Reputation: 3276
I have Jmter set up with distributed servers - single client and number of Jmeter servers.
I set mode=StrippedDiskStore
to reduce impact on load generation.
I use Stepping Thread Group to control number of threads and time. The whole solution works fine when there is relatively small load generated.
But when I increase it – in my case to over 1000 per machine I notice that when test reaches ‘stop’ time, defined in Stepping Thread Group, Jmeter starts stopping threads but it can’t do that immediately as I configured it – e.g. in 1 second.
I can see that Jmeter servers close most of the threads but still ‘wait’ for some last ones. Definitely our application under test is overloaded with requests and responses are very delayed.
But I would like to just ‘cut’ whatever is left and finish the test. Otherwise I need to wait a long time – double or more the time that is defined originally for the scenario. If I kill the servers or send shutdown message via scripts provided with Jmeter then I can’t download results from Jmter servers as StrippedDiskStore mode is used.
Is there any configuration parameter or something else that would help me here?
I also monitor machines with Jmeter servers and CPU and memory are well below 100% so I think that’s not the problem in this case.
Upvotes: 1
Views: 13688
Reputation: 34516
Your issue might be due to missing timeouts on Http Request which you can set using HTTP Request Defaults.
Now regarding the immediate stop, you can use Stop Test in JMeter GUI (only during scripting :-) ) or :
bin/stoptest.sh
bin/stoptest.cmd
This will interrupt the samplers (if Interruptable, Http Request is) but will result in those sampler being marked in error (which is regular)
As a side note, I am not sure StrippedDiskStore is the best option. Why not use StrippedAsynch ?
Upvotes: 4