[ejabberd] http_bind & high CPU load: help needed

George Hazan george.hazan at gmail.com
Mon Dec 3 13:09:05 MSK 2007


   Hello, Stefan!
   Fri, 30 Nov 2007 17:17:10 +0100 you wrote:

>> I tried to turn on epoll support (Linux Red Hat Enterprise), but the CPU
>> usage is still too high (140-150%). Postgres database is on another
>> server, and this server does almost nothing, the overall load there is
>> less than 5%.
SS> Have you tried turning off logging?

Afaik the logging can slow down the whole process, but it can't cause the 
CPU overload on the SCSI disks...

SS> I haven't done benchmarks yet with this version of ejabberd. We're using
SS> ejabberd-1.1.1 and are able to handle about 10.000 users per machine
SS> (which are less powerful than yours).

At no doubt ejabberd can handle 10K users :) I just wanna understand what am 
I doing wrong.

SS> Are you using kernel-poll? (+K true) Dunno if that's the same as epoll.

Yes, epoll is a kqueue for linux... I couldn't run the native kqueue for 
FreeBsd, even with the recent kernel patch, so I just got a machine with 
Linux where kqueue support works ok.

SS> Are you having support for SMP enabled?

Sure, otherwise I couldn't get a 140% CPU usage.

-- 
WMBR, George Hazan (ghazan at postman.ru) 




More information about the ejabberd mailing list