You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@spamassassin.apache.org by bu...@bugzilla.spamassassin.org on 2005/10/10 19:44:05 UTC
[Bug 4624] New: spamd crashes under heavy load after upgrade.
http://bugzilla.spamassassin.org/show_bug.cgi?id=4624
Summary: spamd crashes under heavy load after upgrade.
Product: Spamassassin
Version: 3.1.0
Platform: DEC
OS/Version: OSF/1
Status: NEW
Severity: normal
Priority: P5
Component: spamc/spamd
AssignedTo: dev@spamassassin.apache.org
ReportedBy: cantelow@csd.net
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.
[Bug 4624] spamd crashes under heavy load after upgrade.
Posted by bu...@bugzilla.spamassassin.org.
http://issues.apache.org/SpamAssassin/show_bug.cgi?id=4624
------- Additional Comments From jm@jmason.org 2007-04-16 05:31 -------
btw it might be worth trying 3.2.0 rc2 or later -- there are a number of fixes
in the prefork code
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.
[Bug 4624] spamd crashes under heavy load after upgrade.
Posted by bu...@bugzilla.spamassassin.org.
http://bugzilla.spamassassin.org/show_bug.cgi?id=4624
------- Additional Comments From jm@jmason.org 2005-10-10 11:36 -------
If you can get a system-call trace (ie. from "strace") of both the parent and
child processes (preferably something like "strace -f" output), that would be
very helpful.
BTW --round-robin should certainly work around it, and under your levels of load
is probably better anyway -- assuming you have enough RAM for the max number of
children to be in RAM simultaneously.
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.
[Bug 4624] spamd crashes under heavy load after upgrade.
Posted by bu...@bugzilla.spamassassin.org.
http://bugzilla.spamassassin.org/show_bug.cgi?id=4624
------- Additional Comments From cantelow@csd.net 2005-10-10 11:10 -------
Addendum #2, sorry:
Our spamd switches at time of crash:
/usr/local/bin/spamd -D -m 90 -d -p 22874 -x -u spamasn -i 0.0.0.0 -s local3 -A
127.0.0.1,204.181.152.50,204.181.152.5
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.
[Bug 4624] spamd crashes under heavy load after upgrade.
Posted by bu...@bugzilla.spamassassin.org.
http://bugzilla.spamassassin.org/show_bug.cgi?id=4624
------- Additional Comments From cantelow@csd.net 2005-10-11 08:07 -------
(In reply to comment #4)
I'll see if I can get a trace today, thanks. Interestingly, the number of spamd
sessions stay down below 10 for most of the time but several times a day we
apparently get huge but brief waves of activity. Having the client states log
output has been an interesting education. -Ann
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.
[Bug 4624] spamd crashes under heavy load after upgrade.
Posted by bu...@bugzilla.spamassassin.org.
http://bugzilla.spamassassin.org/show_bug.cgi?id=4624
------- Additional Comments From cantelow@csd.net 2005-10-10 10:56 -------
Sorry to have sent this as a blank message before by accident.
We're on HP Tru64 version 5.1.
Upgraded from 3.0.4 to 3.1.0 and started having 1-4 crashes a day.
In our local.cf file:
use_auto_whitelist 0
lock_method flock
We were having crashes when the number of processes in use reached near the
number in the -m switch in spamd. We increased this gradually from 30 to 90.
The last time, spamd still crashed at 50 processes, though I had seen it at 70
processes earlier. We added --round-robin to our spamd switches, and no crashes
have happened in the 18 hours since, so we're doing fine now.
In our log, grepping for prefork, this activity in the last 5 minutes before a
crash without --round-robin may have bearing:
Oct 9 12:53:14 atlas spamd[366464]: prefork: child closed connection
Oct 9 12:53:14 atlas spamd[366464]: prefork: child states:
BBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBIIBBBBBKB
Oct 9 12:53:14 atlas spamd[227559]: prefork: periodic ping from spamd parent
Oct 9 12:53:14 atlas spamd[227559]: prefork: sysread(9) not ready, wait max 300
secs
Oct 9 12:53:14 atlas spamd[167740]: prefork: periodic ping from spamd parent
Oct 9 12:53:14 atlas spamd[167740]: prefork: sysread(43) not ready, wait max
300 secs
Oct 9 12:53:15 atlas spamd[235278]: prefork: periodic ping from spamd parent
Oct 9 12:53:15 atlas spamd[235278]: prefork: sysread(47) not ready, wait max
300 secs
Oct 9 12:53:17 atlas spamd[155119]: prefork: periodic ping from spamd parent
Oct 9 12:53:17 atlas spamd[155119]: prefork: sysread(18) not ready, wait max
300 secs
Oct 9 12:53:17 atlas spamd[258179]: prefork: periodic ping from spamd parent
Oct 9 12:53:17 atlas spamd[258179]: prefork: sysread(24) not ready, wait max
300 secs
... and more, the rest of the processes?
We normally get the sysread-not-ready messages sporadically, but this time it
was a big group of them together in 5 mins.
And many msgs like these, over about 5 seconds:
Oct 9 12:53:57 atlas spamd[253409]: prefork: parent closed, exiting
Oct 9 12:53:57 atlas spamd[197507]: prefork: parent closed, exiting
Oct 9 12:53:57 atlas spamd[246309]: prefork: parent closed, exiting
Oct 9 12:53:57 atlas spamd[246991]: prefork: parent closed, exiting
Oct 9 12:53:57 atlas spamd[250496]: prefork: parent closed, exiting
Oct 9 12:53:57 atlas spamd[159250]: prefork: parent closed, exiting
Oct 9 12:53:57 atlas spamd[258960]: prefork: parent closed, exiting
...
But there was a periodic ping from parent afterwards:
Oct 9 12:54:00 atlas spamd[227559]: prefork: parent closed, exiting
Oct 9 12:54:00 atlas spamd[16779]: prefork: parent closed, exiting
Oct 9 12:57:01 atlas spamd[13042]: prefork: periodic ping from spamd parent
Oct 9 12:57:01 atlas spamd[13042]: prefork: parent closed, exiting
12:57 is when all activity in the log stops with a crash.
Hope that helps. Thanks tons for all your work.
Ann
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.
[Bug 4624] spamd crashes under heavy load after upgrade.
Posted by bu...@bugzilla.spamassassin.org.
http://bugzilla.spamassassin.org/show_bug.cgi?id=4624
------- Additional Comments From cantelow@csd.net 2005-10-10 10:58 -------
Forgot to mention- our perl version is 5.6.1. It is compiled under native
cc, as is our spamassassin package.
Thanks, Ann
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.