Hello,

I'm using pound in a enviroment with 9 webservers and peaks of 500 
reqs/sec. Everything is find (no load on the webservers, fast database 
and so on), but pound sometimes says without any reason "The service is 
not available. Please try again later." to the client.

I have absolutely no idea what causes that problem. Any ideas? Maybe a 
internal problem in pound? The machine running pound has load average 
under 0.3 even in peak-times.

"Alive" is set to 30, "Timeout" to 300. Should I also set the "Client" - 
var to avoid this?


Thanks in advance and best regards,
Robert

It seems I have to reply my own posting ;-)

I found the error, it was the "ulimit"-problem from my 
Debian-installation (default: 1024). I've just added "ulimit -n 10000" 
to the pound-initscript and it works fine now even with 800 reqs/sec.

Thanks to Robert S. for this great pound-thing,
Robert K.


Robert Klikics schrieb:[...]

Robert,

Thanks for the information. We would have also run into an open file limit
eventually in our testing environment. 

Can I ask how much memory and CPU time Pound is using under a load of
800 reqs/sec and "ulimit -n 10000" ? 

--
 Calomel (at) http://calomel.org
 Open Source Research and Reference


On Wed, Dec 12, 2007 at 01:56:09PM +0100, Robert Klikics wrote:[...][...][...]

Hello,

I've a Pentium 4 (at) 3,2 GHz with 1 GB RAM running as loadbalancer (with 1 
identical fallback via heartbeat) and it sleeps almost the whole day, 
even with 800 reqs/sec and 6000+ connections (poundctl says that). In 
fact, the load average is never over 0,3 (15 minutes avg). The CPU is 
mostly under 15% I as far as I know and "atsar" monitoring told me.

The only bad thing is that the logfile increases very fast. 10 Gbyte 
with LogLevel 3 are no problem ... ;-)

Greetings,
Robert


Calomel schrieb:[...][...]
>>> Hello,
>>>
>>> I'm using pound in a enviroment with 9 webservers and peaks of 500

>>> reqs/sec. Everything is find (no load on the webservers, fast
database 
>>> and so on), but pound sometimes says without any reason "The
service is 
>>> not available. Please try again later." to the client.
>>>
>>> I have absolutely no idea what causes that problem. Any ideas?
Maybe a 
>>> internal problem in pound? The machine running pound has load
average 
>>> under 0.3 even in peak-times.
>>>
>>> "Alive" is set to 30, "Timeout" to 300. Should I also set the
"Client" - 
>>> var to avoid this?
>>>
>>>
>>> Thanks in advance and best regards,
>>> Robert
>>>
>>> -- 
>>> To unsubscribe send an email with subject unsubscribe to
pound(at)apsis.ch.
>>> Please contact roseg(at)apsis.ch for questions.[...][...]