-------- Original Message -------- Subject: Re: very basic SA-Learn performance question: is 90 seconds or so per token really, really slow or roughly normal? From: David Jones <djo...@ena.com> To: users@spamassassin.apache.org Date: Sat Nov 04 2017 16:35:02 GMT+0300 (AST)
> On 11/04/2017 08:09 AM, David Gessel wrote: >> so days later, still chunking away, not making much progress. >> >> If I kill the process (doesn't stop sa-learn, just kills current script), it >> always returns >> ^Cplugin: eval failed: interrupted at /usr/local/bin/sa-learn line 511. >> >> which is >> >> 0509 sub killed { >> 0510 $spamtest->finish_learner(); >> 0511 die "interrupted"; >> 0512 } >> >> The only difference in sa-learn I'm running from 3.4.1 at >> https://svn.apache.org/repos/asf/spamassassin/tags/spamassassin_release_3_4_1/ >> >> is line 50 >> 0050 $searchrelative = 1; # disabled during "make install": >> REMOVEFORINST >> (which I assume is removed given "REMOVEFORINST") >> >> So I assume given the changes in lines 19-21, that my server is running >> 3.4.1 release. >> >> I note that 3.4.2p3 has one difference from 3.4.1, which is comment out use >> bytes; at line 21 (this has been there or not there a few times over various >> versions and so may be slightly meaningful to something) >> >> 0021 # use bytes; >> >> I'm not sufficiently perl savvy to have any idea whether that's useful to my >> performance issues or not, but it an easy enough mod to try. >> >> Any thoughts? >> > > Can you setup a generic SA installation on your desktop/laptop/VM (some other > location), copy your ham/spam corpus, and try to train the bayes on another > instance? This may take a little effort but prove insightful. > > If that proves to be fast, then you could copy over your > /etc/mail/spamassassin directory and see if that causes the same slow down > which would point to some setting in /etc/mail/spamassassin. You may want to > backup the /etc/mail/spamassassin before copying over yours for reference > just in case you need to diff some config files later. > > P.S. A very fast, easy way to setup a test VM with SA (plus Postfix, Dovecot, > Roundcube webmail, etc.) is http://iredmail.org. > Smol update. I updated the whole system (portsnap update, portmaster -Rafd) and after getting through a few update quirks (DEFAULT_VERSIONS+=ssl=openssl warning is no longer optional...), everything is working. This updates: spamassassin-3.4.1_11 perl5-5.24.3 among others. And which brought me back to this: Use of uninitialized value $msgscore in addition (+) at /usr/local/lib/perl5/site_perl/Mail/SpamAssassin/Plugin/TxRep.pm line 1415. Use of uninitialized value $msgscore in subtraction (-) at /usr/local/lib/perl5/site_perl/Mail/SpamAssassin/Plugin/TxRep.pm line 1415. which this version fixes https://svn.apache.org/viewvc/spamassassin/trunk/lib/Mail/SpamAssassin/Plugin/TxRep.pm?view=markup&pathrev=1720440 but which hasn't been rolled into the port yet. Perhaps because it is really slow? I'll try to do some timing experiments with the installed version vs. the rev 1720440 version tomorrow (2:49 am my time now). -David