[Question]How to understand prometheus_remote_storage_sedn_latency_seconds?

[guanglinlv]

hi, i'm a little poor on the histogram and quantile slot, what's the graph means ?

thanks a lot.

[grobie]

In Prometheus we usually use 100-quantiles also called percentiles. They represent the range of values in 100 equally sized groups.

In the example above the latency of all calls to the OpenTSDB remote storage is recorded in seconds. The 0.5 quantile means 50% of all calls took at most $value seconds, 0.9 means 90% of all calls took at most $value seconds, etc. So at the end of the time range you can see that 90% of all calls to OpenTSDB were not slower than 1s.

[guanglinlv]

oh, percentiles is fit for me. is it means that we don't need the statistic of sends in various time region [0.1,0.5,1,2,5]s? named like prometheus_remote_storage_send_latency_seconds_bucket. prometheus will do the statistic internal and calculate it to the percentiles.

thanks @grobie .

[grobie]

The difference are summaries (quantile labels) vs. histograms (_bucket metrics). Read more about the differences here: https://prometheus.io/docs/practices/histograms/

[guanglinlv]

thanks. closing it.

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.