Spamd Child Forks

[Ray Dzek <Ra...@specialized.com>]

spamd: handled cleanup of child pid 19888 due to SIGCHLD
spamd[5262]: prefork: child states: II
 
I see a lot of mention of these messages showing up in 3.1, but I have
not seen a definitive answer as to if I need to pay more attention to
them or they are just a minor/visual annoyance in the logs and will be
cleaned up with a patch or the next release.
 
Some clarification, please, from those in the know?
 
Thanks!
 


Ray Dzek
Net Ops / Helpdesk Supervisor
Specialized Bicycle Components 

 

Re: Spamd Child Forks, and scaling problems

[Andrew Donkin <ar...@waikato.ac.nz>]

Ray wrote:

>> spamd: handled cleanup of child pid 19888 due to SIGCHLD
>> spamd[5262]: prefork: child states: II

Ray, the one you want to look for is "prefork: server reached
--max-clients setting, consider raising it".  If you see that, you may
want to look at your free RAM and, well, consider using
--max-CHILDREN.  There have been posts here about how to do that
properly (to summarise: gradually).

Mike replied:

> It's log noise. Nothing to be alarmed about.

The signal notification could be relegated to a debug level but the
second of those lines can be useful when diagnosing spamd's problems.
Example:

20:39:53 prefork: child states: IIIIIIIIIIIIIIIIIIII 
20:39:59 prefork: child states: IBBBBBBIIIIIIIIIIIII 
20:40:06 prefork: child states: BBBBBBBBBBBBBBBBBBBB 
20:40:07 prefork: child states: BBBBBBIBBBBBBBBBBBBBB 
20:40:09 prefork: child states: BBBBBBBIBBBBBBBBBBBBBB 
20:40:10 prefork: child states: IBBBBBBBIBBBBBBBBBBBBBB 
20:40:13 prefork: child states: BIBBBBBBBIBBBBBBBBBBBBBB 
20:40:15 prefork: child states: BBBBIBBBBBIBBBBBIBBBIIIBI 
20:40:18 prefork: child states: BBBBBBBBBBBIBIBIBBIIIBBBBB 
20:40:22 prefork: child states: BBBBBBBBBBBBBBBBBBBBBBBBBBB 
20:40:31 prefork: child states: BBBIBBBBBBBBBBIBBBBBBBBBBBBB 
20:40:36 prefork: child states: BBBIBBBBBBBBBBIBBBBBBBBBBBBB 

This continues, until (columns removed to reduce line length):

20:51:41 BBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBB 
20:51:43 BBBBBIBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBB 
20:51:43 BBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBB 
20:51:43 BBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBB 
20:51:44 BBBBBBBBIBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBB 
20:51:45 BBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBB 
20:51:47 BBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBB 
20:51:49 BBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBB 
20:51:50 BIBBBBBBBBBBIBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBIB 
20:51:52 BBBBBBBBBBBBBIIIBBBIBIBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBB 

While this is going on, spamc is timing out and letting email through
unchecked.  spamd goes to some effort to not panic-spawn children, so
I have circumvented it by reverting to the old algorithm via
--round-robin.

-- 
_________________________________________________________________________
Andrew Donkin                  Waikato University, Hamilton,  New Zealand

Re: Spamd Child Forks

[Mike Jackson <mj...@barking-dog.net>]

> spamd: handled cleanup of child pid 19888 due to SIGCHLD
> spamd[5262]: prefork: child states: II
>
> I see a lot of mention of these messages showing up in 3.1, but I have
> not seen a definitive answer as to if I need to pay more attention to
> them or they are just a minor/visual annoyance in the logs and will be
> cleaned up with a patch or the next release.

It's log noise. Nothing to be alarmed about. By default, 3.1 keeps two child 
processes running, and one as a spare. If it's processing two messages at a 
time, it will start up another child as a spare. It will keep doing that 
until you've hit the max-children setting. When there's more than two idle 
processes again, it will kill them off until there's only two. That's what 
you're seeing.