問題文
A team wants both a distribution of request durations and the ability to investigate one unusually slow request. Which arrangement satisfies both, and what does it cost?
選択肢
- Traces alone, since the backend can derive the distribution from them.
- A Histogram alone, because the individual measurements that went into each bucket can be recovered from the exported data points when a query narrows the range far enough.
- A Gauge for both.
- A Histogram for the distribution together with traces for the individual requests, at the cost of emitting two signals for the same operation.