Closed dcrankshaw closed 9 years ago
Good question. I think the timer should record time elapsed between request received and response sent. So if we have a situation where client sends request to A
which forwards request to B
, A
s timer should include the time to forward the request to B
and receive the response, and B
s timer should record timing as normal.
Okay, but that also means the aggregated counts for # of queries across all nodes wouldn't be accurate. Is that alright?
Yeah that's fine. For these metrics I think we're more interested in machine workload than application workload.
Fixed by #64
@dcrankshaw How should this be represented via the timer metrics?