1 Agent(s) exited unexpectedly with exit code '137'

Hey @jwerner
Here, I am running my Framework on windows as MasterController and AWS Linux Instance as AgentController, on running the Load Test suddenly the Agent is exited with a “returned with exit code '137” getting the below error, please refer to the attached screenshot as well.

1 Agent(s) exited unexpectedly:
localhost
Agent ‘ac001_00’ returned with exit code ‘137’*

Please refer to the Issue and let us know what are the causes for this kind of Issue.

Thanks.

Hi vishnu3948,

Looks like the agent process on the AWS machine was killed by the OS, likely by the “out-of-memory killer”. Please check if the file “xlt/results/[timestamp]/ac001_00/agent-stderr.log” on the master controller machine contains something like “Killed”.
If this is the case, the machine probably got short of memory so the OS had to do something and killed the biggest process. Maybe the machine is simply too small in terms of RAM or the Java heap of the agent process needs to be limited to match the available RAM.

  • What is the RAM size of the AWS machine?
  • What is the current value of the JVM option “-Xmx” in the file “[your-test-suite]/config/jvmargs.cfg”?
  • How many parallel agent processes do you run on an agent machine? Look for the property “com.xceptance.xlt.mastercontroller.agentcontrollers.default.agents” in “xlt/config/mastercontroller.properties” on the master controller machine.

The master controller machine contains something like “Killed” —> Yes I can see this in “agent-stderr.log” file

RAM size of the AWS machine --> 1CPO 1 GB RAM
The current value of the JVM option “-Xmx” --> -Xmx512m
parallel agent processes --> 4

Ah, okay, then the machine is really too small. Please reconfigure (in mastercontroller.properties) the number of agents per machine to 1. Then it may work for you, but the machine is still a bit too small.

Typically, we reserve 512 MB for the system to breathe. Per agent process we assume about 512 MB for the JVM plus the configured maximum heap size. So, for 4 agents per machine with each 512 MB heap size, we would roughly need 0,5 GB + 4 * (0.5 GB + 0.5 GB) = 4.5 GB of RAM. 4 GB are probably sufficient. With this amount of RAM you could also configure 2 agents per agent machine with a heap size of 1 GB each.

We also recommend to use more than 1 CPU, but this also depends on the configured load profile.

Thanks for the Info,
As I have reconfigured the number of agents per machine to 1, but still facing the same issue, please refer to the Attachment below:
com.xceptance.xlt.mastercontroller.agentcontrollers.default.agents = 1
The current value of the JVM option “-Xmx” Set maximum permitted memory -Xmx512m

Yeah, even with only one agent process the machine is a bit too small. What about increasing the RAM of the AWS machine to 2 GB and see how it goes?

Just to make sure: I assume that your test scenarios use the HtmlUnit-based approach (and not real browsers) and XLT is the only application on the AWS machine. If this is not the case, we would need to take the resource usage of the browsers/the other applications into consideration when sizing the machine.