Server won't accept more than 128 connections... (tried everything)

Drowned

Member
Dec 5, 2003
19
0
151
Th new EasyApache system is kind of confusing to me. Before I would just open up httpd.conf and change stuff, save, restart apache and be done.

But I tried this, did the "apache_conf_distiller --update" and the "build_apache_conf" without any changes in my status... is there something else I'm missing? It's showing "128 requests currently being processed" all the time and my CPU is almost idling and I have tons of ram available... I'm trying to set it to 250 clients. My httpd.conf file DOES show 250 but apache is still being limited to 128...

StartServers 10
MinSpareServers 10
MaxSpareServers 15
ServerLimit 250
MaxClients 250
MaxRequestsPerChild 1000

I read about SERVER_HARD_LIMIT but have no idea where it is or what file should I add it to :(
 
Last edited:

Voltar

Well-Known Member
Apr 30, 2007
267
0
168
Bakersfield, California
Which version of Apache are you using? There is a compiled in SERVER_HARD_LIMIT in Apache 1.3, but not in 2.x. After rebuilding httpd.conf did you check to make sure your changes were retained?
 

Drowned

Member
Dec 5, 2003
19
0
151
Which version of Apache are you using? There is a compiled in SERVER_HARD_LIMIT in Apache 1.3, but not in 2.x. After rebuilding httpd.conf did you check to make sure your changes were retained?
Server version: Apache/2.2.11 (Unix)

Yeah, after doing some reading I found out 2.x doesn't have one.

Yes, changes are being retained but ignored. I checked my error_log and no errors are given when I restart.
 

Drowned

Member
Dec 5, 2003
19
0
151
I just noticed that the setting IS being used, but apache is only working with 128 requests all the time and leaving the other as open slots.

This is how my status looks like all the time :ugh:

Code:
WWWCCKKCKKKCKKCKKCCCWCCCKKCWKKKCWKCWCKWWWWCCCKWCKKCKWKKCKCKKKCKK
CCWWWKCKCWCCWWWKKCCKKCWKWKWKKKWWWWKWWCKWKCWWKKWWKKWKWCCCKWWKCWCK
................................................................
........
I have thousands of requests being put on hold although apache still has more than 100 open slots?! :eek: Why would this happen? I have over 2GB of free ram and the cpu is almost idling with just 128 connections :confused:
 

Drowned

Member
Dec 5, 2003
19
0
151
Come on, somebody has to know about some kind of limit that will rule over MaxClients and ServerLimit :( :(:(:(:(:(
 

Drowned

Member
Dec 5, 2003
19
0
151
In case this happens to you, mod_proxy was the one at fault. It was giving me segmentation fault errors and killing the extra connections. I simply compiled apache without it and everything went back to normal :)
 

Voltar

Well-Known Member
Apr 30, 2007
267
0
168
Bakersfield, California
In case this happens to you, mod_proxy was the one at fault. It was giving me segmentation fault errors and killing the extra connections. I simply compiled apache without it and everything went back to normal :)
If you disable mod_proxy you'll lose the virtualhost proxy subdomains for WHM, cPanel, webdisk and webmail.

Did you get any coredumps and if so did you run gdb on them to see what happened?
 

Drowned

Member
Dec 5, 2003
19
0
151
If you disable mod_proxy you'll lose the virtualhost proxy subdomains for WHM, cPanel, webdisk and webmail.

Did you get any coredumps and if so did you run gdb on them to see what happened?
The log said this:

Code:
[debug] proxy_util.c(1625): proxy: grabbed  
scoreboard slot 0 in child 1380 for worker proxy:reverse
[Fri Dec 22 14:15:37 2006] [debug] proxy_util.c(1644): proxy: worker  
proxy:reverse already initialized
and then came a segmentation fault error:

Code:
[notice]child pid 1644 exit signal Segmentation fault (11)
Do I really need mod_proxy?
 

Voltar

Well-Known Member
Apr 30, 2007
267
0
168
Bakersfield, California
If you or any of your users use the proxy subdomains because they are behind a firewall or for any other reason, cpanel.example.com or webmail.example.com won't work as you don't have the proxy module installed.

Do you have any coredumps to examine? A segfault can also be caused by bad/failing memory, so that might be something to look into.