issue with use case

[dc3o]

We're having a script that measures a user activity on the server. Script outputs the data in text file(can make it be prometheus friendly). Problem is that intervals are really short ~ 10ms. Is prometheus proper tool for scraping these targets? Goal is to graph this data and make sure I can alert on some criteria.

[Ben Kochie]

Do you really need every sample every 10ms? Or maybe you need to put your measurements into a histogram for more efficient aggregation?

What are you actually measuring and what is your actual alert criteria?

On Tue, Jun 21, 2022 at 10:18 AM dc3o <deln...@gmail.com> wrote:

We're having a script that measures a user activity on the server. Script outputs the data in text file(can make it be prometheus friendly). Problem is that intervals are really short ~ 10ms. Is prometheus proper tool for scraping these targets? Goal is to graph this data and make sure I can alert on some criteria.

--
You received this message because you are subscribed to the Google Groups "Prometheus Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to prometheus-use...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/prometheus-users/1af3ea74-625e-47f3-9ddd-9dcd9a87758dn%40googlegroups.com.

[Ben Kochie]

This is the perfect use case for a Histogram.

Rather than try and sample every N milliseconds, measure every user delay as a histogram Observe(duration) and have Prometheus collect the histogram at the usual 15 second interval.

Also, rather than try and write a script, simply build the Prometheus client library into your service such that the histogram observer is integrated. This reduces the overhead to the minimum.

On Tue, Jun 21, 2022 at 10:33 AM Nemanja Delic <deln...@gmail.com> wrote:

Main thing is to detect delays in user activity - since they are connecting remotely we are trying to catch response times in their regular activity. We are able to get this data - but don't see how this can be ingested in Prometheus.

[Brian Candler]

Also if you want to take this further, you can attach an "exemplar" to each bucket - one specific example request which generated a latency point within that range. This makes it very easy to jump to a specific log or trace with a latency within that range.

Google for "prometheus exemplar" for more info.  Note that it's currently an experimental feature which needs to be enabled.

https://prometheus.io/docs/prometheus/latest/configuration/configuration/#exemplars

https://prometheus.io/docs/prometheus/latest/feature_flags/#exemplars-storage

[Nemanja Delic]

Many thanks for your help. Will try it

To view this discussion on the web visit https://groups.google.com/d/msgid/prometheus-users/d3539815-f7a8-4e2e-9b7e-b2c25f6b22a6n%40googlegroups.com.