How many VUs & LGs can be used for open source CLI version?

I would like to know the following when running a tests via open source - CLI?

  1. How many VUs can be supported?
  2. How many LGs can be used?
  3. Duration of test?

I might be wrong but I believe I read a limit of 30k-40k VUs for a computer/VM/container, obviously depending on the computer/VM/container specifications.

Regarding the time, I expect k6 to be able to execute Soak tests (Endurance tests) where you test your system to prove the uptime SLA. I have been in projects where it was 24h and others where the app had to run 7 days without any downtime… so I expect that at least 24h are supported without issues. IMHO, 8 to 24h under normal system load should give enough data on the performance, if it degrades, etc… to accurately determine a system uptime.

Hope it helps :slight_smile:

1 Like

Hey Jabastin,

k6 OSS doesn’t impose any type of artificial limit in terms of number of VUs that are supported on a single instance. To answer your questions more directly, this means that you won’t be limited with number of VUs or duration. As said in the previous comment here, you will be limited only by the hardware you are running on. This article explains more in depth about running single instance of k6 with large amount of VUs while this article focuses on OS tuning and limitations.

In terms of your second question, running distributed load generators is not supported natively in k6 at this point, you can check out this issue for more info and updates on that topic. With that being said, you can orchestrate multiple k6 instances yourself using a solution like this one: GitHub - grafana/k6-operator: An operator for running distributed k6 tests.


Thanks much for the detailed informations.