Metric combining `apollo_router_processing_time` and `apollo_router_query_planning_time` for requests that require query planning

[samjcombs]

Metric combining `apollo_router_processing_time` and `apollo_router_query_planning_time` for requests that require query planning

[xuorig]

@samjcombs what would that metric actually represent? My understanding is that apollo_router_processing_time does include time spent planning, but it only includes the time actually spent waiting for the plan to be ready for that specific request, which does seem to align with "exactly how much time requests are spent in Router". That seems like the important metric to track when it comes to overhead per request.

Edit: worth noting that even the planning_time is not exactly what we'd think it is at first glance: apollographql/federation-rs#455

[samjcombs]

Ahh, thanks for confirming that

time actually spent waiting for the plan to be ready

is included in processing_time

I got misled by looking at p99s instead of max for processing_time. Requests that wait for query plans are likely well below 1% of all requests, so I was "missing" it.

Thanks for the clarification here!

[samjcombs]

Closing this feature request as obsolete because time spent waiting for query planning is included in processing_time. (Verified just now comparing max processing_time to max query planning time and they line up perfectly)
This would also catch any super-long query plans that require multiple requests to plan. Thanks, @xuorig, for the help here.