Really useful! I thought I was going mad trying to understand how this was working. Thank you very much for your thorough explanation!
On Thursday, February 4, 2021 at 5:21:17 PM UTC-5 [email protected] wrote: > This is an effect of the automatic join (vector matching) that Prometheus > does around binary operators. By default, a binary operator will look for > series with exactly identical label sets on the left and right side of the > operator, and then make an identically labeled output series with that > operation applied. For series that do *not* find an exact correspondence on > the other side, they are simply dropped from the result. So if one of your > sides is a subset of the other (as in your case), only that subset will > find a label match and make it into the result. You will still have some > cost though by first selecting those extra series that then get thrown away > by the binary operator matching, so it may or may not be a good idea > (efficiency-wise) to do the filtering everywhere (if otherwise the set of > selected series would be very large). > > I visualized it with a somewhat simpler scenario here, see the "Explain" > tab on this query showing how two of the three series on the right side do > not find a match on the left side: > > https://demo.promlens.com/?l=bqxzm3klwDa > > On Thu, Feb 4, 2021 at 7:33 PM Kristopher Kahn <[email protected]> wrote: > >> I have this query: >> >> *((node_memory_MemTotal{instance="hostname.example.com:9100 >> <http://hostname.example.com:9100>"} - (node_memory_MemFree + >> node_memory_Buffers + node_memory_Cached)) / node_memory_MemTotal) * 100 * >> >> I'm surprised to find that every other metric after: >> *node_memory_MemTotal{instance="hostname.example.com:9100 >> <http://hostname.example.com:9100>"}* >> >> "knows" to use that instance, such as node_memory_MemFree, and that I >> don't need to declare that instance every time thereafter so that the query >> grows into: >> >> *((node_memory_MemTotal{instance="hostname.example.com:9100 >> <http://hostname.example.com:9100>"} - >> (node_memory_MemFree{instance="hostname.example.com:9100 >> <http://hostname.example.com:9100>"} + >> node_memory_Buffers{instance="hostname.example.com:9100 >> <http://hostname.example.com:9100>"} + >> node_memory_Cached{instance="hostname.example.com:9100 >> <http://hostname.example.com:9100>"})) / >> node_memory_MemTotal{instance="hostname.example.com:9100 >> <http://hostname.example.com:9100>"}) * 100* >> >> How does Prometheus know to keep and use that same returned instance for >> the entire query? >> >> -- >> 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/06b8d835-2c30-43dd-9467-16f6c032e28an%40googlegroups.com >> >> <https://groups.google.com/d/msgid/prometheus-users/06b8d835-2c30-43dd-9467-16f6c032e28an%40googlegroups.com?utm_medium=email&utm_source=footer> >> . >> > > > -- > Julius Volz > PromLabs - promlabs.com > -- 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/daaf6caf-c1a5-450c-9e79-f0c497705f64n%40googlegroups.com.
