Hi @Elibarick
Apologies for the delay in getting back to you.
It would be best if you looked at what you are trying to load test, and try to simplify your scenarios. For example:
- If you are trying to load test getting a token, how many concurrent users will be able to get one, device a scenario for that only. And I would not mix it with the upload. Those two endpoints might have different requirements.
- For the upload endpoint, try to get the token during the setup for all VUs, and reuse it. So you can test the upload alone. And again see if that is hitting a bottleneck and where the issue is.
I imagine, based on the error you shared, that the HTTP responses might not contain a valid reply and the check fails. But I can’t say for sure since you do not share at what lines the errors happen in your actual script.
If you cant’t figure this out, seeing at what line/s it fails in the script, I would log the responses using console.log(), so you can see what is being returned when you see the failures. Or use http debugging and investigate what happens around that request.
As to what causes this when you increase the number of VUs it can be both on the system under test (SUT) side, or the load generator. Have you reviewed both your SUT logs & metrics, and the load generator, to see if there is a bottleneck that causes requests to fail, and where the bottleneck is?
Some help in Running large tests, and potentially in Unable to connect to gRPC server running locally - #7 by mstoykov.
And many thanks for helping other users in the forum, much appreciated 
Cheers!