Hi.

We get alot of '500 - internal server error'.
I've read the threads and ulimit was raised from 2048 to 4096 for user
who run pound. This did not have much of an impact, we went from aprox
1100 '500' messages to around 600.
We have aprox 4 milion requests a day.

We also have scripts that performs requests on all backend servers
every minute, so I know for a fact that all(in some cases 9/10)
servers are responsive.

Question is what else this can be, any clues?
[...]

Hello,

are you sure you did set the ulimit right in the pound init-script? For 
me it sounds exactly like the ulimit-problem. We had the same problem 
some time ago and an increased ulimit finally fixed it ...

4mio reqs/day should not be a problem, or is the machine running pound 
under heavy load?

Regards,
Robert


Mattias Berge schrieb:[...]

Mattias,

It is possible your backend servers need more time to respond to clients.

Try setting the TimeOut directive higher. (the default is 15 seconds) Then
test if the errors have ceased.

Pound secure reverse proxy "how to"
http://calomel.org/pound.html

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


On Sun, Jan 20, 2008 at 10:23:02PM +0100, Mattias Berge wrote:[...]

Alive 30
Client: 120
TimeOut 900
No grace.

We run apache (timeout 120) and behind that tomcat (timeout 20).

The values of pound is being configured by our hoster, and they are
not able to help me find the root of the problems.

TimeOut 900 makes no sense to me, but could this really be the origin
of our problems?

On 1/20/08, Calomel <poundlist(at)calomel.org> wrote:[...]

[...]

you might also (if possible) find out how long your webservers are 
taking to deliver the page to pound once the request was made



Mattias Berge wrote:[...][...]
>>> Hi.
>>>
>>> We get alot of '500 - internal server error'.
>>> I've read the threads and ulimit was raised from 2048 to 4096 for
user
>>> who run pound. This did not have much of an impact, we went from
aprox
>>> 1100 '500' messages to around 600.
>>> We have aprox 4 milion requests a day.
>>>
>>> We also have scripts that performs requests on all backend servers
>>> every minute, so I know for a fact that all(in some cases 9/10)
>>> servers are responsive.
>>>
>>> Question is what else this can be, any clues?
>>>
>>> --
>>> Mattias Berge
>>> Operations
>>>
>>> Travelstart Nordic AB
>>> Tel +46 (0)40-690 3825
>>>
>>> --
>>> To unsubscribe send an email with subject unsubscribe to
pound(at)apsis.ch.
>>> Please contact roseg(at)apsis.ch for questions.
>>>       [...][...]

[...]

How?

On 1/21/08, Jimmy Brake <jimmy(at)dwalliance.com> wrote:[...]
[...]

I put a command to the push the scprit name, process id and micro time 
into a log file at the top of my script and again immediately after 
pushing all the data out of the script. Then just calculate the 
difference on questionable results. There are lots of ways to do it.

Our system is pretty easy to do that technique with since nearly all our 
PHP scripts(1000+) load a config file and a template generation file. 
The 'might' and 'if possible' .... really do apply. Some of our scripts 
the ones that are known to have the occasional exceptionally long query 
or build phase have conditional logging to let us know when it is not 
working as promptly as we would like, then we research and upgrade as 
needed. Recently (the last 18 months or so) we have been using ajax+json 
foo to move more of the build phase to the client. Another technique we 
use for large PDF generation(some over 5000 pages) we just email it to 
the client.

Jimmy


Mattias Berge wrote:[...][...]
>>> Alive 30
>>> Client: 120
>>> TimeOut 900
>>> No grace.
>>>
>>> We run apache (timeout 120) and behind that tomcat (timeout 20).
>>>
>>> The values of pound is being configured by our hoster, and they
are
>>> not able to help me find the root of the problems.
>>>
>>> TimeOut 900 makes no sense to me, but could this really be the
origin
>>> of our problems?
>>>
>>> On 1/20/08, Calomel <poundlist(at)calomel.org> wrote:
>>>
>>>       
>>>> Mattias,
>>>>
>>>> It is possible your backend servers need more time to respond
to clients.
>>>>
>>>> Try setting the TimeOut directive higher. (the default is 15
seconds) Then
>>>> test if the errors have ceased.
>>>>
>>>> Pound secure reverse proxy "how to"
>>>> http://calomel.org/pound.html
>>>>
>>>> --
>>>>  Calomel (at) http://calomel.org
>>>>  Open Source Research and Reference
>>>>
>>>>
>>>> On Sun, Jan 20, 2008 at 10:23:02PM +0100, Mattias Berge wrote:
>>>>
>>>>         
>>>>> Hi.
>>>>>
>>>>> We get alot of '500 - internal server error'.
>>>>> I've read the threads and ulimit was raised from 2048 to
4096 for user
>>>>> who run pound. This did not have much of an impact, we
went from aprox
>>>>> 1100 '500' messages to around 600.
>>>>> We have aprox 4 milion requests a day.
>>>>>
>>>>> We also have scripts that performs requests on all backend
servers
>>>>> every minute, so I know for a fact that all(in some cases
9/10)
>>>>> servers are responsive.
>>>>>
>>>>> Question is what else this can be, any clues?
>>>>>
>>>>> --
>>>>> Mattias Berge
>>>>> Operations
>>>>>
>>>>> Travelstart Nordic AB
>>>>> Tel +46 (0)40-690 3825
>>>>>
>>>>> --
>>>>> To unsubscribe send an email with subject unsubscribe to
pound(at)apsis.ch.
>>>>> Please contact roseg(at)apsis.ch for questions.
>>>>>
>>>>>           
>>>> --
>>>> To unsubscribe send an email with subject unsubscribe to
pound(at)apsis.ch.
>>>> Please contact roseg(at)apsis.ch for questions.
>>>>
>>>>
>>>>
>>>>         
>>>
>>>       [...][...]

[...]