I have pack of stats which I’m receiving over module that exports VPP stats to Prometheus – this is a vector in prometheus terms and measured in nanosecs (I think) that was spend on some operation.

Current (and incorrect) PromQL query looks like this (avg (rate(_ideco_dataplane_cpu_stat{}[30s])/1000000) by (group))*100 – it gives just flat graph even under heavy load

The only usable number I was able to get out of this stats is count of used cores – scalar(max(count(_ideco_dataplane_cpu_stat) by (node)))

What I’m trying to get is percent of CPU load for each group.