appmetrics-prometheus icon indicating copy to clipboard operation
appmetrics-prometheus copied to clipboard

Regarding percentile calculation on Summary

Open abhay40711cs opened this issue 7 years ago • 0 comments

This is moreover a query on summary matrices exposed by the package. I agree it does expose direct percentile which is great but in order to make the clear picture of application response time, I should be able to aggregate the same. I received the matrices in the below format :

http_request_duration_microseconds{handler="/xyz/xyz/xyz.js",quantile="0.5"} 5.052877 http_request_duration_microseconds{handler="/xyz/xyz/xyz.js",quantile="0.9"} 5.861079 http_request_duration_microseconds{handler="/xyz/xyz/xyz.js",quantile="0.99"} 5.861079 http_request_duration_microseconds_sum{handler="/xyz/xyz/xyz.js"} 28.973924 http_request_duration_microseconds_count{handler="/xyz/xyz/xyz.js"} 6

http_request_duration_microseconds{handler="/abc/abc/abc.js",quantile="0.5"} 5.052877 http_request_duration_microseconds{handler="/abc/abc/abc.js",quantile="0.9"} 5.861079 http_request_duration_microseconds{handler="/abc/abc/abc.js",quantile="0.99"} 5.861079 http_request_duration_microseconds_sum{handler="/abc/abc/abc.js"} 28.973924 http_request_duration_microseconds_count{handler="/abc/abc/abc.js"} 61.

How will i able to aggregate upon the various request received per operation basis , question: What is the 90th percentile response time for all those requests received last minute?

I think the need of Histogram metrices is required here.

abhay40711cs avatar Sep 20 '17 14:09 abhay40711cs