Hi Philippe

My responses are inline

> Hi Tomasz,
> I don't think it's related to cpum changes (it's another executable).
I also recompiled sipp and manager, not only cpum, with these changes, because 
the CPU usage on the sipp host was also wrongly measured (always 100%). And 
on the CPU host I don't run CPUM - its obviously the manager that measures 
CPU here.

Yesterday evening I reverted to original sipp and re-ran the test and it went 
normally without freezing. Earlier I ran it two times with the systemstats.* 
changes and it froze these two times.

> How much users are you using per instance ?
6000

> Are you saturating the TCP interface ? (even if it doesn't seems to be a
> high CPS)... Can you also check the cpu utilisation on the sipp/manager
> side? (ex: using top)
No. I have ran much heavier load on the same environment before. The CPU 
utilization on the sipp host was below 20% (there was not only manager and 
sipp, but also a java program)

> The message means that we have prepared calls 2 seconds upfront
> (PrepOffset), and once it's supposed to be really executed; we don't
> have calls in the SYNC (=ready) state.
> One way to debug it is to start sipp instances manually (not in
> background), so you can check the screen and get more detailed
> information about the calls states.
I was running it in foreground. Unfortunately I don't have the screen any 
more, but I don't remember anything strange there. There were some calls 
started and some of them finished. Generally looked like normally running 
sipp except for the fact that it was not updating its screen. It happened 
after several minutes from starting the test. At the beginning it was fine.

The issue is not very important, because I can use the original sipp/manager 
which measures 100% cpu on the sipp host, because the CPU usage on the SUT 
was always the critical factor in my system, not the CPU on the sipp. And the 
SUT is measured with cpum, which does not cause any problems after 
introducing your changes.

Best regards
-- 
Tomasz Radziszewski
Senior Software Engineer
Ericpol Telecom sp. z o.o.
Madalinskiego 9, 30-303 Krakow, Poland
e-mail: [EMAIL PROTECTED]
http://www.ericpol.pl/

>
> Regards,
> Philippe
>
> -----Original Message-----
> From: [EMAIL PROTECTED]
> [mailto:[EMAIL PROTECTED] On Behalf Of Tomasz
> Radziszewski
> Sent: 06 May 2008 16:49
> To: [email protected]
> Subject: Re: [Sipp-users] IMS bench SIPp: problem with cpumem
>
> After making this change, my sipp "freezes" after running about 10-15
> minutes
> of test (it seems to be random amount of time - not always the same). As
> for
> now, I haven't made enough tests to make sure that it is related to this
>
> change.
>
> I'll switch back to old version. Do you know if such effect could result
> from
> these changes? I'm attaching logs.
>
> The "frozen" sipp stops updating its screen (even the Total-time counter
>
> doesn't go forward). It happens to all sipp instances (I have 4 of
> them).
> After the sipp "freezes", the manager only prints CPU stats, not call
> stats.
> Neither sipp nor manager responds to "q" key, but they both respond to
> Ctrl-C. However, stopping manager with Ctrl-C doesn't stop sipp as it
> normally does.
>
> Characteristic entry that repeats many times in sipp logs:
>
> !! ERROR !! There should have been calls in SYNC !!.
>
> Full logs are attached.
>
> I'll appreciate any help



-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
Don't miss this year's exciting event. There's still time to save $100. 
Use priority code J8TL2D2. 
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
_______________________________________________
Sipp-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/sipp-users

Reply via email to