[Rspamd-Users] rspamd per-user statistic user limit

[Gabriele Nencioni]

On 7/4/20 10:12 AM, Gabriele Nencioni wrote:
> Sorry, I have forgot to say I'm trying to perform a massive learning of
> thousand ham and spam messages from a directory.
> 
> So how can I do that avoiding the counters have been reset to zero?

Hi all again,
about this problem, after I have performed severl tests, I would add
some news.

I think it's a statfile counter issue because all the keys are properly
stored on redis backend.

Moreover using the same redis backend for rspamd 1.8.3 and 2.4.2 servers:
~# rspamd -v
Rspamd daemon version 1.8.3
~# rspamadm configdump | grep _servers
        write_servers = "172.20.235.61";
        read_servers = "172.20.235.62";

~# rspamd -v
Rspamd daemon version 2.4
~# rspamadm configdump | grep _servers
        write_servers = "172.20.235.61";
        read_servers = "172.20.235.62";


I have the following rspamc stat output result:
~# rspamd -v
Rspamd daemon version 1.8.3
~# rspamc stat
...
Statfile: BAYES_SPAM_USER type: redis; length: 0; free blocks: 0; total
blocks: 0; free: 0.00%; learned: 3953; users: 1594; languages: 0
Statfile: BAYES_HAM_USER type: redis; length: 0; free blocks: 0; total
blocks: 0; free: 0.00%; learned: 1948; users: 1321; languages: 0
Total learns: 5901

~# rspamd -v
Rspamd daemon version 2.4
~# rspamc stat
...
Statfile: BAYES_SPAM_USER type: redis; length: 0; free blocks: 0; total
blocks: 0; free: 0.00%; learned: 1535; users: 594; languages: 0
Statfile: BAYES_HAM_USER type: redis; length: 0; free blocks: 0; total
blocks: 0; free: 0.00%; learned: 444; users: 321; languages: 0
Total learns: 1979

so exactly 1000 users less for each statfile.
Moreover also the "Total learns" counter is different as you can see
from above output.
The issue occurs only on peruser classifier.

Since the whole Keyspace is ok, the rspamc scan properly works and each
message (learned) properly triggered the BAYES_HAM_USER or
BAYES_SPAM_USER symbols.

So can you confirm it is a statfile counter visualization problem?
If so, how can I fix it?


Thanks in advance
Regards,
-- 
Gabriele Nencioni