jmeter-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From sebb <seb...@gmail.com>
Subject Re: Increasing CPU during BOSH Test using JMETER
Date Fri, 04 Jun 2010 16:11:57 GMT
On 04/06/2010, frogg6 <fdunne2@yahoo.ie> wrote:
>
>  Yes, I am looking at the max CPU comsumption. It should not go over ~75%.
>
>  The high/low bursts may be explained by the way the script works. I have 4
>  instances, each instance logs in 2500 users/threads, sends a burst of IMs,
>  then sends a presence change, burst of IM's, presence change etc...
>
>  Nothing else is running on the jmeter boxes.
>
>  It is the latest OS version with up to dates patches.
>
>  It is in an isolated network.
>
>  Is jmeter capable of running such a test over 5 days?
>

Yes, provided that you take note of:

http://jakarta.apache.org/jmeter/usermanual/best-practices.html#lean_mean

However, I'm not convinced that this is a JMeter problem.

In your original posting, you say:

"... the presence server's CPU
reaches about 90% which is a test failure. I have checked my presence server
logs and everything looks fine so the problem must lie with the toolboxes."

If I've understood this correctly, the JMeter systems are not the ones
that are suffering from increasing CPU. Therefore the most likely
problem is a resource leak or other problem in the code on the
presence server.

If the request load on the presence server is within its design
limits, it should not be possible for an external source (such as
JMeter) to cause the presence server to misbehave in this way.

If the JMeter host is having problems running JMeter, then the load at
the target server is likely to be reduced, not increased, and you
won't be able to maintain the desired throughput. But that does not
appear to be the problem here.

>  Deepak Brijnandan Goel wrote:
>  >
>  > Hey
>  >
>  > Namaskara~Nalama~Guten Tag
>  >
>  > You already have transactions/sec in the reports (messages/sec,
>  > presence/sec), which are maintained  at a constant rate.
>  >
>  > Also i checked the CPU consumed, it looks ok to me. You might need to run
>  > for longer durations to check for abnormal behavior. The avg CPU rate has
>  > increased by 10% over a days time (Are you looking at max CPU consumption
>  > in
>  > your reports?).
>  >
>  > Also the CPU consumption seems to drop in between (10000 to 14000) and
>  > then
>  > seems to increase (14000 to 17000)? This does suggest that the CPU
>  > consumption due to the test might be ok, but there are some other threads
>  > running on the test servers.
>  >
>  > Is there anything else running on the test servers? Like some maintenance
>  > threads?
>  >
>  > Some other questions:
>  >
>  > 1. Is your system up to date with all OS patches? Is the latest OS system?
>  > 2. Is it on a isolated network environment?
>  > 3. We might also need to check whether Jmeter can run test at such high
>  > rates for long duration of times (The test have been running for almost 2
>  > days at a stretch). (Most of the memory leaks show up within a few hours.
>  > One rarely needs to run test for such long times to check application
>  > abnormality)
>  >
>  > Deepak
>  >   --
>  > Keigu
>  >
>  > Deepak
>  > +91-9765089593
>  > deicool@gmail.com
>  >
>  > Skype: thumsupdeicool
>  > Google talk: deicool
>  > Blog: http://loveandfearless.wordpress.com
>  > Facebook: http://www.facebook.com/deicool
>  >
>  > Check out my Work at:
>  > LinkedIn: http://in.linkedin.com/in/thumsupdeicool
>  >
>  > "Contribute to the world, environment and more :
>  > http://www.gridrepublic.org
>  > "
>  >
>  >
>  > On Fri, Jun 4, 2010 at 7:12 PM, frogg6 <fdunne2@yahoo.ie> wrote:
>  >
>  >>
>  >>  I sent you on my results. I will have to run a trace to check if the
>  >> transactions/sec is decreasing over time. If so, what would this suggest?
>  >> How can it be fixed?
>  >>
>  >>
>  >> Deepak Brijnandan Goel wrote:
>  >> >
>  >> > Hey
>  >> >
>  >> > Namaskara~Nalama~Guten Tag
>  >> >
>  >> > Could you please share the Jmeter test reports? Is the transactions/sec
>  >> > decreasing and response time increasing over time?
>  >> >
>  >> > Deepak
>  >> >   --
>  >> > Keigu
>  >> >
>  >> > Deepak
>  >> > +91-9765089593
>  >> > deicool@gmail.com
>  >> >
>  >> > Skype: thumsupdeicool
>  >> > Google talk: deicool
>  >> > Blog: http://loveandfearless.wordpress.com
>  >> > Facebook: http://www.facebook.com/deicool
>  >> >
>  >> > Check out my Work at:
>  >> > LinkedIn: http://in.linkedin.com/in/thumsupdeicool
>  >> >
>  >> > "Contribute to the world, environment and more :
>  >> > http://www.gridrepublic.org
>  >> > "
>  >> >
>  >> >
>  >> > On Fri, Jun 4, 2010 at 6:29 PM, frogg6 <fdunne2@yahoo.ie> wrote:
>  >> >
>  >> >>
>  >> >> Hi again,
>  >> >>
>  >> >> I don't think the problem lies with the presence sever as everything
>  >> >> looks
>  >> >> ok on that side. It is something to do with the jmeter toolboxes setup
>  >> -
>  >> >> possiblr the heap size/new size? I am not storing up IM & presence
>  >> >> requests,
>  >> >> I simply fire them off to the presence server at a set rate in a
>  >> infinite
>  >> >> loop once a user has logged in. I have logging enabled to tell me
if a
>  >> >> request failed/timedout but everything looks ok. Right now I am
>  >> running
>  >> 2
>  >> >> instances of Jmeter [each of 1250 users] per toolbox. Would it be
a
>  >> >> better
>  >> >> idea to run 4 instances over 4 boxes?
>  >> >>
>  >> >>
>  >> >> Deepak Brijnandan Goel wrote:
>  >> >> >
>  >> >> > Hey
>  >> >> >
>  >> >> > Namaskara~Nalama~Guten Tag
>  >> >> >
>  >> >> > This gets a bit tricky. How does your application server the
>  >> request?
>  >> >> Does
>  >> >> > it keep the history of all the request by each user in the memory?
>  >> What
>  >> >> is
>  >> >> > the caching mechanism? Are there any objects created by each
request
>  >> >> which
>  >> >> > is not released? Can profiling of the application help in finding
>  >> out
>  >> >> > memory
>  >> >> > leaks?
>  >> >> >
>  >> >> > Deepak
>  >> >> >   --
>  >> >> > Keigu
>  >> >> >
>  >> >> > Deepak
>  >> >> > +91-9765089593
>  >> >> > deicool@gmail.com
>  >> >> >
>  >> >> > Skype: thumsupdeicool
>  >> >> > Google talk: deicool
>  >> >> > Blog: http://loveandfearless.wordpress.com
>  >> >> > Facebook: http://www.facebook.com/deicool
>  >> >> >
>  >> >> > Check out my Work at:
>  >> >> > LinkedIn: http://in.linkedin.com/in/thumsupdeicool
>  >> >> >
>  >> >> > "Contribute to the world, environment and more :
>  >> >> > http://www.gridrepublic.org
>  >> >> > "
>  >> >> >
>  >> >> >
>  >> >> > On Fri, Jun 4, 2010 at 5:20 PM, frogg6 <fdunne2@yahoo.ie>
wrote:
>  >> >> >
>  >> >> >>
>  >> >> >> Hey,
>  >> >> >>
>  >> >> >> Thanks for your quick reply! Once a user logs in, he sends
a
>  >> constant
>  >> >> >> rate
>  >> >> >> of presence and IM to the presense server in a infinite loop.
In
>  >> order
>  >> >> to
>  >> >> >> keep sending Presence and IM he will need to stay logged
in for the
>  >> >> >> duration
>  >> >> >> of the test. So in answer to your question, I don't log out
and
>  >> users
>  >> >> and
>  >> >> >> their tasks never finish [until of course the test is ended
>  >> manually].
>  >> >> >>
>  >> >> >> Is it better to use a higher heap size? Can this cause any
>  >> negaitive
>  >> >> >> effects?
>  >> >> >>
>  >> >> >>
>  >> >> >>
>  >> >> >> Deepak Brijnandan Goel wrote:
>  >> >> >> >
>  >> >> >> > Hey
>  >> >> >> >
>  >> >> >> > Namaskara~Nalama~Guten Tag
>  >> >> >> >
>  >> >> >> > You might like to check for memory leaks in the application.
Are
>  >> >> your
>  >> >> >> > users
>  >> >> >> > logging out after finishing their task?
>  >> >> >> >
>  >> >> >> > You might also like to increase your max heap size to
1 GB.
>  >> >> >> >
>  >> >> >> > Deepak
>  >> >> >> >   --
>  >> >> >> > Keigu
>  >> >> >> >
>  >> >> >> > Deepak
>  >> >> >> > +91-9765089593
>  >> >> >> > deicool@gmail.com
>  >> >> >> >
>  >> >> >> > Skype: thumsupdeicool
>  >> >> >> > Google talk: deicool
>  >> >> >> > Blog: http://loveandfearless.wordpress.com
>  >> >> >> > Facebook: http://www.facebook.com/deicool
>  >> >> >> >
>  >> >> >> > Check out my Work at:
>  >> >> >> > LinkedIn: http://in.linkedin.com/in/thumsupdeicool
>  >> >> >> >
>  >> >> >> > "Contribute to the world, environment and more :
>  >> >> >> > http://www.gridrepublic.org
>  >> >> >> > "
>  >> >> >> >
>  >> >> >> >
>  >> >> >> > On Fri, Jun 4, 2010 at 4:35 PM, frogg6 <fdunne2@yahoo.ie>
wrote:
>  >> >> >> >
>  >> >> >> >>
>  >> >> >> >> I am using Jmeter to generate BOSH login, IM &
Presence traffic
>  >> to
>  >> >> my
>  >> >> >> >> presence server. The test is for 5000 users and
the traffic from
>  >> >> >> jmeter
>  >> >> >> >> runs
>  >> >> >> >> through a proxy server before going to the presence
server. I
>  >> have
>  >> >> >> setup
>  >> >> >> >> JMeter on 2 servers running 2 identical instances
of my test
>  >> >> scripts
>  >> >> >> on
>  >> >> >> >> each
>  >> >> >> >> [2 x 2 x 1250 users/script = 5000 users]. The login
period for
>  >> 5000
>  >> >> >> uers
>  >> >> >> >> is
>  >> >> >> >> 20 mins after which time each user generates IM
and Presence
>  >> >> requests
>  >> >> >> to
>  >> >> >> >> the
>  >> >> >> >> presence server. Everything is fine for the first
24 hours, I
>  >> see
>  >> a
>  >> >> >> nice
>  >> >> >> >> steady CPU rate of around 60 % on my presence server.
However
>  >> after
>  >> >> 24
>  >> >> >> >> hours
>  >> >> >> >> the CPU gradually gets larger and larger until the
presence
>  >> >> server's
>  >> >> >> CPU
>  >> >> >> >> reaches about 90% which is a test failure. I have
checked my
>  >> >> presence
>  >> >> >> >> server
>  >> >> >> >> logs and everything looks fine so the problem must
lie with the
>  >> >> >> >> toolboxes.
>  >> >> >> >> Any idea what might be causing the CPU to rise gradually
after
>  >> 24
>  >> >> >> hours?
>  >> >> >> >> Are
>  >> >> >> >> there any memory configurations etc that need to
be set prior to
>  >> >> >> running
>  >> >> >> >> a
>  >> >> >> >> test? I am using a HeapSize of min: 256, max: 512
which for each
>  >> >> >> >> instance.
>  >> >> >> >> Do the NewSize values have to match these in terms
of ratio? My
>  >> >> >> toolboxes
>  >> >> >> >> also contain 4 processors so I am tying each jmeter
instance to
>  >> a
>  >> >> >> >> separate
>  >> >> >> >> processor which seems to fix the problem intermittently.
Any
>  >> help
>  >> >> >> would
>  >> >> >> >> be
>  >> >> >> >> much appreciated :-)
>  >> >> >> >> --
>  >> >> >> >> View this message in context:
>  >> >> >> >>
>  >> >> >>
>  >> >>
>  >> http://old.nabble.com/Increasing-CPU-during-BOSH-Test-using-JMETER-tp28778525p28778525.html
>  >> >> >> >> Sent from the JMeter - User mailing list archive
at Nabble.com.
>  >> >> >> >>
>  >> >> >> >>
>  >> >> >> >>
>  >> >> ---------------------------------------------------------------------
>  >> >> >> >> To unsubscribe, e-mail:
>  >> jmeter-user-unsubscribe@jakarta.apache.org
>  >> >> >> >> For additional commands, e-mail:
>  >> >> jmeter-user-help@jakarta.apache.org
>  >> >> >> >>
>  >> >> >> >>
>  >> >> >> >
>  >> >> >> >
>  >> >> >>
>  >> >> >> --
>  >> >> >> View this message in context:
>  >> >> >>
>  >> >>
>  >> http://old.nabble.com/Increasing-CPU-during-BOSH-Test-using-JMETER-tp28778525p28778903.html
>  >> >> >> Sent from the JMeter - User mailing list archive at Nabble.com.
>  >> >> >>
>  >> >> >>
>  >> >> >>
>  >> ---------------------------------------------------------------------
>  >> >> >> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
>  >> >> >> For additional commands, e-mail:
>  >> jmeter-user-help@jakarta.apache.org
>  >> >> >>
>  >> >> >>
>  >> >> >
>  >> >> >
>  >> >>
>  >> >> --
>  >> >> View this message in context:
>  >> >>
>  >> http://old.nabble.com/Increasing-CPU-during-BOSH-Test-using-JMETER-tp28778525p28779651.html
>  >> >> Sent from the JMeter - User mailing list archive at Nabble.com.
>  >> >>
>  >> >>
>  >> >> ---------------------------------------------------------------------
>  >> >> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
>  >> >> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
>  >> >>
>  >> >>
>  >> >
>  >> >
>  >>
>  >> --
>  >> View this message in context:
>  >> http://old.nabble.com/Increasing-CPU-during-BOSH-Test-using-JMETER-tp28778525p28780112.html
>  >> Sent from the JMeter - User mailing list archive at Nabble.com.
>  >>
>  >>
>  >> ---------------------------------------------------------------------
>  >> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
>  >> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
>  >>
>  >>
>  >
>  >
>
>
> --
>  View this message in context: http://old.nabble.com/Increasing-CPU-during-BOSH-Test-using-JMETER-tp28778525p28781903.html
>
> Sent from the JMeter - User mailing list archive at Nabble.com.
>
>
>  ---------------------------------------------------------------------
>  To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
>  For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
>
>

---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org


Mime
View raw message