prometheus trigger type

[jeffgriffith]

Question about the prometheus trigger type. The examples queries show using a sum and indeed the HPA thresholding seems to indicate an (avg). The metric result gets divided by the number of pods automatically then. This seems to map to the HPA target types of Value vs AverageValue. Is there a way to tell KEDA to use a raw metric value against the threshold rather than having it do an avg of the sum based on # of pods? Thanks!

triggers:
- type: prometheus
  metadata:
    # Required
    serverAddress: http://<prometheus-host>:9090
    metricName: http_requests_total
    threshold: '100'
    query: sum(rate(http_requests_total{deployment="my-deployment"}[2m])) # Note: query must return a vector/scalar single element response

[zroubalik]

As you noticed, KEDA currently feeds those metrics to HPA as TargetValue. It is not possible to tell KEDA to not use this approach at the moment. But with a certain change in the code, it won't not a problem to add an option to use different value type. I haven't had a time to thought properly about any problems this (potential change) could brought to other scalers, but it should be ok. But this change should be made for all scalers, not just Prometheus. Are you willing to contribute this? Feel free to open an issue with a proposal.