RPS is zero when request count > 0

[mihacretu]

Hi, I have a simple test that runs with KeepConstant, copies=1, for 5 minutes. In the html report for this scenario we can see the request reach 0.2 rps.
But in the scenario details the overall RPS is 0. This is also displayed in Grafana dashboard, showing for this scenario the RPS = 0.

Unfortunately, because of the security policies of the project, i cannot share more information, but here are two screenshots to show the issue.
My question is if the RPS has double precision and can show more than 1 decimal point. From my math, it should display RPS=14(request)/300(seconds)=0.046

Thanks for taking the time to check this.

[AntyaDev]

Hi @mihacretu
Based on your stats table, out of 14 requests, only 13 succeeded — so we should calculate:
Ok RPS =13 / 300 = 0.043

I checked the code and noticed that we also apply Math.Round(RPS, 1), which is why you’re getting 0.

Additionally, most other load testing tools typically represent RPS without decimal points.

Does it make sense?

[mihacretu]

Hi @AntyaDev, thanks for your reply.
We want to compare NBomber with other performance tools, and for the other tools we use, we report the results in JTL Reporter, where we see the rps with 2 decimal points. Unfortunately is not intuitive for non technical people which reads the result why the RPS is 0 if we have succeded requests. Can this be configurable somehow?

[AntyaDev]

Hi @mihacretu
It's not currently supported, but it's not a significant feature for us to add.
I think we can add this for you. It sounds logical to have such an option.
And it would be better to round to 2 decimals by default instead of 1.

[mihacretu]

Also, if I may ask, in Grafana dashboard for rps metric, the query is taking the last* metric, which in the scenario i posted above is 0. Is there any reason why?
Please see this screenshot, the scenario from above is the last row in the table. and as you can see, there are other scenarios for which the rps is 0. Is this happening because of the rounding you mentioned?

[AntyaDev]

@mihacretu to be honest, I don't remember in detail why do we use last, probably because we display the latest updated value.
I recommend trying NBomber Studio:

• It brings a familiar UI like in NBomber HTML report.
• It will soon support running and deploying (with NATS dependency) the load test in Kubernetes and other cloud providers.
• It will soon support the "compare runs" feature, where you will be able to select runs and compare them.

[mihacretu]

Thanks for the response. This is what we plan, we just wanted to test first. Does Nbomber studio supports network isolated envs?

[AntyaDev]

Could you elaborate on the network-isolated environment?
NBomber Studio should be deployed in your private environment. It supports SSO / OIDC providers for authorization.

[AntyaDev]

@mihacretu
By a network-isolated environment, you mean to work without external access to the internet?
Then, yes, it supports this exact model.

[mihacretu]

Indeed, this is the case. Then for sure we will give it a try.

Thanks a lot for your support, I very much appreciate.

[AntyaDev]

@mihacretu
Feel free to ask any questions; I'm happy to help.
Regarding NBomber Studio, I mentioned that it will soon support deploying to K8s and other clouds, basically it will automate all these actions: https://nbomber.com/docs/deployment/deploy-to-k8s

[AntyaDev]

@mihacretu we started working on this.
Tomorrow, we will release a beta with this improvement.

[mihacretu]

@AntyaDev Thank you! I will check it soon it is available. That's awesome.