Hello Matthias, Yes, it was about rate query function.
Thank you very much. Le jeu. 19 nov. 2020 à 10:05, Matthias Rampke <[email protected]> a écrit : > What is "rate_1m" in this case? Is this a metric that has rate in the > name, or are you asking about the rate query function > <https://prometheus.io/docs/prometheus/latest/querying/functions/#rate>? > > Ingestion is mostly independent from query evaluation. It is in principle > possible to overload the server with queries, and thus impact ingestion, > but you'd have to try quite hard. > > For any functions that take a range vector, the time range does have an > impact, but the difference between 1m and 5m is small. There will be more > samples to iterate over, but in my experience CPU is almost never the > limiting factor. What is more important is *how much data* there is to > load. For the most part, samples within a few minutes of each other will be > stored together. If you loaded 1m worth of data you probably loaded the > other 4 minutes as well. It gets more tricky if you are trying to use 30+ > minute ranges. > > Also keep in mind that the *number* of time series plays a role – you can > get away with ranges over several hours or even days on a single time > series, but trying to query a day's worth of a few thousand time series is > going to be painful. > > The best way around this is recording rules: say, you want to graph a > service's HTTP request rate over a long period of time; you can make the > *rate* calculation immediately after ingestion, when the data is still in > memory; this is cheap even over many many time series. At the same time, > you can *aggregate* into a small number of series: when graphing a long > time, you are probably not interested in a breakdown by ephemeral > instances; sum that away and you reduce the data needs at query time by a > lot. > > All in all, I would not worry about it until you have a problem though. > > /MR > > > On Wed, Nov 18, 2020 at 9:47 AM Geoffrey Graveaud < > [email protected]> wrote: > >> >> Hello everyone. >> How are you ? >> >> Please, I have a question regarding the "*rate*" in prometheus metrics. >> >> Does it affect *performance*? maybe *thoughtput *or *ingestion *? >> I think so but I am having difficulty identifying it clearly. >> >> For example, "*rate_1m*" vs "*rate_5mn*". >> If you have some answers from experience, I'm a taker . >> >> Regards, >> Geoffrey Graveaud >> >> -- >> 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 [email protected]. >> To view this discussion on the web visit >> https://groups.google.com/d/msgid/prometheus-users/4d13fc24-fd29-45b7-8578-690718407711n%40googlegroups.com >> <https://groups.google.com/d/msgid/prometheus-users/4d13fc24-fd29-45b7-8578-690718407711n%40googlegroups.com?utm_medium=email&utm_source=footer> >> . >> > -- 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 [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/prometheus-users/CAJx0yzwV%3DLbFNwLbh0m63_M2w6AW3ePhEt3r_%2Bo3YD6BcM7W3w%40mail.gmail.com.
