On Thu, Jul 17, 2014 at 05:08:15PM +0200, Benny Pedersen wrote:
> > HAM: 183595 (150000 required)
> >SPAM: 52692 (150000 required)
> >Insufficient spam corpus to generate scores; aborting.
> >Exit Status 9 is not zero for do-nightly-rescore-example
> 
> could it be solved with longer backlog for known spam & ham, not
> extending minimal spam & ham, just thinking on how to solve this
> long standing problem, since i see it comes daily now :(

I know my corpora are small enough not to make much difference (about
20.000 ham altogether, plus a little bit of spam), but: Would it be
possible to increase the time span during which mass-checks are accepted?
I.e. run whatever uses them later that day?

At the moment I am supposed to start mass-check not before 9:00 AM UTC.
Yesterday, the "do-nightly-rescore-example 9" message appeared here at
12:07 PM UTC.  That's little more than three hours.  Even if my laptop is
running at 9:00 AM and starts crunching messages immediately, it probably
takes more than three hours for the slow "weekly" run.  If the laptop is
not running at 9:00 AM, I seem to have no chance of making the deadline
that day, even with the fast "dailies".

Of course, I may be missing something. :)  For example, the Date: header
of that "do-nightly-rescore-example" message is usually at least eight
hours before the last "Received:" header (taking the declared time zones
into account, of course).  So either one of the machines involved has a
really bad RTC chip, or that message is delayed and actually refers to
the previous day?

Regards,
Marc

Reply via email to