Skip to content

Adjudication: otlp-logs/timestamp-nanosecond-precision — does time_unix_nano have to survive a round trip? #12

Description

@DeviousCardi

Check: cases/otlp-logs/timestamp-nanosecond-precision.yaml

What each implementation does

Backend Version Observed
Parseable v2.9.4 "…123456789" (string, nanoseconds) → "2026-09-08T09:31:55.123" (string, milliseconds)
OpenObserve v0.92.2 "…123456789" (string, nanoseconds) → 1788859915123456 (number, microseconds)
Quickwit 0.8.2 "…123456789" (string, nanoseconds) → 1788859957123456000 (number, nanoseconds holding whole microseconds)
VictoriaLogs v1.52.0 "…123456789" (string, nanoseconds) → "2026-09-08T14:36:09.123456789Z" (string, nanoseconds — kept intact)
Loki 3.1.1 "…123456789" (string, nanoseconds) → "1788878689123456789" (string, nanoseconds — kept intact)

Full reproductions with the exact request are in the case's own notes:.

What the specification says, and where it stops

opentelemetry-proto logs/v1, LogRecord.time_unix_nano:

time_unix_nano is the time when the event occurred, expressed as UNIX
Epoch time in nanoseconds since 00:00:00 UTC on 1 January 1970.

The specification defines what the field means and the precision it is sent
at. It does not say what a receiving store must preserve. Two of five stores
here keep the value intact, proving the precision is keepable — that is not
the same as a rule saying it must be kept.

The question put upstream

Not yet filed. Per this project's own policy on posting to repositories it
does not maintain, filing this with the OpenTelemetry specification's owners
needs explicit approval before it goes out — noted here rather than guessed
at, per ADJUDICATION.md: "An issue with no upstream question linked stays
open with nothing to promote."

The question, as it would be put: does the specification intend
time_unix_nano to be an opaque instant a conformant receiver may store at
reduced precision, or does "the time the event occurred" imply the sender's
own precision is part of what is being recorded and ought to survive a round
trip where the backend has the means to keep it (as two of the five
implementations already do)?

Their answer

Pending. This issue stays open until the question above is actually put to
the specification's owners and they answer it.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    adjudicationA check where two implementations disagree and the spec is silent

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions