You should use rate() instead of irate(), and then if you are using grafana 7.2 or later <https://grafana.com/docs/grafana/latest/datasources/prometheus/#using-__rate_interval> :
* In your data source definition, set the scrape interval to 5 seconds * Use rate(foo[$__rate_interval]) in your queries However, it's slightly broken. What you'll actually get is a correct foo[...] range at large zoom, but at small zoom it won't go any smaller than foo[20s], which means it smears together three time intervals (i.e. it shows the average rate over 15 seconds). As a workaround, you can set the data source scrape interval to 2.5 seconds if it accepts fractions, or 3 seconds if it doesn't. More details here: https://github.com/grafana/grafana/issues/32169 This is considered a "working as intended, we won't fix" by Grafana. On Friday, 30 July 2021 at 03:45:25 UTC+1 [email protected] wrote: > My exporter scrapes metrics every 5 seconds. I set the range-vector of the > query to 10 seconds, which is 2x of scrape interval, in order to catch > 'spikes and dips', > > The problem is Grafana automatically controls the resolution(step) of the > graph, and I'm losing some data points when I set a long time range. > > Trivial solution is using '$__interval' or '$__rate_interval' variable > instead of constant range vector. > > But if I use variable for range-vector, I will lose many data points when > I monitor a long time range graph. Because irate() considers only the last > two samples. > > Is it okay to use irate() in Grafana? Or is there any better solution? > -- 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/489d6e33-c921-4cb2-a0e3-6be4a417cf91n%40googlegroups.com.
