Skip to content
HTTP metrics semantics: duration, population, sampling and paths
lelik112
OP

8 days ago

  1. Which exact duration field feeds HTTP p50/p90/p95/p99, and does its interval end at first byte, upstream completion or full response completion? How are aborted/streamed requests counted?

  2. Which requests are included/excluded from totalRequests and latency population: health, edge rejection, upstream failure, cancellation and zero-body responses? Are latency and 5xx denominators identical?

  3. Are observations sampled or dropped; what completeness, ingestion-delay and finalization guarantees exist for returned buckets, especially partial edge buckets?

  4. Is path the raw request path or a normalized application route template? What cardinality/redaction behavior applies to unknown or identifier-bearing paths?

$20 Bounty

1 Reply

Status changed to Awaiting Railway Response Railway • 8 days ago


Railway
BOT

8 days ago

This thread has been opened as a bounty so the community can help solve it.

Status changed to Open Railway • 8 days ago


  1. The fields are already aggregated. And IIRC, it's full response completion (from edge network to proxy to app's response time).
  2. The 5xx and latency denominators are identical.
  3. Results are bucketed at variable intervals depending on the time range being viewed.
  4. It's just a string filter that shows the path the user tried to access.

Welcome!

Sign in to your Railway account to join the conversation.

Loading...