You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit e1b803d
Browse filesBrowse the repository at this point in the historyBrowse files
fix(fi_instrumentation): keep register()'s exporter when another span processor is added (TH-8727)
register() marked its own processor as a default (otel.py:244,
`tracer_provider._default_processor = True`). The TracerProvider
add_span_processor override shuts a default down and removes it, so the
next add_span_processor() call on the provider register() returns (a
console exporter, a second backend, LiveKit's privacy filter) dropped the
Future AGI exporter, and traces silently stopped reaching Future AGI.
Upstream OpenTelemetry's add_span_processor() only adds.
register() now removes the provider's construction default itself and
adds its processor as an ordinary one, so later processors run next to
it. A TracerProvider() constructed directly is unchanged: its first
add_span_processor() call still replaces the default SimpleSpanProcessor,
and now logs that at INFO. register(span_exporter=...) stays the way to
export somewhere else instead.
The livekit test that pinned the old register() behaviour now expects
the behaviour of the installed fi-instrumentation-otel: traceAI-livekit
allows 1.1.0, so its TH-8665 workaround in enable_*() stays.
Release: fi-instrumentation-otel 1.1.1 (version, CHANGELOG, README).
Not published.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
To export somewhere other than Future AGI, pass your exporter to `register(span_exporter=...)`. Before fi-instrumentation-otel 1.1.1, the first `add_span_processor()` call replaced the Future AGI exporter, and spans stopped reaching Future AGI.
0 commit comments