Smokeping_prober 0.7.1 - amount or buckets

[Alexander Wilke]

Hello,

I would Like to know If there are any technical disadvantages or limitations If I use a high(er) amount of buckets?

And is there a possibility to Aggregate buckets later in PromQL/Grafana?

My Idea was so unsere more buckets to have detailed View in latencies If I want to investigate Something specific in a specific time frame.

But as for Overview i do Not need all the fine grained bucket steps.

For time range i can use rate.

But is there a possibility to lets say Aggregate buckets

0.5-1.0s

1.0-2.0s

2.0-3.0s

To one bucket 0.5-3.0s

TL;DR.

Any technical limitations or disadvantages with Higher amount of buckets

How to Aggregate buckets for Overview but expand it for detailed Analysis If needed?

[Ben Kochie]

More buckets cost more to store and process, but thankfully there is now Prometheus "native histograms", which give you high resolution for less cost.

https://prometheus.io/docs/prometheus/latest/feature_flags/#native-histograms

https://www.usenix.org/conference/srecon23emea/presentation/rabenstein

This is already supported by the smokeping_prober, but you may need to enable it in Prometheus.

--
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/345a4c69-11cb-4e72-8af8-22dab5dc23e0n%40googlegroups.com.

[Alexander Wilke]

Thanks for your feedback.

I had a look at Prometheus and it looks like this feature is still in development and maybe will change.

So probably I will stay with what I have right now and enable it in the future if the feature is stable and has reached its final design.