Today I realized that it was indeed the sa-learn that is slowing down my Archiving.

I added an «&» on the end of the line that calls the sa-learn (called by the report-ham.sieve script):

exec /usr/bin/sa-learn -u ${1} --ham &

Does anyone thing this is a bad idea? I noticed that archiving got way faster this way…

Best,

Francis

On 9 May 2024, at 23:09, dovecot-request@dovecot.org wrote:

...

On 9 May 2024, at 19:45, dovecot-request@dovecot.org mailto:dovecot-request@dovecot.org wrote:

Francis Augusto Medeiros-Logeay via dovecot skrev den 2024-05-09 12:29:

...

...

If you want to analyze emails, why not do it with scheduled crons after they are archived?

maybe ask for skip older then one day in here https://github.com/isbg/isbg/tree/master

while roundcube just set the needed flag for retest ?

...

wouldn't it reanalyze all archived messages anyway?

sometimes email is ham at recpt stage, while its spam later on test, this is why isbg is made imho :)

I see.

But I wonder if there is something wrong with what I already have. It worked fine for years, and it works fine for most of my users. I tested with a few accounts, and Archving (with its sa-learn script) works blazingly fast. But with one user - happens to be mine - it is super slow when archiving, most likely due the sa-learn.

Best,

Francis