Out of memory error.

Hi,

How much RAM does your computer have? Which version of loadUI are you running? If it's a stable version from SourceForge, please try downloading the latest nightly build to see if that works better!

Regards,
Dain
eviware.com

Hi,

Its a pretty recent version from source forge downloaded a weeks or so ago.
My laptop has 4GB of RAM.
Although Im not testing any of the services I need to yet. When I do, and if it all goes horribly wrong, I'll let you know. Thanks.

Would you expect that polling your average wcf service could cause an error like this. Given that loadui allows you to ramp up request levels to quite mind boggling rates, you wouldn't have thought that this would happen on a pc with 4GB after about 30secs. I think I rammped it up to 200/sec over 2 mins and it failed after 30secs.
The variation in the request times increased hugely too. The average starting at around 200ms and finishing when the test failed at about 1sec but the std dev ending up way higher than that at 5-6 seconds.

I have some more accurate stats on this.
My pc has 4 gig 3.24 usable.
It was using a fixed rate of just 4, with an interval of just 10ms.
Polling the yahooapis (oops I should stop doing that.)
It failed after about 2 mins with an out of memory error but looking in task manager the memory usage only went up to 1.55 gig. And when loadui was closed it dropped to 1.1gig.
I have an i5 cpu - something is surely amiss here?

Hi,

I could really do with some help with this issue now. I thought it may be due to some sort of limit imposed by the service I was hitting. This isn't the case as I'm now load testing a service within our network and whenever the load is ramped up to around 1200 I get an 'out of memory' error. I have installed windows virtual machine and have done this to see if running 2 instances of loadUI results in anything different. Interestingly both instances of LoadUI (running on the same laptop) got to around a load of 1200 before failing with the same message.
I can't believe this hasn't been noticed by others - so what am I doing wrong?

Hello!

Do you use the Web Page Runner or the soapUI Runner? If you use the soapUI Runner, what does your soapUI TestCase look like?

What number does the Running field reach approximately before failing?

/Henrik
eviware.com

Yeah I use the SOAP UI runner.

The Load is set to reach way above 1200 something like 10-20 thousand, but it never gets there because LOADUI crashes and its the 'Running' figure that gets to around 1200.

This is what the test case looks like - It's test case 3 that I'm running.

Am I understanding you correct? Have you on the soapUI Runner clicked "Settings > Advanced > concurrentSamples" and changed it to more than 10000 and done the same thing from the Workspace view with "Settings > Execution > Max internal threads"? That's the only way I can see that would make you surpass 1000 concurrent requests.

If that's indeed what you've done (and what you want to do), then you could change "-Xmx768m" in the file loadUI.jnlp (open it in notepad or similar) to "-Xmx1024m" or even higher (-Xmx1024m means use at most 1024 MB of memory). When increasing this doesn't help anymore (you're machine is no longer fast enough), you switch to Distributed mode and start distributing your TestCases to either other machines of yours, or to Amazon EC2.

For your particular soapUI TestCase though, I'd replace the soapUI Runner with a Web Page Runner (since you're just seem to be doing a HTTP GET request). A Web Page Runner will most likely require less memory.

I just successfully tested ramping up a soapUI runner (with a simple soapUI TestCase) to 9000 concurrently execution soapUI TestCases with "-Xmx1024m" in Local mode on my machine. That means 9000 x n, where n is the number of agents.

Regards!

/Henrik

Now there's a secret. The only thing I thought of was that it was a limitation of my machine (and it was - sort of - but it didn't look like it).

I will try this. Yes you are right I have moved concurrent samples up to over 10k within the project and globally.

I had thought running on agents would be my only option in order to ramp up the load - but then there are a finite number of VMs you can ask your IT dept for and I thought 1200 was a little low at any rate for a load testing tool and a machine with 4gig RAM.

Thanks for the help.

p.s. I just tried this and jumped right up to 2048 at which point I got a message saying 'unable to start java virtual machine' when I launched LoadUI.
Setting it to 1024 results in the test just stopping at around the same level. (The graph just flattens out and the counters stall in the stats component but the time keeps ticking on).

....and there was me getting hopefull.........

Hmm, ok. Would it be possible for you to send the soapUI and loadUI projects to henrik@[the-same-domain-as-this-forum]?

/Henrik
loadUI developer