I have moved jenkins master from win7-64 to ubuntu 12.04 and it didn't
change the situation. Once the load increases, reload of the main dashboard
produces a couple of 304s (Not Changed) with response time over 5 seconds.

However, once jenkins is deployed with tomcat problem is gone.
The response time for static content went down. On the same network segment
all 304 complete under 20ms.


The following is used as a load test:
4 machines, each running:
while /bin/true; do time wget -q -E -H -k -K -p -e robots=off
http://whatever/jenkins; done
+
browser pointing to front page with auto reload enabled.


cheers
pawel




On Sun, Jan 20, 2013 at 12:06 PM, Yitzhak Zuriel <zurie...@gmail.com> wrote:

> We also use the winstone http server that comes out of the box with
> Jenkins. In our case we run Jenkins as a Windows service.
>
> On Saturday, January 19, 2013 9:07:25 AM UTC+2, Richard J wrote:
>
>>  Yes.
>>
>> I just run java.exe Jenkins.war
>>
>> or some such command line. i.e. no container app like tomcat.
>>
>>
>>
>> *From:* jenkins...@googlegroups.**com [mailto:jenkins...@**
>> googlegroups.com] *On Behalf Of *Pawel Jasinski
>>
>> *Sent:* Friday, January 18, 2013 2:52 PM
>> *To:* jenkins...@googlegroups.**com
>>
>> *Subject:* Re: Slow Jenkins response to http after upgrade
>>
>>
>>
>> what do you use as container, winstone?
>>
>>
>>
>>
>>
>> On Thu, Jan 17, 2013 at 9:51 AM, Yitzhak Zuriel <zuri...@gmail.com>
>> wrote:
>>
>> We recently upgraded our Jenkins CI servers to the latest LTS version
>> (4.480.1 -- before this we were using version 4.447.2)  After this upgrade,
>> we found that on our busiest servers (especially on a master of 8 slaves;
>> all run on Windows Server 2008 r2) that the http response becomes very
>> slow. It's inconsistent, but approximately half of the time the GUI
>> responds either sluggishly or seems stuck. Has anyone experienced this?
>> Does anyone have suggestions for how we can investigate the cause of this?
>> We thought it might be due to intensive SCM polling activity of many
>> projects' commit jobs, so we increased the polling intervals and spread
>> them out more, and this had no effect. I'd appreciate any ideas.  A small
>> related question: one of the things we want to do for our investigation is
>> to record logs of http activity. Which logger name(s) should we use to
>> capture such logs?
>>
>>
>>
>

Reply via email to