Analysing the View Result Tree Listener

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
15 messages Options
Reply | Threaded
Open this post in threaded view
|

Analysing the View Result Tree Listener

akshay saraswat
This post has NOT been accepted by the mailing list yet.
This post was updated on .
Hi All,

I conducted a performance test on an application, however when i view the view result tree listener results, i am unable to guess where exactly the problem is,is it a bandwidth issue, a network issue,a hardware issue or issue with the application?

Attached is the screenshot of the view result tree listener, can anybody suggest as to what could be the problem
Regards,

Akshay Saraswat
ZK
Reply | Threaded
Open this post in threaded view
|

Re: Analysing the View Result Tree Listener

ZK
hi,
1st thing I notice is you are not using any assertions on your HTTP request
However; this doesn't explain your issue, but I would advise you do add a response assertion

How many threads are you running?
Do you get this issue with just 1 thread? or do you have a higher load? (I personally do not go higher than 300 threads per load injector, I also limit each thread group to 1 request per second)

If this issue doesn't happen with 1 thread but a greater load then the 1st place I would check would be your application logs


HTH
ZK
Reply | Threaded
Open this post in threaded view
|

Re: Analysing the View Result Tree Listener

akshay saraswat
This post has NOT been accepted by the mailing list yet.
hi,

thanks for the prompt reply zk, I am new to JMeter hence it was sort of a mock test.I would definitely like to know what all elements should i use in my test plan, which all listener's give a clear picture of the performance.Do i need several instances of Jmeter to handle higher loads or one would suffice?

As to your question:

I am using a higher level of load,ranging from 100 to 2000 users with a ramp up time between 1-25 sec.As can be seen, the tree listener reflects a Java Socket connection error, does this imply that the server is unable to handle such enormous requests or the application cannot handle it?

Akshay

Regards,

Akshay Saraswat
Reply | Threaded
Open this post in threaded view
|

Re: Analysing the View Result Tree Listener

prabhu
This post has NOT been accepted by the mailing list yet.
Hi,
   Can you post one of the full Error response( response data tab in View results tree)?

Prabhu
ZK
Reply | Threaded
Open this post in threaded view
|

Re: Analysing the View Result Tree Listener

ZK
This post was updated on .
In reply to this post by akshay saraswat
Hi,
if I were you, I would run the test several times, each time with a higher load
start with 100 threads, then the next test 200 and so on
After each test look at your Average response times in the Aggregate Report
(I normally use the 'Aggregate Report (not the 'Statistical Aggregate Report')+ View Results tree as my listeners... no others)

You may notice after a certain load these average response times become unacceptably high thus causing a timeout
Also does your application server have any timeout value set? (maybe 20 seconds?)
You may be able to correlate these 2 points


Another point would be if you are getting issues with say 500 threads, try distributed testing, i.e use 2 load injectors with each running 250 threads each and see if the issue still occurs

I honestly think 2000 threads for 1 load injector is way too much (personally I don't go higher than 300)


Really do what you can to ensure JMeter is not a bottleneck! otherwise your results won't be valid


HTH
ZK
Reply | Threaded
Open this post in threaded view
|

RE: Analysing the View Result Tree Listener

Robin D. Wilson
My experience with the View Results Tree listener suggest that using it has a big affect on the performance characteristics of the
test. I noticed 2 things:

1) The test generally runs slower when the View Results Tree listener is collecting sample results. (I attribute that to the fact
that it is having to write all that data to disk).

2) When you reach a certain point (several thousand samples), the test responses are substantially altered by the View Results Tree
listener (filling up some buffer somewhere?)...

I would suggest that you use the View Results Tree with some caveats:

1) Don't use it for a load-test, or performance-test scenario, it will interfere with the accuracy of your results.
2) Use it for 'Errors' recording only (in load/perf tests), since it will log only the responses that had an error (this is really
helpful in debugging what broke).
3) Use other listeners for tracking the performance and load of your system (the perfmon listener is handy, and summary results is
useful too).

--
Robin D. Wilson
Sr. Director of Web Development
KingsIsle Entertainment, Inc.
VOICE: 512-777-1861
www.KingsIsle.com


-----Original Message-----
From: ZK [mailto:[hidden email]]
Sent: Thursday, July 05, 2012 9:37 AM
To: [hidden email]
Subject: Re: Analysing the View Result Tree Listener

Hi,
if I were you, I would run the test several times, each time with a higher
load
start with 100 threads, then the next test 200 and so on
After each test look at your Average response times in the Aggregate Report

You may notice after a certain load these average response times become
unacceptably high thus causing a timeout
Also does your application have any timeout value set?

you may be able to correlate these 2 points


Another point would be if you are getting issues with say 500 threads, try
distributed testing, i.e use 2 load injectors with each running 250 threads
each and see if the issue still occurs


HTH
ZK

--
View this message in context: http://jmeter.512774.n5.nabble.com/Analysing-the-View-Result-Tree-Listener-tp5713858p5713886.html
Sent from the JMeter - User mailing list archive at Nabble.com.

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]


---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

ZK
Reply | Threaded
Open this post in threaded view
|

RE: Analysing the View Result Tree Listener

ZK
This post was updated on .
2) Use it for 'Errors' recording only (in load/perf tests), since it will log only the responses that had an error (this is really
helpful in debugging what broke).


I agree completely with this point Robin, once you are satisfied your test is executing correctly, only log errors.
I have had issues in the past as Jmeter runs out of memory if logging all results with this listener.


ZK
Reply | Threaded
Open this post in threaded view
|

RE: Analysing the View Result Tree Listener

Dzmitry_Kashlach
This post was updated on .
  How much users do you need to simulate?
  You can try our  JMeter testing cloud, it can give you up to 18000 users.

   Also we have custom JMeter Plugin JMeter Plugin,  so you can run tests locally but use Blazemeter for analyzing.
Reply | Threaded
Open this post in threaded view
|

RE: Analysing the View Result Tree Listener

akshay saraswat
This post has NOT been accepted by the mailing list yet.
In reply to this post by ZK
hi,

I need to simulate around 1000 users , I used an incremental approach and with varying ramp up's i.e I started from a mere 50 users and went upto 2000 users, i repeated these tests with ramp up range from 1-25 seconds, the results are fine till the load is around 500 users with a ramp up between 1-10 sec, however as soon as the load is increased beyond that,no matter what the ramp up is, errors start pouring in.

Also I have a few doubts:
 
1)the  response time's abruptly increase beyond 500-600 users, does this mean, peak handling capacity of the application is this number and we shouldn't load test the application beyond this ?
2) Will using multiple instances of Jmeter (load injectors) give a better picture?
3) Does using response assertions gives us an insight into the application?

Thanks,
Akshay
Regards,

Akshay Saraswat
Reply | Threaded
Open this post in threaded view
|

RE: Analysing the View Result Tree Listener

prabhu
This post has NOT been accepted by the mailing list yet.
Hi,
   You cannot just infer from this that 500-600 is the breakdown threshold of your application server. Maybe response time increase can be taken as a factor of your server's request handling ability. You also need to consider the error responses and error code you are getting. Error code like 500, internal server error or any 5XX Http errors signifies that the error is due to inablity of your server to process the requests. If you are getting connection Socket exceptions under high load, it may be due to crowding of the requests at the port you are using to send the requests to the server or the maximum amount of requests that the port can handle is exceeded. So Set yourself a goal of so many transactions to be handled ex. 500 users execute 25000 transactions in 30 min. Add the constant throughput timer accordingly. if it is going through well, increase the users and throughput. In this way you can arrive at requests handling capacity of your server.
If your application server is running in different instances, Bypass the Load balancer and hit different application instances directly by running different Jmeter instances( in different load machines with different ports). In this way you can arrive at realistic load handling capacity of your server.

Thanks
Prabhu
Reply | Threaded
Open this post in threaded view
|

RE: Analysing the View Result Tree Listener

Sriram BillaKota
This post has NOT been accepted by the mailing list yet.
This post was updated on .
Hi,


>>>1)the  response time's abruptly increase beyond 500-600 users, does this mean, peak handling capacity of >>>>the application is this number and we shouldn't load test the application beyond this ?

Bandwidth, might also be a crucial factor for your response time increase. Recently we had a load test and we see a very high response times over a period of time. After debugging further we figured out that the bandwidth is the bottleneck. So, you better calculate the data transferred for all your threads and see if you have enough bandwidth or not.

>>>>2) Will using multiple instances of Jmeter (load injectors) give a better picture?
I don't think its of much use, unless these two instances are on different locations and don't share the internet connection.

>>>>3) Does using response assertions gives us an insight into the application?
Response assertions are always a good option to fully validate your webpage requests. Though excessive usage of assertions may increase the processing time of Jmeter, you can always use them where ever necessary to see the correct page is loaded.

Thanks
pmd
Reply | Threaded
Open this post in threaded view
|

Re: Analysing the View Result Tree Listener

pmd
In reply to this post by Dzmitry_Kashlach
http://jmeter.apache.org/usermanual/best-practices.html

Tree listener is for debug , never ever use it for a load test with lot of
users or itérations .

Regards
Philippe
On Friday, July 6, 2012, Dzmitry_Kashlach wrote:

>   How much users do you need to simulate?
>   You can try our JMeter testing cloud, http://blazemeter.com, it can give
> you up to 18000 users.
>
> --
> View this message in context:
> http://jmeter.512774.n5.nabble.com/Analysing-the-View-Result-Tree-Listener-tp5713858p5713901.html
> Sent from the JMeter - User mailing list archive at Nabble.com.
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [hidden email] <javascript:;>
> For additional commands, e-mail: [hidden email]<javascript:;>
>
>

--
Cordialement.
Philippe Mouawad.
Reply | Threaded
Open this post in threaded view
|

Re: Analysing the View Result Tree Listener

Naresh Reddy
This post has NOT been accepted by the mailing list yet.
Hi All,
    Actually i am new to the Jmeter i have successfully run the script in jmeter with 10 concurrent users by taking the ramp-up period  270 seconds when i take the ramp-up  is below 250 then among all 10 users only 6 users request was successful remaining is failed.
now i  don't know how to  analyze the results based on the reports generated by jmeter so can any body help me.

Thanks in Advance.
Reply | Threaded
Open this post in threaded view
|

Re: Analysing the View Result Tree Listener

Rajendra Prasad
This post has NOT been accepted by the mailing list yet.
Hi Everyone,

I have recorded a script & add regular expression for Sign In & Sign Out . Sign In is passed but sign out is failed .










Pls let how to anlayis the error from View Result tree


Ph:9848757104
Gmail: rajendracuteboy@gmail.com
ZK
Reply | Threaded
Open this post in threaded view
|

Re: Analysing the View Result Tree Listener

ZK
Hi,
your problem is you are not signing in to your application
You should add an assertion to the 'sign in' sampler to check the user has actually signed in
for example is the 'sign out' link there after the user signs in?

Also...
some of your variables have not been set :
VIEWSTATE = ${viewState}

You should check your Regular Expressions are correct

I would advise adding a Debug Sampler to the end of your test script and checking your values there (within the View Results Tree)


ZK