Unusual behavior from graphs

[yurko006]

Description of issue / feature request

I am running locust.io on aws with t2.micro instances. One master and two slaves on Elastic Beanstalk. I am getting some strange behavior when it comes to the graphs.

Expected behavior

Would like to see a smooth continuous graph such as the one below. NOTE: This was running off of a Mid 2015 MacbookPro locally, non distributed.

Actual behavior

This is what it looks like when running on Elastic Beanstalk.

Environment settings (for bug reports)

• OS: 64bit Amazon Linux 2017.09 v2.6.0 running Java 8
• Locust version: 0.8.2

[heyman]

That does look weird. It's hard to debug without a way to reproduce it. Does it happen consistently? Have you tried with the Locust master branch (there been some changes to the stats & chart code). Do you use python 3 or 2?

[yurko006]

I used this repo as an example of how to deploy this framework on Elastic Beanstalk with some changes as to the framework version and instance type I am running on. It happens every time I run the code. What is also interesting is that I am seeing the number of slaves fluctuate between one and two when I run a test with the aforementioned configuration. I am running locust 0.8.1 which I grab from pip. To be honest, I am not completely sure how to check which python version it is running. Is there a version which the framework defaults to, or is there an easy way which I can check which version the framework is using?

[heyman]

Hmm, I haven't used Elastic Beanstalk. Could it be related to some autoscaling mechanism that is used?

Also, did you see the following in Readme of the repo you linked:

?

You should be able see the Python version simply by running python --version, if the Elastic-Beanstalk environment lets you do that.

[abhiksingh]

I believe this may be related to the fact that locust is CPU intensive and t2.micro instances only allow you to burst to full CPU for 6 minutes/hour. See https://aws.amazon.com/ec2/instance-types/#burst for more details.

[cgoldberg]

@abhiksingh interesting!