Why JMmeter hangs if we set:
300 threads
600 ramp up time
one hour duration
In summary report JMeter hangs on 77000 samples count.
You are probably using GUI Mode for Load Testing, you shouldn't , see:
http://www.ubik-ingenierie.com/blog/jmeter_performance_tuning_tips/
http://jmeter.apache.org/usermanual/best-practices.html
Another possible cause is your server hanging, so add a Connection and Response Timeout in Http Request so that it fails when the server hangs instead of waiting infinitely for a response
Related
I run a performance test with load 25 users with jmeter integrated with jenkins.Once the build is triggered all my thread groups are completed but the console process is running for close to 2 hours.
image here
In your image I see 500 users, not 25.
Also I see that 498 users have finished and 2 are still running.
I cannot state for sure why they are running, it's more a question to you. I can think of the following reasons:
You're not following JMeter Best Practices
JMeter lacks resources and got stuck i.e. in endless garbage collection (which is a subset of point 1)
JMeter waits for response from the server and the server fails to respond. By default JMeter's HTTP Request samplers don't have any timeout defined and it means that JMeter will wait forever, so make sure to set reasonable connect/response timeouts, the setting lives under "Advanced" tab of the HTTP Request sampler (or better use HTTP Request Defaults so you could set the timeout for all the HTTP Request samplers in one shot)
If nothing helps you can check what exactly threads are doing and where they're stuck by taking a thread dump
I tried 500 users with 100 Ramp up/Hatch rate per seconds. The load generated was monitored for tomcat using JConsole.
For same set of request JMeter requests were using 100-80% CPU usage where as locust was around 60-30% CPU usage.
I wanted to get the bottle-neck for the server how many maximum request it can server.
Can i simulate same using locust I tried setting min-max-wait to 0 but no use.
I tried 1000 concurrent users in locust but the RPS was same. Using Jmeter i was able to find that the RPS more that this is supported by my server. RPS value by locust gives 160 where as JMeter gives 250 which is huge difference.
Locust can't generate load on my tomcat server.
Can any one please help me to explain what i am missing to simulate the load similar to JMeter.
Thanks in Advance
Looking into Locust Issue # 464 my expectation is that Locust doesn't consider so called "embedded resources", to wit images, scripts, styles, fonts, etc. which actually generate lots of extra asynchronous requests and produce the main load.
Check what exact requests are being sent by Locust and JMeter using either Tomcat access log or a sniffer tool like Wireshark - the requests should be the same in terms of amount/nature. Ideally they also should match the requests which are being sent by real user using real browser.
Follow #aldenpeterson-wf suggestion of setting min_wait and max_wait with 1000 ms to get rate per second
Have you looked at min and max wait? Setting them to both be 1000 (ms) would result in locust making 1 request/sec per client which seems to be exactly what you are looking to do.
I am trying to stress test a web application which is composed of login,view page,other pages and log out. The full flow contains 14 request and I have created 300 users to complete the flow.
I have the following Thread Group configuration:
According to the online resources since I have 300 users and the ramp up period is 6, for each 1 second there will be 50 thread added. Therefore all the 300 thread will be up and running after 6 second.
So can I conclude that after 6 second Jmeter will have 300 active thread accessing the website at the same time?
My second question is when I execute the load test of more than 100 users when I view the Result Tree Listener in the Sampler Result tab the following error is triggered only for js and css files but when I open the response data tab for that request it is displayed correctly.
Response code: 200
Response message: Embedded resource download
javax.net.ssl.SSLHandshakeException message:Non HTTP response message: Remote host closed connection during handshake,
Is it a performance issue of my website or Jmeter cannot download all the js/css files?
Thanks in advance
With regards to your Threads configuration, the actual concurrency will depend on your application response time. JMeter acts as follows:
Each 1 second JMeter will start 50 users
Each of 50 users will start executing your 14 requests upside down
When a user will finish executing all requests it will be shut down
So given your application average response time for all 14 requests is > 500 ms you should have 300 concurrent users. You can always check how many users were online using Active Threads Over Time listener. See JMeter Test Results: Why the Actual Users Number is Lower than Expected article for more detailed information on the topic
Too little information to provide the answer check jmeter.log and your application under test logs for any clues. One thing is obvious: you should definitely NOT be running JMeter in GUI mode especially with View Results Tree listener enabled as it is too resource intensive and side effects might be unpredictable. So repeat your test in non-GUI mode with all listeners disabled and if the issue will remain - inspect log files.
Using jmeter, I have 10000 users need to hit the server and to respond back with in 40 sec.
During execution (in distributed mode) only 600 users really hitting (Checked in db) in server.
But in AGGREGATE REPORT it shows all requests hits the server.
What is the issue behind this? why the number of requests hits the server isn't consistent between db and jmeter listener?
Probably your test configuration is a little bit wrong.
JMeter acts as follows:
JMeter starts all the threads (virtual users) within the bounds of ramp-up period specified in the Thread Group configuration
Each thread (virtual user) starts executing samplers upside down
When there are no more samplers to execute and no more loops to iterate the thread is being shut down.
You can check how many threads were actually active using Active Threads Over Time Listener or the same chart of the HTML Reporting Dashboard
So you may run into a situation when some samplers have already done their work and some haven't been yet started. Basically you need to provide enough loops to make sure all 10k threads will be up and running for the required test duration.
See JMeter Test Results: Why the Actual Users Number is Lower than Expected guide for more information.
I am using JMeter 2.11 for simulate 10000 thread users. I use CSV data set config to simulate 10000 user load and take 2-3 HTTP request in my test plan. All 10000 user load applied successfully by Jmeter but main problem is when my last request goes idle (as I put my Jmeter http request on home page of my website) all sessions goes timed out after some time.
I also configure "app pool's Idle time out=0" and my application's web config's "session state timeout is 20" (I also increase this time from 20 min but no way). Also Keep Alive is checked on all http request.
Please suggest me what to do for keep my last request's session alive.
JMeter threads are never idle, if thread doesn't have any more samplers to execute or no more loops to iterate it's being shut down. 10 000 users is quite a high load, it might be a problem with your JMeter configuration, i.e. it is not capable of generating and sustaining 10 000 threads. In 99% of cases jmeter.log file has enough troubleshooting information.
In the meantime few recommendations:
Upgrade to the latest version of JMeter (for the moment it's Apache JMeter 2.13)
Make sure you use 64-bit server JRE or JDK
Provide JMeter enough Java Heap space, by default it comes with 512Mb only which is not enough to simulate 10k users
Run JMeter test in non-GUI mode
Disable all Listeners during test run
See 9 Easy Solutions for a JMeter Load Test “Out of Memory” Failure for above points explained and few more tips and tricks.