PromQl: Calculating p99 over a time interval

Viewed 18

I’ve got multiple time series latency metrics and one for each kubernetes pod. I need to calculate the p99 latency metric for the entire service ,i.e, all the pods combined, over every 5min interval. How do I do that?

Here’s the sample time series data:

service_latency(app=“payments”,pod_name=“pod1”)
service_latency(app=“payments”,pod_name=“pod2”)
service_latency(app=“payments”,pod_name=“pod3”)

PromQl that I have so far:

quantile_over_time(0.99, service_latency{app=“payments”}[5m])


However, the above PromQL returns p99 metric for each pod rather than for the entire service. I also am not sure if it is indeed calculating the p99 for every 5 minutes like I intended it to.

Can a PromQL expert here please help!

0 Answers
Related