einstein1981
Guru
Joined: 10 Mar 2004
Posts: 395
Location: Venezuela , Caracas
|
 Posted: Fri Mar 27, 2009 7:40 pm Post subject: apache2 + php configs +kannel bridging |
 |
|
hello,
I'm running a server with a simple site that is used for making an smpp/customhttp service bridge
Basically I have 2 servers, one (host A)is running kannel ( smsc) and is connected with a MNO, which is sending us some sms, I have kannel posting these messages to my webserver on host B, that basically proceses these sms and formats them according to a specification, and then POSTS ( via curl_lib) to another remote host (that is not under my control). ( hope i ve not lost you yet).
This remote host after some procesing returns some results to host B ( another php script) that re-interprets it ( pretty simple) and sends it back to kannel on host A to send back to the mobile.
So basically its a 3 host route, 2way.
Everything works properly, except when the throughput of kannel gets high ( 18 sms/s), at that point it seems that apache simply cannot handle the load, and stats dropping conections on the way back ( script 2),since kannel holds the sms until it gets a 200ok message from the server, it is constantly re-filling up the socket pool ( I imagine) so basically its DoSing the 2d script.
I tried configuring apache to listen on more ports so that the uncontrolled script keeps posting to port 80 and i configured kannel to post to 2 other ports ( 8081 and 8082) to see if kannel would use up the socket pool on those ports, while the other script would have access to post back to the port 80. This did not work.
Maybe there are some config tips you could give me so that i can make this work.
some basic info on the servers is that they are not simple house machines, they actually are HP or dell servers with a big conection ( I think its T1, the server is remote and I did not set it up in the first place), 2 xeons etc.
also I want to stress that both scripts are really short and not cpu intensive, they just validate some parameters with some regexp and if all is ok curl it to somewhere else, and it would seem ( with netstat) that many conections stay open ( even the scripts have die() at each end (die sends http 200 OK) )so it makes no sense that the socket pool would be used up so quickly.
Thank you in advance
My httpd.conf is ( pertinent parts)
| Code: |
ServerRoot "/usr/lib/apache2"
<IfModule !perchild.c>
</IfModule>
#
PidFile "/var/run/apache2.pid"
Timeout 10
KeepAlive On
MaxKeepAliveRequests 20000
KeepAliveTimeout 3
<IfModule prefork.c>
StartServers 32
MinSpareServers 32
MaxSpareServers 64
MaxClients 220
MaxRequestsPerChild 100000
</IfModule>
<IfModule worker.c>
StartServers 5
MaxClients 200
MinSpareThreads 25
MaxSpareThreads 25
ThreadsPerChild 25
MaxRequestsPerChild 0
</IfModule>
<IfModule perchild.c>
NumServers 5
StartThreads 5
MinSpareThreads 5
MaxSpareThreads 10
MaxThreadsPerChild 20
MaxRequestsPerChild 0
</IfModule>
<IfModule peruser.c>
ServerLimit 256
MaxClients 256
MinSpareProcessors 2
MaxProcessors 10
MaxRequestsPerChild 1000
# kill off idle processors after this many seconds
# set to 0 to disable
ExpireTimeout 1800
Multiplexer nobody nobody
Processor apache apache
# chroot dir is optional:
# Processor user group /path/to/chroot
</IfModule>
<IfModule itk.c>
StartServers 5
MinSpareServers 2
MaxSpareServers 10
MaxClients 150
MaxRequestsPerChild 1000
</IfModule>
Listen 80
#Listen 8081
#Listen 8082
|
_________________
thesis. |
|
Networking & Security
