You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@jmeter.apache.org by "ohaya@yahoo.com.INVALID" <oh...@yahoo.com.INVALID> on 2019/08/21 17:15:14 UTC
Mismatch of OS Process Sampler/Beanshell Post processor vs.
requests
Hi,
I have a Jmeter test plan where I have:
Thread group - OS Process Sampler - Beanshell Post Processor - Cookie Manager
- HTTP Request - Summary
The Beanshell Post Process just haas:
response = prev.getResponseDataAsString();
vars.put("SAMLResponseBody", response);
and the HTTP Request has in the BODY DATA:
${SAMLResponseBody}
Basically, the OS Process Sampler executes a Java app, and then the Beanshell Post Processor moves the output from the Java app into a Jmeter variable, "SampleResponseBody", and then the HTTP Request sends that to a URL.
However, when I run the test plan with 100 threads with scheduler set for 30 seconds, and I look at the Summary after the test has stopped, I am seeing:
OS Process Sampler: # Samples = 106HTTP Request: # Samples = 24
And there are 0.00% Errors.
I *expected* that the #Samples for the OS Process Sampler would be the same as the #Samples for the HTTP Request, i.e., there should be one HTTP Request for each body/string that gets produced.
Can someone tell me why that is not the case?
Thanks,Jim
Re: Mismatch of OS Process Sampler/Beanshell Post processor vs. requests
Posted by Felix Schumacher <fe...@internetallee.de>.
Am 21. August 2019 23:53:22 MESZ schrieb "ohaya@yahoo.com.INVALID" <oh...@yahoo.com.INVALID>:
> I think that the HTTP POST requests are taking a really long time.
Then take a walk to the server and find where it is blocking :)
Felix
>
>On Wednesday, August 21, 2019, 8:23:07 PM UTC, ohaya@yahoo.com
><oh...@yahoo.com> wrote:
>
> Hi,
>The logging is set for DEBUG already and I don't see any terminated"
>messages at all in jmeter.log.
>I tried a test with only 10 threads for 10 seconds, and endup with
>OS Process Sampler: # Samples = 20
>HTTP Request: # Samples = 10
>So even with much smaller number of threads, the # Samples is still not
>the same.
>Also the number on the upper left of the Jmeter window is 0/10.
>This is really strange :(!!
>Jim
>
>
>On Wednesday, August 21, 2019, 7:32:46 PM UTC, Ivan Rancati
><iv...@gmail.com> wrote:
>
>You could try to log with DEBUG level, and see if jmeter.log has
>messages
>like
>"thread 1-37 terminated because the scheduler's time is reached".
>
>Perhaps it's just the web server that takes a long time to answer when
>100
>requests arrive concurrently.
>Running the test for, say, 20 minutes instead of 30 seconds might shed
>some
>light.
>Would the difference between completed OS Process Samplers and
>completed
>HTTP Samplers still be about 80 (therefore much smaller in %), or would
>you
>still have about 1 completed HTTP Sampler for approx. every 4 completed
>OS
>Process Samplers?
>
>Best regards,
>Ivan
>
>On Wed, Aug 21, 2019 at 7:15 PM ohaya@yahoo.com.INVALID
><oh...@yahoo.com.invalid> wrote:
>
>> Hi,
>> I have a Jmeter test plan where I have:
>> Thread group - OS Process Sampler - Beanshell Post Processor
>-
>> Cookie Manager
>> - HTTP Request - Summary
>>
>> The Beanshell Post Process just haas:
>> response = prev.getResponseDataAsString();
>> vars.put("SAMLResponseBody", response);
>> and the HTTP Request has in the BODY DATA:
>> ${SAMLResponseBody}
>>
>> Basically, the OS Process Sampler executes a Java app, and then the
>> Beanshell Post Processor moves the output from the Java app into a
>Jmeter
>> variable, "SampleResponseBody", and then the HTTP Request sends that
>to a
>> URL.
>> However, when I run the test plan with 100 threads with scheduler set
>for
>> 30 seconds, and I look at the Summary after the test has stopped, I
>am
>> seeing:
>> OS Process Sampler: # Samples = 106HTTP Request: # Samples = 24
>> And there are 0.00% Errors.
>>
>> I *expected* that the #Samples for the OS Process Sampler would be
>the
>> same as the #Samples for the HTTP Request, i.e., there should be one
>HTTP
>> Request for each body/string that gets produced.
>> Can someone tell me why that is not the case?
>> Thanks,Jim
>>
>>
>>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: user-unsubscribe@jmeter.apache.org
For additional commands, e-mail: user-help@jmeter.apache.org
Re: Mismatch of OS Process Sampler/Beanshell Post processor vs.
requests
Posted by "ohaya@yahoo.com.INVALID" <oh...@yahoo.com.INVALID>.
I think that the HTTP POST requests are taking a really long time.
On Wednesday, August 21, 2019, 8:23:07 PM UTC, ohaya@yahoo.com <oh...@yahoo.com> wrote:
Hi,
The logging is set for DEBUG already and I don't see any terminated" messages at all in jmeter.log.
I tried a test with only 10 threads for 10 seconds, and endup with
OS Process Sampler: # Samples = 20
HTTP Request: # Samples = 10
So even with much smaller number of threads, the # Samples is still not the same.
Also the number on the upper left of the Jmeter window is 0/10.
This is really strange :(!!
Jim
On Wednesday, August 21, 2019, 7:32:46 PM UTC, Ivan Rancati <iv...@gmail.com> wrote:
You could try to log with DEBUG level, and see if jmeter.log has messages
like
"thread 1-37 terminated because the scheduler's time is reached".
Perhaps it's just the web server that takes a long time to answer when 100
requests arrive concurrently.
Running the test for, say, 20 minutes instead of 30 seconds might shed some
light.
Would the difference between completed OS Process Samplers and completed
HTTP Samplers still be about 80 (therefore much smaller in %), or would you
still have about 1 completed HTTP Sampler for approx. every 4 completed OS
Process Samplers?
Best regards,
Ivan
On Wed, Aug 21, 2019 at 7:15 PM ohaya@yahoo.com.INVALID
<oh...@yahoo.com.invalid> wrote:
> Hi,
> I have a Jmeter test plan where I have:
> Thread group - OS Process Sampler - Beanshell Post Processor -
> Cookie Manager
> - HTTP Request - Summary
>
> The Beanshell Post Process just haas:
> response = prev.getResponseDataAsString();
> vars.put("SAMLResponseBody", response);
> and the HTTP Request has in the BODY DATA:
> ${SAMLResponseBody}
>
> Basically, the OS Process Sampler executes a Java app, and then the
> Beanshell Post Processor moves the output from the Java app into a Jmeter
> variable, "SampleResponseBody", and then the HTTP Request sends that to a
> URL.
> However, when I run the test plan with 100 threads with scheduler set for
> 30 seconds, and I look at the Summary after the test has stopped, I am
> seeing:
> OS Process Sampler: # Samples = 106HTTP Request: # Samples = 24
> And there are 0.00% Errors.
>
> I *expected* that the #Samples for the OS Process Sampler would be the
> same as the #Samples for the HTTP Request, i.e., there should be one HTTP
> Request for each body/string that gets produced.
> Can someone tell me why that is not the case?
> Thanks,Jim
>
>
>
Re: Mismatch of OS Process Sampler/Beanshell Post processor vs. requests
Posted by sebb <se...@gmail.com>.
On Sat, 24 Aug 2019 at 04:12, ohaya@yahoo.com <oh...@yahoo.com> wrote:
>
> Hi,
>
> Maybe I am misunderstanding you, but I think that that (setting the loop count, is what I am doing, e.g., I have:
>
> - Threads: 100
> - Loop: 10
>
Ok, but then don't also set the scheduler.
>
> On Saturday, August 24, 2019, 12:19:09 AM UTC, sebb <se...@gmail.com> wrote:
>
>
> On Sat, 24 Aug 2019 at 01:04, ohaya@yahoo.com.INVALID
> <oh...@yahoo.com.invalid> wrote:
> >
> > Hi,
> > Currently, we are trying to simulate a scenario where we generate a signed SAML message, and then send that message (it is encoded) as the BODY in POST request.
> > So currently in the test plan, we use the OS Process Sampler to run a Java app to create (and sign and encode) the SAML message and output that to stdout. The Beanshell Post processor captures that output into a Jmeter var, which we then use as the BODY Data in an HTTP Request.
>
> You could perhaps wrap the Java app as JavaSamplerClient.
>
> > As I asked earlier, is there any way to specify how long Jmeter will wait to consider the test plan run complete, e.g., we would like to increase to try to get more of the requests/responses.
>
> Why not limit the thread repeat count and let the test finish naturally?
>
>
> > Jim
> >
> > On Friday, August 23, 2019, 7:05:33 PM UTC, Ivan Rancati <iv...@gmail.com> wrote:
> >
> > You have an interesting setup: JMeter and Java app (is it part of the
> > System under test, btw?) on the same computer.
> >
> > If you can't run JMeter and Java app on two different computers, this might
> > work:
> > -Test 1, that runs the Java app, to generate a file with URLs and/or POST
> > contents
> > -Test 2, to run later, reads the above file and tests the web application
> >
> > Ivan
> >
> > On Thu, Aug 22, 2019 at 1:30 PM ohaya@yahoo.com.INVALID
> > <oh...@yahoo.com.invalid> wrote:
> >
> > > Hi,
> > > Is there a way to increase/set the amount of time that Jmeter will wait
> > > for more responses, after the scheduler time has elapsed?
> > > With this test, I think the java app that is run is taking so much
> > > processor time that the HTTP Requests are not being firede up.
> > >
> > > Thanks,Jim
> > >
> > >
> > >
> >
---------------------------------------------------------------------
To unsubscribe, e-mail: user-unsubscribe@jmeter.apache.org
For additional commands, e-mail: user-help@jmeter.apache.org
Re: Mismatch of OS Process Sampler/Beanshell Post processor vs.
requests
Posted by "ohaya@yahoo.com.INVALID" <oh...@yahoo.com.INVALID>.
Hi,
Maybe I am misunderstanding you, but I think that that (setting the loop count, is what I am doing, e.g., I have:
- Threads: 100- Loop: 10
On Saturday, August 24, 2019, 12:19:09 AM UTC, sebb <se...@gmail.com> wrote:
On Sat, 24 Aug 2019 at 01:04, ohaya@yahoo.com.INVALID
<oh...@yahoo.com.invalid> wrote:
>
> Hi,
> Currently, we are trying to simulate a scenario where we generate a signed SAML message, and then send that message (it is encoded) as the BODY in POST request.
> So currently in the test plan, we use the OS Process Sampler to run a Java app to create (and sign and encode) the SAML message and output that to stdout. The Beanshell Post processor captures that output into a Jmeter var, which we then use as the BODY Data in an HTTP Request.
You could perhaps wrap the Java app as JavaSamplerClient.
> As I asked earlier, is there any way to specify how long Jmeter will wait to consider the test plan run complete, e.g., we would like to increase to try to get more of the requests/responses.
Why not limit the thread repeat count and let the test finish naturally?
> Jim
>
> On Friday, August 23, 2019, 7:05:33 PM UTC, Ivan Rancati <iv...@gmail.com> wrote:
>
> You have an interesting setup: JMeter and Java app (is it part of the
> System under test, btw?) on the same computer.
>
> If you can't run JMeter and Java app on two different computers, this might
> work:
> -Test 1, that runs the Java app, to generate a file with URLs and/or POST
> contents
> -Test 2, to run later, reads the above file and tests the web application
>
> Ivan
>
> On Thu, Aug 22, 2019 at 1:30 PM ohaya@yahoo.com.INVALID
> <oh...@yahoo.com.invalid> wrote:
>
> > Hi,
> > Is there a way to increase/set the amount of time that Jmeter will wait
> > for more responses, after the scheduler time has elapsed?
> > With this test, I think the java app that is run is taking so much
> > processor time that the HTTP Requests are not being firede up.
> >
> > Thanks,Jim
> >
> >
> >
>
Re: Mismatch of OS Process Sampler/Beanshell Post processor vs. requests
Posted by sebb <se...@gmail.com>.
On Sat, 24 Aug 2019 at 01:04, ohaya@yahoo.com.INVALID
<oh...@yahoo.com.invalid> wrote:
>
> Hi,
> Currently, we are trying to simulate a scenario where we generate a signed SAML message, and then send that message (it is encoded) as the BODY in POST request.
> So currently in the test plan, we use the OS Process Sampler to run a Java app to create (and sign and encode) the SAML message and output that to stdout. The Beanshell Post processor captures that output into a Jmeter var, which we then use as the BODY Data in an HTTP Request.
You could perhaps wrap the Java app as JavaSamplerClient.
> As I asked earlier, is there any way to specify how long Jmeter will wait to consider the test plan run complete, e.g., we would like to increase to try to get more of the requests/responses.
Why not limit the thread repeat count and let the test finish naturally?
> Jim
>
> On Friday, August 23, 2019, 7:05:33 PM UTC, Ivan Rancati <iv...@gmail.com> wrote:
>
> You have an interesting setup: JMeter and Java app (is it part of the
> System under test, btw?) on the same computer.
>
> If you can't run JMeter and Java app on two different computers, this might
> work:
> -Test 1, that runs the Java app, to generate a file with URLs and/or POST
> contents
> -Test 2, to run later, reads the above file and tests the web application
>
> Ivan
>
> On Thu, Aug 22, 2019 at 1:30 PM ohaya@yahoo.com.INVALID
> <oh...@yahoo.com.invalid> wrote:
>
> > Hi,
> > Is there a way to increase/set the amount of time that Jmeter will wait
> > for more responses, after the scheduler time has elapsed?
> > With this test, I think the java app that is run is taking so much
> > processor time that the HTTP Requests are not being firede up.
> >
> > Thanks,Jim
> >
> >
> >
>
---------------------------------------------------------------------
To unsubscribe, e-mail: user-unsubscribe@jmeter.apache.org
For additional commands, e-mail: user-help@jmeter.apache.org
Re: Mismatch of OS Process Sampler/Beanshell Post processor vs.
requests
Posted by "ohaya@yahoo.com.INVALID" <oh...@yahoo.com.INVALID>.
Hi,
Currently, we are trying to simulate a scenario where we generate a signed SAML message, and then send that message (it is encoded) as the BODY in POST request.
So currently in the test plan, we use the OS Process Sampler to run a Java app to create (and sign and encode) the SAML message and output that to stdout. The Beanshell Post processor captures that output into a Jmeter var, which we then use as the BODY Data in an HTTP Request.
As I asked earlier, is there any way to specify how long Jmeter will wait to consider the test plan run complete, e.g., we would like to increase to try to get more of the requests/responses.
Jim
On Friday, August 23, 2019, 7:05:33 PM UTC, Ivan Rancati <iv...@gmail.com> wrote:
You have an interesting setup: JMeter and Java app (is it part of the
System under test, btw?) on the same computer.
If you can't run JMeter and Java app on two different computers, this might
work:
-Test 1, that runs the Java app, to generate a file with URLs and/or POST
contents
-Test 2, to run later, reads the above file and tests the web application
Ivan
On Thu, Aug 22, 2019 at 1:30 PM ohaya@yahoo.com.INVALID
<oh...@yahoo.com.invalid> wrote:
> Hi,
> Is there a way to increase/set the amount of time that Jmeter will wait
> for more responses, after the scheduler time has elapsed?
> With this test, I think the java app that is run is taking so much
> processor time that the HTTP Requests are not being firede up.
>
> Thanks,Jim
>
>
>
Re: Mismatch of OS Process Sampler/Beanshell Post processor vs. requests
Posted by Ivan Rancati <iv...@gmail.com>.
You have an interesting setup: JMeter and Java app (is it part of the
System under test, btw?) on the same computer.
If you can't run JMeter and Java app on two different computers, this might
work:
-Test 1, that runs the Java app, to generate a file with URLs and/or POST
contents
-Test 2, to run later, reads the above file and tests the web application
Ivan
On Thu, Aug 22, 2019 at 1:30 PM ohaya@yahoo.com.INVALID
<oh...@yahoo.com.invalid> wrote:
> Hi,
> Is there a way to increase/set the amount of time that Jmeter will wait
> for more responses, after the scheduler time has elapsed?
> With this test, I think the java app that is run is taking so much
> processor time that the HTTP Requests are not being firede up.
>
> Thanks,Jim
>
>
>
Re: Mismatch of OS Process Sampler/Beanshell Post processor vs.
requests
Posted by "ohaya@yahoo.com.INVALID" <oh...@yahoo.com.INVALID>.
Hi,
Is there a way to increase/set the amount of time that Jmeter will wait for more responses, after the scheduler time has elapsed?
With this test, I think the java app that is run is taking so much processor time that the HTTP Requests are not being firede up.
Thanks,Jim
On Thursday, August 22, 2019, 9:23:34 AM UTC, ohaya@yahoo.com.INVALID <oh...@yahoo.com.INVALID> wrote:
RE. >Also the number on the upper left of the Jmeter window is 0/10.
When is it showing those values?
answer: I meant to say that the 0/10 is what shows after the test plan is completed/stopped.
On Thursday, August 22, 2019, 8:48:40 AM UTC, Felix Schumacher <fe...@internetallee.de> wrote:
Am 21. August 2019 22:22:58 MESZ schrieb "ohaya@yahoo.com.INVALID" <oh...@yahoo.com.INVALID>:
> Hi,
>The logging is set for DEBUG already and I don't see any terminated"
>messages at all in jmeter.log.
>I tried a test with only 10 threads for 10 seconds, and endup with
>OS Process Sampler: # Samples = 20
>HTTP Request: # Samples = 10
>So even with much smaller number of threads, the # Samples is still not
>the same.
It could still be the case, that the server could not cope and the http samples were not processed in time and therefore didn't count.
Can you have a look at the server logs? Did it get all requests? Did it log them? How long did they take? (note that some servers will log requests quit the incoming time, so be sure to add the duration of the server requests to them)
Are the and exceptions or other strange messages in the logs of jmeter? Compare it to a run with disabled http sampler.
Take threaddumps while running the second second test and see, where jmeter is spending its time.
>Also the number on the upper left of the Jmeter window is 0/10.
When is it showing those values?
Regards
Felix
>This is really strange :(!!
>Jim
>
>
>On Wednesday, August 21, 2019, 7:32:46 PM UTC, Ivan Rancati
><iv...@gmail.com> wrote:
>
>You could try to log with DEBUG level, and see if jmeter.log has
>messages
>like
>"thread 1-37 terminated because the scheduler's time is reached".
>
>Perhaps it's just the web server that takes a long time to answer when
>100
>requests arrive concurrently.
>Running the test for, say, 20 minutes instead of 30 seconds might shed
>some
>light.
>Would the difference between completed OS Process Samplers and
>completed
>HTTP Samplers still be about 80 (therefore much smaller in %), or would
>you
>still have about 1 completed HTTP Sampler for approx. every 4 completed
>OS
>Process Samplers?
>
>Best regards,
>Ivan
>
>On Wed, Aug 21, 2019 at 7:15 PM ohaya@yahoo.com.INVALID
><oh...@yahoo.com.invalid> wrote:
>
>> Hi,
>> I have a Jmeter test plan where I have:
>> Thread group - OS Process Sampler - Beanshell Post Processor
>-
>> Cookie Manager
>> - HTTP Request - Summary
>>
>> The Beanshell Post Process just haas:
>> response = prev.getResponseDataAsString();
>> vars.put("SAMLResponseBody", response);
>> and the HTTP Request has in the BODY DATA:
>> ${SAMLResponseBody}
>>
>> Basically, the OS Process Sampler executes a Java app, and then the
>> Beanshell Post Processor moves the output from the Java app into a
>Jmeter
>> variable, "SampleResponseBody", and then the HTTP Request sends that
>to a
>> URL.
>> However, when I run the test plan with 100 threads with scheduler set
>for
>> 30 seconds, and I look at the Summary after the test has stopped, I
>am
>> seeing:
>> OS Process Sampler: # Samples = 106HTTP Request: # Samples = 24
>> And there are 0.00% Errors.
>>
>> I *expected* that the #Samples for the OS Process Sampler would be
>the
>> same as the #Samples for the HTTP Request, i.e., there should be one
>HTTP
>> Request for each body/string that gets produced.
>> Can someone tell me why that is not the case?
>> Thanks,Jim
>>
>>
>>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: user-unsubscribe@jmeter.apache.org
For additional commands, e-mail: user-help@jmeter.apache.org
Re: Mismatch of OS Process Sampler/Beanshell Post processor vs.
requests
Posted by "ohaya@yahoo.com.INVALID" <oh...@yahoo.com.INVALID>.
RE. >Also the number on the upper left of the Jmeter window is 0/10.
When is it showing those values?
answer: I meant to say that the 0/10 is what shows after the test plan is completed/stopped.
On Thursday, August 22, 2019, 8:48:40 AM UTC, Felix Schumacher <fe...@internetallee.de> wrote:
Am 21. August 2019 22:22:58 MESZ schrieb "ohaya@yahoo.com.INVALID" <oh...@yahoo.com.INVALID>:
> Hi,
>The logging is set for DEBUG already and I don't see any terminated"
>messages at all in jmeter.log.
>I tried a test with only 10 threads for 10 seconds, and endup with
>OS Process Sampler: # Samples = 20
>HTTP Request: # Samples = 10
>So even with much smaller number of threads, the # Samples is still not
>the same.
It could still be the case, that the server could not cope and the http samples were not processed in time and therefore didn't count.
Can you have a look at the server logs? Did it get all requests? Did it log them? How long did they take? (note that some servers will log requests quit the incoming time, so be sure to add the duration of the server requests to them)
Are the and exceptions or other strange messages in the logs of jmeter? Compare it to a run with disabled http sampler.
Take threaddumps while running the second second test and see, where jmeter is spending its time.
>Also the number on the upper left of the Jmeter window is 0/10.
When is it showing those values?
Regards
Felix
>This is really strange :(!!
>Jim
>
>
>On Wednesday, August 21, 2019, 7:32:46 PM UTC, Ivan Rancati
><iv...@gmail.com> wrote:
>
>You could try to log with DEBUG level, and see if jmeter.log has
>messages
>like
>"thread 1-37 terminated because the scheduler's time is reached".
>
>Perhaps it's just the web server that takes a long time to answer when
>100
>requests arrive concurrently.
>Running the test for, say, 20 minutes instead of 30 seconds might shed
>some
>light.
>Would the difference between completed OS Process Samplers and
>completed
>HTTP Samplers still be about 80 (therefore much smaller in %), or would
>you
>still have about 1 completed HTTP Sampler for approx. every 4 completed
>OS
>Process Samplers?
>
>Best regards,
>Ivan
>
>On Wed, Aug 21, 2019 at 7:15 PM ohaya@yahoo.com.INVALID
><oh...@yahoo.com.invalid> wrote:
>
>> Hi,
>> I have a Jmeter test plan where I have:
>> Thread group - OS Process Sampler - Beanshell Post Processor
>-
>> Cookie Manager
>> - HTTP Request - Summary
>>
>> The Beanshell Post Process just haas:
>> response = prev.getResponseDataAsString();
>> vars.put("SAMLResponseBody", response);
>> and the HTTP Request has in the BODY DATA:
>> ${SAMLResponseBody}
>>
>> Basically, the OS Process Sampler executes a Java app, and then the
>> Beanshell Post Processor moves the output from the Java app into a
>Jmeter
>> variable, "SampleResponseBody", and then the HTTP Request sends that
>to a
>> URL.
>> However, when I run the test plan with 100 threads with scheduler set
>for
>> 30 seconds, and I look at the Summary after the test has stopped, I
>am
>> seeing:
>> OS Process Sampler: # Samples = 106HTTP Request: # Samples = 24
>> And there are 0.00% Errors.
>>
>> I *expected* that the #Samples for the OS Process Sampler would be
>the
>> same as the #Samples for the HTTP Request, i.e., there should be one
>HTTP
>> Request for each body/string that gets produced.
>> Can someone tell me why that is not the case?
>> Thanks,Jim
>>
>>
>>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: user-unsubscribe@jmeter.apache.org
For additional commands, e-mail: user-help@jmeter.apache.org
Re: Mismatch of OS Process Sampler/Beanshell Post processor vs. requests
Posted by Felix Schumacher <fe...@internetallee.de>.
Am 21. August 2019 22:22:58 MESZ schrieb "ohaya@yahoo.com.INVALID" <oh...@yahoo.com.INVALID>:
> Hi,
>The logging is set for DEBUG already and I don't see any terminated"
>messages at all in jmeter.log.
>I tried a test with only 10 threads for 10 seconds, and endup with
>OS Process Sampler: # Samples = 20
>HTTP Request: # Samples = 10
>So even with much smaller number of threads, the # Samples is still not
>the same.
It could still be the case, that the server could not cope and the http samples were not processed in time and therefore didn't count.
Can you have a look at the server logs? Did it get all requests? Did it log them? How long did they take? (note that some servers will log requests quit the incoming time, so be sure to add the duration of the server requests to them)
Are the and exceptions or other strange messages in the logs of jmeter? Compare it to a run with disabled http sampler.
Take threaddumps while running the second second test and see, where jmeter is spending its time.
>Also the number on the upper left of the Jmeter window is 0/10.
When is it showing those values?
Regards
Felix
>This is really strange :(!!
>Jim
>
>
>On Wednesday, August 21, 2019, 7:32:46 PM UTC, Ivan Rancati
><iv...@gmail.com> wrote:
>
>You could try to log with DEBUG level, and see if jmeter.log has
>messages
>like
>"thread 1-37 terminated because the scheduler's time is reached".
>
>Perhaps it's just the web server that takes a long time to answer when
>100
>requests arrive concurrently.
>Running the test for, say, 20 minutes instead of 30 seconds might shed
>some
>light.
>Would the difference between completed OS Process Samplers and
>completed
>HTTP Samplers still be about 80 (therefore much smaller in %), or would
>you
>still have about 1 completed HTTP Sampler for approx. every 4 completed
>OS
>Process Samplers?
>
>Best regards,
>Ivan
>
>On Wed, Aug 21, 2019 at 7:15 PM ohaya@yahoo.com.INVALID
><oh...@yahoo.com.invalid> wrote:
>
>> Hi,
>> I have a Jmeter test plan where I have:
>> Thread group - OS Process Sampler - Beanshell Post Processor
>-
>> Cookie Manager
>> - HTTP Request - Summary
>>
>> The Beanshell Post Process just haas:
>> response = prev.getResponseDataAsString();
>> vars.put("SAMLResponseBody", response);
>> and the HTTP Request has in the BODY DATA:
>> ${SAMLResponseBody}
>>
>> Basically, the OS Process Sampler executes a Java app, and then the
>> Beanshell Post Processor moves the output from the Java app into a
>Jmeter
>> variable, "SampleResponseBody", and then the HTTP Request sends that
>to a
>> URL.
>> However, when I run the test plan with 100 threads with scheduler set
>for
>> 30 seconds, and I look at the Summary after the test has stopped, I
>am
>> seeing:
>> OS Process Sampler: # Samples = 106HTTP Request: # Samples = 24
>> And there are 0.00% Errors.
>>
>> I *expected* that the #Samples for the OS Process Sampler would be
>the
>> same as the #Samples for the HTTP Request, i.e., there should be one
>HTTP
>> Request for each body/string that gets produced.
>> Can someone tell me why that is not the case?
>> Thanks,Jim
>>
>>
>>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: user-unsubscribe@jmeter.apache.org
For additional commands, e-mail: user-help@jmeter.apache.org
Re: Mismatch of OS Process Sampler/Beanshell Post processor vs.
requests
Posted by "ohaya@yahoo.com.INVALID" <oh...@yahoo.com.INVALID>.
Hi,
The logging is set for DEBUG already and I don't see any terminated" messages at all in jmeter.log.
I tried a test with only 10 threads for 10 seconds, and endup with
OS Process Sampler: # Samples = 20
HTTP Request: # Samples = 10
So even with much smaller number of threads, the # Samples is still not the same.
Also the number on the upper left of the Jmeter window is 0/10.
This is really strange :(!!
Jim
On Wednesday, August 21, 2019, 7:32:46 PM UTC, Ivan Rancati <iv...@gmail.com> wrote:
You could try to log with DEBUG level, and see if jmeter.log has messages
like
"thread 1-37 terminated because the scheduler's time is reached".
Perhaps it's just the web server that takes a long time to answer when 100
requests arrive concurrently.
Running the test for, say, 20 minutes instead of 30 seconds might shed some
light.
Would the difference between completed OS Process Samplers and completed
HTTP Samplers still be about 80 (therefore much smaller in %), or would you
still have about 1 completed HTTP Sampler for approx. every 4 completed OS
Process Samplers?
Best regards,
Ivan
On Wed, Aug 21, 2019 at 7:15 PM ohaya@yahoo.com.INVALID
<oh...@yahoo.com.invalid> wrote:
> Hi,
> I have a Jmeter test plan where I have:
> Thread group - OS Process Sampler - Beanshell Post Processor -
> Cookie Manager
> - HTTP Request - Summary
>
> The Beanshell Post Process just haas:
> response = prev.getResponseDataAsString();
> vars.put("SAMLResponseBody", response);
> and the HTTP Request has in the BODY DATA:
> ${SAMLResponseBody}
>
> Basically, the OS Process Sampler executes a Java app, and then the
> Beanshell Post Processor moves the output from the Java app into a Jmeter
> variable, "SampleResponseBody", and then the HTTP Request sends that to a
> URL.
> However, when I run the test plan with 100 threads with scheduler set for
> 30 seconds, and I look at the Summary after the test has stopped, I am
> seeing:
> OS Process Sampler: # Samples = 106HTTP Request: # Samples = 24
> And there are 0.00% Errors.
>
> I *expected* that the #Samples for the OS Process Sampler would be the
> same as the #Samples for the HTTP Request, i.e., there should be one HTTP
> Request for each body/string that gets produced.
> Can someone tell me why that is not the case?
> Thanks,Jim
>
>
>
Re: Mismatch of OS Process Sampler/Beanshell Post processor vs. requests
Posted by Ivan Rancati <iv...@gmail.com>.
You could try to log with DEBUG level, and see if jmeter.log has messages
like
"thread 1-37 terminated because the scheduler's time is reached".
Perhaps it's just the web server that takes a long time to answer when 100
requests arrive concurrently.
Running the test for, say, 20 minutes instead of 30 seconds might shed some
light.
Would the difference between completed OS Process Samplers and completed
HTTP Samplers still be about 80 (therefore much smaller in %), or would you
still have about 1 completed HTTP Sampler for approx. every 4 completed OS
Process Samplers?
Best regards,
Ivan
On Wed, Aug 21, 2019 at 7:15 PM ohaya@yahoo.com.INVALID
<oh...@yahoo.com.invalid> wrote:
> Hi,
> I have a Jmeter test plan where I have:
> Thread group - OS Process Sampler - Beanshell Post Processor -
> Cookie Manager
> - HTTP Request - Summary
>
> The Beanshell Post Process just haas:
> response = prev.getResponseDataAsString();
> vars.put("SAMLResponseBody", response);
> and the HTTP Request has in the BODY DATA:
> ${SAMLResponseBody}
>
> Basically, the OS Process Sampler executes a Java app, and then the
> Beanshell Post Processor moves the output from the Java app into a Jmeter
> variable, "SampleResponseBody", and then the HTTP Request sends that to a
> URL.
> However, when I run the test plan with 100 threads with scheduler set for
> 30 seconds, and I look at the Summary after the test has stopped, I am
> seeing:
> OS Process Sampler: # Samples = 106HTTP Request: # Samples = 24
> And there are 0.00% Errors.
>
> I *expected* that the #Samples for the OS Process Sampler would be the
> same as the #Samples for the HTTP Request, i.e., there should be one HTTP
> Request for each body/string that gets produced.
> Can someone tell me why that is not the case?
> Thanks,Jim
>
>
>