I have a load test in which I begin with a scenario that connects many WebSockets to my microservice system (roughly 5000VUs ws) and then a scenario that sends ramping arrival rate of http requests over those WSs.
In every test, I can see in the metrics that the http messages that were sent are greater than the messages that were received by the WSs, in both I am using the native metrics (http_reqs, ws_msgs_received - statsd format).
The http message sender scenario sends about 300RPS with spikes of 600RPS across all open WSs randomly.
Is there any problem with the bandwidth of the k6/ws (this is what I use) ?
Any suggestion for making this kind of test more efficient in terms of not losing messages ?
Important to mention that I could not see any errors from my servers side and also when using another load testing provider, could not see any problem with messages losing as well.