly over months and I couldn't explain exactly why.
In  most  cases  problem  with  redis memory usage is because per user
settings  enabled  and  autolearn  function is turned on.


WH> Now shortly before the
WH> maximum load I updated to rspamd 2.0 and changed the bayes classifier: I changed
WH> my per-user-config (as shown here:
WH> https://rspamd.com/doc/configuration/statistic.html#redis-statistics) back to
WH> the default.

WH> After a reboot and a delete of the redis DB, the tokens are no longer saved:

WH> 2019-10-15 12:48:59 #2856(rspamd_proxy) <2af2d5>; proxy;
WH> rspamd_redis_finalize_process: cannot retreive stat tokens from Redis: skip
WH> obtaining bayes tokens for BAYES_SPAM: not enough learns 0; 200 required
WH> 2019-10-15 12:48:59 #2856(rspamd_proxy) <2af2d5>; proxy;
WH> rspamd_redis_finalize_process: cannot retreive stat tokens from Redis: skip
WH> obtaining bayes tokens for BAYES_HAM: not enough learns 0; 200 required

In  case  of empty database you need to learn manually 200 messages as
ham  and  200  messages as spam, before classifying can work, or change
thresholds (not recomennded).


-- 
Aste