The prover provides a substantial amount of timing and metrics information, but the current design (centered around the TimingTree and timing! macro) needs some tweaking to work well in a production cluster context.
[ ] Investigate the current solution – what is good and what's bad?
[ ] Can we tweak the timing! macro to optionally send metrics to prometheus as well?
[ ] Is Prometheus "standard enough" or do we need to be generic here (so that users can plug in some other metrics collection gizmo)
[ ] Add configuration facilities to the prover to turn on metrics collection with prometheus
The prover provides a substantial amount of timing and metrics information, but the current design (centered around the TimingTree and
timing!
macro) needs some tweaking to work well in a production cluster context.timing!
macro to optionally send metrics to prometheus as well?