Bug 65592 - mpm_event does not create new child after MaxConnectionsPerChild is hit
Summary: mpm_event does not create new child after MaxConnectionsPerChild is hit
Status: RESOLVED FIXED
Alias: None
Product: Apache httpd-2
Classification: Unclassified
Component: mpm_event (show other bugs)
Version: 2.4.49
Hardware: PC Linux
: P2 regression with 3 votes (vote)
Target Milestone: ---
Assignee: Apache HTTPD Bugs Mailing List
URL:
Keywords: FixedInTrunk, PatchAvailable
Depends on:
Blocks:
 
Reported: 2021-09-22 15:38 UTC by Matt Zuba
Modified: 2021-09-27 08:40 UTC (History)
1 user (show)


Attachments
Correctly account processes (976 bytes, patch)
2021-09-22 16:08 UTC, Ruediger Pluem 		Details | Diff
View All 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 Matt Zuba 2021-09-22 15:38:06 UTC 
I've encountered issues on multiple servers running Apache 2.4.49 where we have MaxConnectionsPerChild set for the event mpm.  When child processes die after this limit his hit, the mpm is not creating new processes which stop Apache from serving any requests.  This issue did not exist in 2.4.48.

The easiest way for me to reproduce with the most simplistic example is with the httpd docker images.  When modifying the mpm_event conf and setting all values to 1 (for ease of testing), 2.4.49 will cease to serve any requests, whereas 2.4.48 will continue to serve request after request.

=====================
<IfModule mpm_event_module>
    StartServers             1
    MinSpareThreads         1
    MaxSpareThreads        1
    ThreadsPerChild         1
    MaxRequestWorkers      1
    MaxConnectionsPerChild   1
</IfModule>
=====================

I can produce some repos with Dockerfiles in them if necessary.
Comment 1 Ruediger Pluem 2021-09-22 16:08:46 UTC 
Created attachment 38046 [details]
Correctly account processes

Can you please try if the attached patch fixes your issue?
Comment 2 Matt Zuba 2021-09-22 17:51:29 UTC 
Thank you for the quick patch.  I've applied it on top of the Apache source for Ubuntu from Ondřej Surý found at https://launchpad.net/~ondrej/+archive/ubuntu/apache2 and it has solved the problem in 2.4.49.
Comment 3 Ruediger Pluem 2021-09-22 18:51:25 UTC 
Committed to trunk a r1893520.
Comment 4 Graham Leggett 2021-09-26 14:36:30 UTC 
Backported to v2.4.50.
Comment 5 jerry 2021-09-27 02:35:16 UTC 
(In reply to Matt Zuba from comment #2)
> Thank you for the quick patch.  I've applied it on top of the Apache source
> for Ubuntu from Ondřej Surý found at
> https://launchpad.net/~ondrej/+archive/ubuntu/apache2 and it has solved the
> problem in 2.4.49.

This sounds exactly like what I am experiencing with my recent upgrade to 2.4.49. I submitted my issue to the mailing list (https://httpd.markmail.org/search/?q=#query:+page:1+mid:mwn5orzpqoq4vwii+state:results).

May I ask how exactly you "applied the patch on top of the Apache source"?

Also, what are the chances of an immediate release with this patch given the significance of this issue in production servers?

Thanks!
Comment 6 jerry 2021-09-27 08:40:59 UTC 
I found instructions to patch Apache here:
https://www.linuxquestions.org/questions/solaris-opensolaris-20/steps-to-apply-the-patch-apache-918907/#post4551988

Thanks for the patch!