Another approach is to turn it around using the "unless" operator. This will *only* give timeseries from the vector on the LHS, and will suppress any which have matching label sets on the RHS.
sum by (nodename) (....) unless on (nodename) machine_cpu_cores [assuming that machine_cpu_cores has a "nodename" label] On Wednesday, 3 May 2023 at 17:18:37 UTC+1 Brian Candler wrote: > I think you need to describe: > * what you actually see > * what you would like to see instead > * the results of each of the subexpressions (i.e. left and right of "or") > in the PromQL browser > > Then it should be clearer how to combine them to achieve the result you > want. > > For more info on how the "or" operator works, see: > https://prometheus.io/docs/prometheus/latest/querying/operators/#logical-set-binary-operators > > The most important thing to note is that it's not a boolean: it's a > union. An expression like "a or b" has a vector of values for "a" and a > vector of values for "b". The result combines both timeseries a and b into > the result set. However, if there are any metrics from "a" and "b" which > match *exactly* the same set of label values, then only the "a" one will be > included in the result set. > > You can restrict the set of labels used for matching using "on(labels...)" > or "ignoring(labels...)" > > In your example: the subexpressions "sum by (nodename) (...)" will only > have a single label {nodename="XXX"}, whilst the subexpression > "machine_cpu_cores" very likely has more labels than that (job, instance > etc) and may not have a "nodename" label at all. Since the labels of the > LHS and RHS of "or" don't match, both sides are included in the result. > > On Wednesday, 3 May 2023 at 15:19:09 UTC+1 Anuj Kumar wrote: > >> HI All, >> >> I am using the below query but getting output for the two queries . I >> need to get the output for each query. Any help would be appreciated. >> >> machine_cpu_cores or sum by(nodename) >> (irate(node_cpu_seconds_total{mode="idle"}[5m]) * on(instance) >> group_left(nodename) node_uname_info) or sum by(nodename) >> (irate(node_cpu_seconds_total{mode!="idle"}[5m]) * on(instance) >> group_left(nodename) node_uname_info) >> >> Thanks, >> Anuj Kumar >> > -- 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/c0c392b6-906d-4781-aa1d-c40fe1d55f8fn%40googlegroups.com.
