Is your feature request related to a problem? Please describe.
Topic metrics are expressed as computed rates, e.g. gauges: pulsar_rate_in pulsar_rate_out. This makes Prometheus scraping miss any change of rate in between the scrape interval.
Describe the solution you'd like
Exposing the metrics as counters instead/in addition to gauges, e.g. pulsar_count_in would allow the rate to be computed in prometheus, and would not miss change of rates within the scraping interval.
Describe alternatives you've considered
Increasing the scrape interval to <15s however it still misses spikes.
I have considered using a sidecar container that would create the counters from gauges, but it adds unwanted complexity.
Is your feature request related to a problem? Please describe.
Topic metrics are expressed as computed rates, e.g. gauges:
pulsar_rate_inpulsar_rate_out. This makes Prometheus scraping miss any change of rate in between the scrape interval.Describe the solution you'd like
Exposing the metrics as counters instead/in addition to gauges, e.g.
pulsar_count_inwould allow the rate to be computed in prometheus, and would not miss change of rates within the scraping interval.Describe alternatives you've considered
Increasing the scrape interval to <15s however it still misses spikes.
I have considered using a sidecar container that would create the counters from gauges, but it adds unwanted complexity.