Hello Grafana k6 community!
I am specifically using the xk6-output-timestream extension to stream metrics to AWS Timestream, but I think my questions here apply to other output streams like CSV and JSON as well since the output appears similar.
My main question here is around the aggregation of custom metrics in the real-time datapoints, specifically around counter custom metrics. My goal was to use counters to aggregate a rate at which certain response content was observed, and it looks like I am doing this by binning and summing the measurements per minute but it was not how I expected it to work so I am wanting to validate my assumptions.
Originally, since the documentation on details of metrics data point streams is pretty light. I figured the counter data would actually look like a cumulative counter with a sum of values added to it. The end-of-test report does this. The data point data coming from a running test is different though and it appears that the actual measurements reflect per VU counts at the end of each VU iteration. This assumption is driven by the observation of 1s in all data points for counter which only has the possibility of incrementing once per VU. Does my assumption seem correct in that the real-time data does not represent an aggregation of a counter used in most/all VUs in a test but each VU iteration’s ending counter value?
Thanks,
Sean