Observability · Flashcard
Which instrument should record request latency?
Why this is the answer
Latency is a distribution, and a Histogram records each measurement into buckets (plus sum and count) so a backend can compute p50/p99 and SLO error rates. A Gauge keeps one value and loses everything else, a Counter's sum gives only a mean that hides tail latency, and an UpDownCounter is for quantities that rise and fall as a total, like active requests.
Read more in the docsMore OpenTelemetry flashcards
- What does a Resource represent in OpenTelemetry, and how does it differ from span attributes?
- Where should the batch processor sit relative to memory_limiter and any sampling or filtering processors?
- A service’s spans never join the spans of the service it calls; each side starts its own trace. What is the usual cause?