Bug 62854 - event.c disabled listensocks and after while nobody is left
Summary: event.c disabled listensocks and after while nobody is left
Status: RESOLVED INVALID
Alias: None
Product: Apache httpd-2
Classification: Unclassified
Component: mpm_event (show other bugs)
Version: 2.4.34
Hardware: Sun Solaris
: P2 normal (vote)
Target Milestone: ---
Assignee: Apache HTTPD Bugs Mailing List
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2018-10-25 12:10 UTC by Petr Sumbera
Modified: 2018-10-25 14:04 UTC (History)
0 users



Attachments
Log file where you can see how processes are disabled and never used again. (36.75 KB, text/plain)
2018-10-25 12:23 UTC, Petr Sumbera
Details

Note You need to log in before you can comment on or make changes to this bug.
Description Petr Sumbera 2018-10-25 12:10:23 UTC

    
Comment 1 Petr Sumbera 2018-10-25 12:18:15 UTC
We are getting regularly Apache HTTPD into state where ALL processes are not accepting new connections. And it's never accepting them again.

..
[Thu Oct 25 11:53:34.198518 2018] [mpm_event:debug] [pid 1954:tid 28] event.c(1844): Too many open connections (26), not accepting new conns in this process
[Thu Oct 25 11:53:34.210734 2018] [mpm_event:debug] [pid 1954:tid 28] event.c(483): AH00457: Accepting new connections again: 3 active conns (1 lingering/0 clogged/0 suspended), 22 idle workers
[Thu Oct 25 11:53:34.211408 2018] [mpm_event:debug] [pid 1954:tid 28] event.c(1844): Too many open connections (2), not accepting new conns in this process
[Thu Oct 25 11:53:39.197077 2018] [mpm_event:debug] [pid 1954:tid 28] event.c(483): AH00457: Accepting new connections again: 0 active conns (0 lingering/0 clogged/0 suspended), 25 idle workers
[Thu Oct 25 11:53:39.198114 2018] [mpm_event:debug] [pid 1954:tid 28] event.c(1844): Too many open connections (2), not accepting new conns in this process
Comment 2 Petr Sumbera 2018-10-25 12:23:32 UTC
Created attachment 36215 [details]
Log file where you can see how processes are disabled and never used again.
Comment 3 Yann Ylavic 2018-10-25 12:43:20 UTC
Can you please share your setting for ThreadsPerChild and AsyncRequestWorkerFactor (if any), and the version of the APR library?

Also (or after) if possible, could you provide the same log file with LogLevel trace6 (at global scope only)?
Comment 4 Petr Sumbera 2018-10-25 12:49:30 UTC
It's default configuration (no ThreadsPerChild nor AsyncRequestWorkerFactor setting).

Server loaded:  APR 1.6.3, APR-UTIL 1.6.1
Compiled using: APR 1.6.3, APR-UTIL 1.6.1
Comment 5 Yann Ylavic 2018-10-25 13:00:25 UTC
Could you update to APR-1.6.5 and notably with r1819938?

I think you need this change for MPM event's wakeup to work properly on Solaris.
Comment 6 Petr Sumbera 2018-10-25 13:06:38 UTC
I have temporary placed requested log file here: http://hostivar.ml/error_log
Comment 7 Petr Sumbera 2018-10-25 14:04:15 UTC
With following I don't seem to be able to reproduce the issue. Thank you!

Server loaded:  APR 1.6.5, APR-UTIL 1.6.1
Compiled using: APR 1.6.3, APR-UTIL 1.6.1