It displays the assigned stage name, the input queue fullness (both in percentage and actual amount) and the instantaneous throughput this iteration was processed at.
Things I'm on the fence about
throughput is a misleading term here, but dunno what else to use
the stage name could instead be spans which might be more appropriate?
Item processed is quite generic and pointless
Its also unclear to me which metrics would be useful here. Probably both queue fullness and throughput would be something we want to measure. However, for metrics the consumer can also compute the averages if they want.
Also note that this throughput isn't measuring over the wait time for sending and receiving as blocking here could lead to misleading values.
This PR adds tracing to the P2P sync stages.
Example trace looks something like this
It displays the assigned stage name, the input queue fullness (both in percentage and actual amount) and the instantaneous throughput this iteration was processed at.
Things I'm on the fence about
throughputis a misleading term here, but dunno what else to usespanswhich might be more appropriate?Item processedis quite generic and pointlessIts also unclear to me which metrics would be useful here. Probably both queue fullness and throughput would be something we want to measure. However, for metrics the consumer can also compute the averages if they want.
Also note that this throughput isn't measuring over the wait time for sending and receiving as blocking here could lead to misleading values.