Bug 3857 - Spamd repeatedly causing unacceptably high system load
Summary: Spamd repeatedly causing unacceptably high system load
Status: RESOLVED FIXED
Alias: None
Product: Spamassassin
Classification: Unclassified
Component: spamc/spamd (show other bugs)
Version: 3.0.0
Hardware: Other Linux
: P5 critical
Target Milestone: Undefined
Assignee: SpamAssassin Developer Mailing List
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2004-10-02 02:13 UTC by Martin
Modified: 2005-01-21 03:18 UTC (History)
0 users


Attachment Type Modified Status Actions Submitter/CLA Status
Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Martin 2004-10-02 02:13:20 UTC 
I keep seeing logcheck emails telling me system load is too high and on 
investigation, it appears its spamd causing the problem. This is a relatively 
low load server in terms of mail volume. It's a 1.7Ghz P4 with 256MB RAM. This 
is happening regularly and was also an issue with 2.63 before I upgraded to 
3.0.0. Most recently it caused system load to remain at 30-34 for seven or 
eight hours till I noticed the problem this morning. Even on start up, each 
thread/child appears to be using 10-15% mem, with four-six threads/childs.
Comment 1 Theo Van Dinter 2004-10-02 11:02:09 UTC 
Subject: Re:  New: Spamd repeatedly causing unacceptably high system load

On Sat, Oct 02, 2004 at 02:13:21AM -0700, bugzilla-daemon@bugzilla.spamassassin.org wrote:
>          Component: spamc/spamd
> 
> is happening regularly and was also an issue with 2.63 before I upgraded to 
> 3.0.0. Most recently it caused system load to remain at 30-34 for seven or 
> eight hours till I noticed the problem this morning. Even on start up, each 
> thread/child appears to be using 10-15% mem, with four-six threads/childs.

Hrm.  First, there is no threading involved running spamd -- it's all process
based.  Second, there's no way SA can generate a load of 30+ if there's only
4-6 spamd processes running.

The issue you describe sounds like you are resource strapped, thereby
causing other processes to sit waiting for swap, driving the load up, etc.

The easy fixes are: put more memory in the machine, turn off other
processes on the machine you don't use, cut back on the number of spamd
children, disable SA features you don't use, disable third-party rulesets
if installed...
Comment 2 Martin 2004-10-02 11:13:54 UTC 
I didn't mean threads, I meant the child processes in 3.0.0. The system may 
indeed have become resource strapped, resulting in the high load, but killing 
SA solves the problem, at least until it happens again. There's not much 
running on the box and resources have never really been a problem, but it 
seems to have gotten gradually worse as the I upgraded each version of SA, 
with 3.0 being the worst.
Comment 3 Dallas Engelken 2004-10-08 12:34:29 UTC 
can you look at bug 3776 and see if that patch fixes your problem?
Comment 4 Martin 2004-10-09 06:02:24 UTC 
patch applied. Now to monitor and wait and see...
Comment 5 Martin 2004-10-14 19:21:24 UTC 
Seeing some 

Oct 15 02:30:14 Socrates spamd[28424]: server hit by SIGCHLD 
Oct 15 02:30:14 Socrates spamd[28424]: handled cleanup of child pid 29000 
Oct 15 02:30:14 Socrates spamd[28424]: server successfully spawned child 
process, pid 16738 
Oct 15 02:30:19 Socrates spamd[28424]: server hit by SIGCHLD 
Oct 15 02:30:19 Socrates spamd[28424]: handled cleanup of child pid 30344 
Oct 15 02:30:19 Socrates spamd[28424]: server successfully spawned child 
process, pid 16768 
Oct 15 02:40:13 Socrates spamd[28424]: server hit by SIGCHLD 
Oct 15 02:40:13 Socrates spamd[28424]: handled cleanup of child pid 30379 
Oct 15 02:40:13 Socrates spamd[28424]: server successfully spawned child 
process, pid 17378 
Oct 15 02:50:09 Socrates spamd[28424]: server hit by SIGCHLD 
Oct 15 02:50:09 Socrates spamd[28424]: handled cleanup of child pid 29766 
Oct 15 02:50:09 Socrates spamd[28424]: server successfully spawned child 
process, pid 18148 

in syslog. I've seen a spot of high load here and there, but not like it was.
Comment 6 Justin Mason 2005-01-21 12:18:32 UTC 
those messages are entirely normal.  marking FIXED