httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Luca Toscano <>
Subject Re: svn commit: r1770771 - /httpd/httpd/trunk/docs/manual/mod/mpm_common.xml
Date Thu, 24 Nov 2016 17:24:48 GMT
Reporting a summary of various chat between me and Daniel over IRC. There
are two things that would need to be documented if I got the code correctly:

1) ServerLimit is now used as upper bound for the sum of total child
processes doing a graceful close plus the active ones serving requests (so
not only MaxRequestWorkers / ThreadsPerChild anymore). This has been
documented and it is clear enough once somebody gets the full picture.
2)  Aggressive process reclaim caused by MaxSpareThreads in special cases
like high traffic load with fluctuations (increase causing more processes
to be spawned, decrease due to temporary MaxSpareThreads limit breached by
lower traffic, then increase again, etc..). With Stefan's changes mpm-event
should not reclaim process slots in these situations, showing it up in the
error log ("Not shutting down child: etc..") and in mod_status (still need
to re-read the patch to have a better idea though).

I think that having more details/comments/notes in ServerLimit might be
confusing, as well as in MaxSpareThreads, so I suppose that we could add a
separate section in the mpm-event's documentation mentioning this new
behavior. If nobody opposes I'd like to add these info to the doc page in
trunk (with corrections from people if what I've said is not true of



2016-11-23 13:13 GMT+01:00 Daniel <>:

> Hello Luca,
> note that "scoreboard is full" is caused for not being enough
> maxsparethreads available and event being constantly reclaming threads, so
> for the case you mention, the comment should rather go on MaxSpareThreads
> rather than ServerLimit.
> 2016-11-23 11:53 GMT+01:00 Luca Toscano <>:
>> Hi Stefan,
>> 2016-11-21 23:13 GMT+01:00 <>:
>>> Author: sf
>>> Date: Mon Nov 21 22:13:05 2016
>>> New Revision: 1770771
>>> URL:
>>> Log:
>>> Describe new behavior of ServerLimit
>>> This has changed a bit after the fix for PR 53555. Better wording is
>>> welcome.
>> first of all thanks a lot for this big effort! I am still reading your
>> code and trying to understand how it works in detail, but I have some
>> suggestions on how to improve the docs:
>> - in
>> it would be great to have a little note section after the "With event,
>> increase this directive if the process number.." with a little background
>> about why event is special. Even something trivial as mentioning that
>> everything originated from reports of the "scoreboard is full, not at
>> MaxRequestWorkers" error could direct people from the search engines to the
>> httpd manual pages rather than random blogs.
>> - Some summaries and explanation that you wrote like
>> (I know that
>> something has been changed in the meantime) would be really great to be
>> captured somewhere, maybe in
>> cs/trunk/mod/event.html.
>> I am happy to help with some work (even after the votes) but since you
>> know really well the code there is a very high chance that the
>> documentation will be more accurate and reliable :)
>> Thanks!
>> Luca
> --
> *Daniel Ferradal*
> IT Specialist
> email         dferradal at
> linkedin

View raw message