Using ultimate thread group for random load spikes

[eya...@gmail.com]

Hi,

I've been using UTG for a while to generate random load scenarios. To do that I create many spawns and pass them to JMeter from command line.

Recently I increased the number of threads being used in every spawn and I quickly run into OutOfMemmory exceptions even though at max I'm using ~1000 concurrent threads. Seems like UTG creates all necessary threads ahead and since I'm running for a long period with many spawns that can sum to thousands of threads.

Any suggestions?

[gli...@gmail.com]

If your JMeter setup lacks JVM heap memory for 1000 threads it will fail no matter which Thread Group derivative you're using, so make sure to follow JMeter Best Practices, at least the essential ones, to wit:

• Increase JVM heap allocated to JMeter to be around 80% of your total available physical RAM
• Run your test in non-GUI mode
• Disable all the Listeners
• Disable all other test elements which are not absolutely required (i.e. Assertions)

If you still experience OOM issues you will have to go for Distributed Testing as due to nature of your test the resources of your machine are not enough to conduct the required load. In any case get used to monitor whether your JMeter engine(s) have enough headroom to operate in terms of CPU, RAM, etc. as even if JMeter is not failing with OOM it might operate not efficiently enough due to i.e. swapping or extensive GC and if JMeter cannot fire requests fast enough it might not produce the required load. The health metrics can be monitored using i.e. JMeter PerfMon Plugin