(Blank image, indeed! ) I see. I have tried playing around with rate and increase some more. For a simple detection, I could make the app alert if the rate of Parcel is > 0, and I do get Grafana Alerts for that. Thank you very much.
But one more thing though, rate and increase give me the per-second increase and "units per x mins" which are decimal values < 1. Is it possible to make it so we know the exact changed value for each scraped point? For example, suppose I have data coming in like this every 5 seconds: [ 10, 20, 10, 23, 12 ,5 , 1050, 1200, 1100, 1050, 1030, 20] . Is it possible to ask Prometheus to turn this above into this below exactly? [ 0, 10, -10, 13, -11, -7, 1045, 150, -100, -50, -20, -1010 ] This is so that I could do an alert if the incoming new scraped value is increased by x amount exactly. On Tuesday, October 27, 2020 at 10:39:53 PM UTC+7 [email protected] wrote: > (You posted a blank image!) > > idelta(metric[period]) only gives the difference between the *last two > points* in a range. Suppose the time period covered has the values (0, > 0, 1, 1). The last two values are (1, 1) and idelta will give zero. > > You say you're scraping data at 5 second intervals, and you're doing > idelta(parcel[20s]). That looks at the last two points in a 20 second range > > Whether or not you see any blip on your graph depends on the graph > resolution. Say you zoom out in your graphing frontend, such that the X > axis only shows one point every 10 seconds. It's quite possible you'll > skip over the increase. It might see: > > (0, 0, 0, 0) at time 0 > (0, 0, 1, 1) at time 10 seconds > (1, 1, 1, 1) at time 20 seconds > > and hence the idelta value for all three points will be zero. You should > use delta not idelta, with a time window big enough to cover at least two > data points. > > Note: since this is a counter, you should be using "increase", "irate" or > "rate", not "(i)delta" > > - "rate" calculates the per-second increase, between the *first and last > data points* in the time window (but skipping ranges where the counter > resets, i.e. goes down). This is the normal one to use. > - "irate" gives the rate between the *last two data points* in the time > window. It suffers from the same problem you saw with idelta. > - "increase" is "rate" multiplied by the time window - instead of "units > per second" you get "units per 5 minutes" or whatever > -- 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/cc0cccca-f001-42d5-acb7-7bd743288915n%40googlegroups.com.
