Spoke too soon. It just took 4 hours longer than 2 days to exhibit the behavior. Is anyone successfully running SpamAssassin 2.43 under perl 5.8?
Debugging output doesn't give any indication as to a source of the problem, so I'm starting to reach for other possibilities... -James -----Original Message----- From: James Bly [mailto:[EMAIL PROTECTED]] Sent: Tuesday, November 05, 2002 10:37 AM To: '[EMAIL PROTECTED]' Subject: RE: [SAtalk] Odd intermittent failure, (RH8, SA 2.42, Perl 5.8) - Memory Leak? Fixed thus far, but why is not clear. The fix was to move my whitelist information and configuration information out of the files in /usr/local/share/spamassassin/*.cf and put it into /etc/mail/spamassassin/local.cf. I moved 3k worth of configuration data from these files (around 80 whitelist entries). I was previously placing them in 60_whitelist.cf because that's how I'd been doing it since the v1.20 days. (The rest of the lines were the rewrite_header, defang_mime etc that I used to set in 10_local.cf.) Is it really that strange that this would cause spamd to get cranky? -James -----Original Message----- From: Matt Kettler [mailto:[EMAIL PROTECTED]] Sent: Tuesday, October 29, 2002 4:43 PM To: James Bly Subject: RE: [SAtalk] Odd intermittent failure, (RH8, SA 2.42, Perl 5.8) - Memory Leak? Hmm, you are correct on that one.. There are still some spamd issues fixed in 2.43, one of which is a "critical reason to upgrade" type bug for all users of spamd. The release announcement for 2.43 indicated: Changes: - AWL change reverted; instead of decreasing the AWL bias gradually to allow frequently-seen addresses to get into the "nonspam" area, it now behaves like 2.31 did, in that the AWL simply represents the long-term average score from that correspondent. - core-dump bug in spamd worked around, _except for the "-m" switch_. The "-m" switch relies on signal handling in the Perl interpreter, which seems to have some bugs we cannot work around reliably on some platforms, so its use is no longer recommended. - some portability fixes for SunOS. The core-dump bug seems to be this one: http://www.hughes-family.org/bugzilla/show_bug.cgi?id=1011 Which is about spamd not cleaning up all it's children and eventually core dumping. Even if this doesn't fix your specific problem, I'd strongly recommend the upgrade if you're using either spamd (which you are) or the AWL. At 03:37 PM 10/29/2002 -0600, James Bly wrote: >I can give 2.43 a round, but I'm having trouble understanding the >connection. This bug was in relation to the file descriptor not being >closed properly under spamc where as the memory leak I'm having is >under spamd... > >-James > >-----Original Message----- >From: Matt Kettler [mailto:[EMAIL PROTECTED]] >Sent: Tuesday, October 29, 2002 3:12 PM >To: James Bly; '[EMAIL PROTECTED]' >Subject: RE: [SAtalk] Odd intermittent failure, (RH8, SA 2.42, Perl >5.8) - Memory Leak? > > >Wait a second.. You're using 2.42.... see this bug > >http://www.hughes-family.org/bugzilla/show_bug.cgi?id=1128 > >This bug appears to be fixed in 2.43. > > >At 02:17 PM 10/29/2002 -0600, James Bly wrote: > >It looks like it may be some sort of memory leak in spamd. > > > >A long ps gave me some additional relevant details to the problem. As > >you can see, two spamd sessions are in lock_p state. There are also > >two spamc connections in FIN_WAIT2. > > > >000 S qd 22432 22431 0 75 0 - 331 wait4 13:49 ? > >00:00:00 /var/maildir/bin/qmail-qfilter /usr/bin/spamc -p > >000 S qd 22352 22351 0 76 0 - 331 wait4 13:42 ? > >00:00:00 /var/maildir/bin/qmail-qfilter /usr/bin/spamc -p > >000 S qd 22432 22431 0 75 0 - 331 wait4 13:49 ? > >00:00:00 /var/maildir/bin/qmail-qfilter /usr/bin/spamc -p > >000 S qd 22352 22351 0 76 0 - 331 wait4 13:42 ? > >00:00:00 /var/maildir/bin/qmail-qfilter /usr/bin/spamc -p > >000 S qd 22351 5516 0 75 0 - 333 wait4 13:42 ? > >00:00:00 /var/maildir/bin/qmail-smtpd > >000 S qd 22431 5516 0 75 0 - 333 wait4 13:49 ? > >00:00:00 /var/maildir/bin/qmail-smtpd > >140 S spamd 910 1 0 75 0 - 4241 schedu Oct24 ? > >00:00:40 /usr/bin/perl /usr/bin/spamd -u spamd -p 1212 > >040 D spamd 22354 910 19 75 0 - 33029 lock_p 13:42 ? > >00:01:14 /usr/bin/perl /usr/bin/spamd -u spamd -p 1212 > >040 D spamd 22434 910 2 75 0 - 4299 lock_p 13:49 ? > >00:00:00 /usr/bin/perl /usr/bin/spamd -u spamd -p 1212 > >000 S qd 22433 22432 0 76 0 - 475 schedu 13:49 ? > >00:00:00 /usr/bin/spamc -p 1212 > >000 S qd 22353 22352 0 76 0 - 475 schedu 13:42 ? > >00:00:00 /usr/bin/spamc -p 1212 > >000 R jbly 22436 22386 0 76 0 - 46 - 13:49 pts/2 > >00:00:00 grep spam > > > >[jbly@boink jbly]$ ps -eo pid,min_flt,maj_flt,cmd | grep spam > > 910 547971 15718 /usr/bin/perl /usr/bin/spamd -u spamd -p 1212 -d > >22352 36 316 /usr/mail/bin/qmail-qfilter /usr/bin/spamc -p 1212 > >22353 22 152 /usr/bin/spamc -p 1212 > >22354 71900 197674 /usr/bin/perl /usr/bin/spamd -u spamd -p 1212 -d > >22847 36 316 /usr/mail/bin/qmail-qfilter /usr/bin/spamc -p 1212 > >22848 17 122 /usr/bin/spamc -p 1212 > >22849 1383 1470 /usr/bin/perl /usr/bin/spamd -u spamd -p 1212 -d > >22854 37 148 grep spam > > > >That 71900 minor page faults and and 197674 major is rather > >concerning to me. (Oh and for the record yes, the system is crawling > >right now.) > > > >I also now see why it's intermittent. Spamd is hitting its bounds for > >memory allocation and gets killed. However this will not happen if > >enough of these stray spamd processes get run at once. The system > >simply can't recover fast enough at that point. > > > >So now I need to figure out where the leak is. Stay tuned. Same > >bat-time, same bat-channel. > > > >-James > > > > > -----Original Message----- > > > ... > > > For your problem, first check any local IPTables firewall rules > > > you have > >an > > > make sure you're not doing something silly that will block fin > > > packets > >that > > > don't have the ack bit set. > > > ... > > > > > > > > >------------------------------------------------------- > >This sf.net email is sponsored by:ThinkGeek > >Welcome to geek heaven. > >http://thinkgeek.com/sf > >_______________________________________________ > >Spamassassin-talk mailing list > >[EMAIL PROTECTED] > >https://lists.sourceforge.net/lists/listinfo/spamassassin-talk ------------------------------------------------------- This sf.net email is sponsored by: See the NEW Palm Tungsten T handheld. Power & Color in a compact size! http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk ------------------------------------------------------- This sf.net email is sponsored by: See the NEW Palm Tungsten T handheld. Power & Color in a compact size! http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
