You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@jmeter.apache.org by "Bartlett, Sarah" <ba...@ctc.com> on 2004/12/08 16:35:32 UTC
Possible bug in monitor results listener
I've been using the Monitor Results Listener in combination with a monitoring test plan to watch the impact of my tests on the server. It seems as though the memory % function is inverted thereby also skewing the load measurement. As my load increases the used memory goes down, which is opposite what I expected. On closer examination, it looks as though the listener is actually computing the percentage of free memory on the server. Is this a bug or was it meant to be this way but just mislabeled?
Sarah Bartlett
CTC Corporation
(814) 248-7545
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
Re: Possible bug in monitor results listener
Posted by Peter Lin <wo...@gmail.com>.
it occurred to me, it could be the graph class and how it draws the
line. if you can give me a bit more information, it would help
diagnose the problem. thanks
peter
On Tue, 18 Jan 2005 14:42:11 +0000 (UTC), Olve Hansen
<ol...@intermedia.uib.no> wrote:
> Peter Lin <woolfel <at> gmail.com> writes:
>
> >
> > yeah, that has been fixed. it was a bug in the caption and in the
> > weight calculation for the health tab. the fix was checked in a few
> > months back.
> >
> > peter
> >
>
> I have experienced the same error, but it is not fixed in 2.0.2, nor the nightly
> build from jakarta-jmeter-2_0.20050109, which is the most recent. Do I have to
> download everything from CVS, or will there be antorher nightly build out,
> incorporating this fix?
>
> I really need the Monitor Results, it is a really great piece of work, but as it
> is now, the load goes down as memory usage increases :-(
>
> Hope this can be fixed soon.
>
> Thanks,
> Olve S. Hansen
>
> ---------------------------------------------------------------------
> 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
Re: Possible bug in monitor results listener
Posted by sebb <se...@gmail.com>.
On Tue, 18 Jan 2005 14:42:11 +0000 (UTC), Olve Hansen
<ol...@intermedia.uib.no> wrote:
> Peter Lin <woolfel <at> gmail.com> writes:
>
> >
> > yeah, that has been fixed. it was a bug in the caption and in the
> > weight calculation for the health tab. the fix was checked in a few
> > months back.
> >
> > peter
> >
>
> I have experienced the same error, but it is not fixed in 2.0.2, nor the nightly
> build from jakarta-jmeter-2_0.20050109, which is the most recent. Do I have to
This is also a 2_0 branch build.
> download everything from CVS, or will there be antorher nightly build out,
> incorporating this fix?
I've just uploaded the latest builds from HEAD (and 2_0) to the
nightly directory.
>
> I really need the Monitor Results, it is a really great piece of work, but as it
> is now, the load goes down as memory usage increases :-(
>
> Hope this can be fixed soon.
>
> Thanks,
> Olve S. Hansen
>
>
> ---------------------------------------------------------------------
> 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
Re: Possible bug in monitor results listener
Posted by Peter Lin <wo...@gmail.com>.
the current implementation does this.
double memdiv = (double)usedMem/(double)totMem;
double memWeight = DEFAULT_MEMORY_FACTOR * memdiv;
the load factor is 50, so as the memory usage increases, memdiv should
get closer to 1. That would mean if the memory used is 90%.
load (blue line) would be 45 + thread weight
memory % (orange line) would be 90
what you should see at runtime, is the load will increase until the
JVM allocates more memory. Once the server reaches a stable point, the
memory usage should remain relatively constant and show a regular
pattern.
http://cvs.apache.org/viewcvs.cgi/jakarta-jmeter/src/monitor/components/org/apache/jmeter/monitor/util/Stats.java?rev=1.9&view=markup
I'll try to run a test at lunch. What kind of behavior do you expect?
I'm not sure I understand the exact problem you are seeing.
peter
On Tue, 18 Jan 2005 14:42:11 +0000 (UTC), Olve Hansen
<ol...@intermedia.uib.no> wrote:
> Peter Lin <woolfel <at> gmail.com> writes:
>
> >
> > yeah, that has been fixed. it was a bug in the caption and in the
> > weight calculation for the health tab. the fix was checked in a few
> > months back.
> >
> > peter
> >
>
> I have experienced the same error, but it is not fixed in 2.0.2, nor the nightly
> build from jakarta-jmeter-2_0.20050109, which is the most recent. Do I have to
> download everything from CVS, or will there be antorher nightly build out,
> incorporating this fix?
>
> I really need the Monitor Results, it is a really great piece of work, but as it
> is now, the load goes down as memory usage increases :-(
>
> Hope this can be fixed soon.
>
> Thanks,
> Olve S. Hansen
>
> ---------------------------------------------------------------------
> 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
Re: Possible bug in monitor results listener
Posted by Peter Lin <wo...@gmail.com>.
Hi Olve,
I ran the test locally on my laptop and the graph is correct. what I
did to test was to create a test plan with 1 thread group with 100
threads hitting "/" on tomcat. A second thread hits the status
servlet.
the ramp up time for the first thread is 100.
I also opened the status servlet with a browser, so I can periodically
look at the memory stats. what I see is that as the test ramps up,
the memory used increases, until the JVM garbage collects and
allocates more memory. After a all 100 threads are active, the server
settles down and the used memory stays below 50%. Every 2-3 minutes
the JVM performs a full GC and I see the memory % drop to about 25%.
I hope that helps.
peter
On Tue, 18 Jan 2005 14:42:11 +0000 (UTC), Olve Hansen
<ol...@intermedia.uib.no> wrote:
> Peter Lin <woolfel <at> gmail.com> writes:
>
> >
> > yeah, that has been fixed. it was a bug in the caption and in the
> > weight calculation for the health tab. the fix was checked in a few
> > months back.
> >
> > peter
> >
>
> I have experienced the same error, but it is not fixed in 2.0.2, nor the nightly
> build from jakarta-jmeter-2_0.20050109, which is the most recent. Do I have to
> download everything from CVS, or will there be antorher nightly build out,
> incorporating this fix?
>
> I really need the Monitor Results, it is a really great piece of work, but as it
> is now, the load goes down as memory usage increases :-(
>
> Hope this can be fixed soon.
>
> Thanks,
> Olve S. Hansen
>
> ---------------------------------------------------------------------
> 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
Re: Possible bug in monitor results listener
Posted by Peter Lin <wo...@gmail.com>.
let me double check, I thought I corrected it, but I could be smoking crack.
peter
On Tue, 18 Jan 2005 14:42:11 +0000 (UTC), Olve Hansen
<ol...@intermedia.uib.no> wrote:
> Peter Lin <woolfel <at> gmail.com> writes:
>
> >
> > yeah, that has been fixed. it was a bug in the caption and in the
> > weight calculation for the health tab. the fix was checked in a few
> > months back.
> >
> > peter
> >
>
> I have experienced the same error, but it is not fixed in 2.0.2, nor the nightly
> build from jakarta-jmeter-2_0.20050109, which is the most recent. Do I have to
> download everything from CVS, or will there be antorher nightly build out,
> incorporating this fix?
>
> I really need the Monitor Results, it is a really great piece of work, but as it
> is now, the load goes down as memory usage increases :-(
>
> Hope this can be fixed soon.
>
> Thanks,
> Olve S. Hansen
>
> ---------------------------------------------------------------------
> 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
Re: Possible bug in monitor results listener
Posted by Olve Hansen <ol...@intermedia.uib.no>.
Peter Lin <woolfel <at> gmail.com> writes:
>
> yeah, that has been fixed. it was a bug in the caption and in the
> weight calculation for the health tab. the fix was checked in a few
> months back.
>
> peter
>
I have experienced the same error, but it is not fixed in 2.0.2, nor the nightly
build from jakarta-jmeter-2_0.20050109, which is the most recent. Do I have to
download everything from CVS, or will there be antorher nightly build out,
incorporating this fix?
I really need the Monitor Results, it is a really great piece of work, but as it
is now, the load goes down as memory usage increases :-(
Hope this can be fixed soon.
Thanks,
Olve S. Hansen
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
Re: Possible bug in monitor results listener
Posted by Peter Lin <wo...@gmail.com>.
yeah, that has been fixed. it was a bug in the caption and in the
weight calculation for the health tab. the fix was checked in a few
months back.
peter
On Wed, 8 Dec 2004 10:35:32 -0500, Bartlett, Sarah <ba...@ctc.com> wrote:
> I've been using the Monitor Results Listener in combination with a monitoring test plan to watch the impact of my tests on the server. It seems as though the memory % function is inverted thereby also skewing the load measurement. As my load increases the used memory goes down, which is opposite what I expected. On closer examination, it looks as though the listener is actually computing the percentage of free memory on the server. Is this a bug or was it meant to be this way but just mislabeled?
>
> Sarah Bartlett
> CTC Corporation
> (814) 248-7545
>
> ---------------------------------------------------------------------
> 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