Fail to load v1.13.0 due to module not found error in Python v3.12

[a03nikki]

Describe the bug

elastic-opentelemetry 1.13.0 fails to load its distro at startup with ModuleNotFoundError: No module named 'opentelemetry._opamp.agent'. The contradiction is that 1.13.0 hard-pins opentelemetry-sdk==1.41.1 as a dependency, but imports a private internal module (opentelemetry._opamp.agent) that does not exist in opentelemetry-sdk 1.41.1. The distro fails entirely, leaving the application with no OTel instrumentation.

Downgrading to elastic-opentelemetry==1.11.0 resolves the issue.

To Reproduce

1. Create a Python 3.12 virtual environment and install elastic-opentelemetry==1.13.0 (e.g. via pip install elastic-opentelemetry==1.13.0)
2. Confirm opentelemetry-sdk==1.41.1 and opentelemetry-opamp-client==0.2b0 are installed as transitive dependencies
3. Run any application with opentelemetry-instrument python -m uvicorn myapp:app ... (or any WSGI/ASGI equivalent)
4. Observe startup error:

Distribution distro configuration failed
Traceback (most recent call last):
  File ".../opentelemetry/instrumentation/auto_instrumentation/_load.py", line 68, in _load_distro
    distro = entry_point.load()()
  File ".../elasticotel/distro/__init__.py", line 55, in <module>
    from opentelemetry._opamp.agent import OpAMPAgent
ModuleNotFoundError: No module named 'opentelemetry._opamp.agent'
Failed to auto initialize OpenTelemetry

Environment (please complete the following information)

• OS: macOS (Darwin 25.5.0, Apple Silicon)
• Python version: 3.12.13
• Framework and version: FastAPI 0.115.x + uvicorn 0.30.x
• Distro version: elastic-opentelemetry 1.13.0

Additional context

The opentelemetry-opamp-client package (0.2b0) is installed — it is correctly listed as a dependency of elastic-opentelemetry 1.13.0 and resolves — but the failing import is opentelemetry._opamp.agent, which is a private internal module of opentelemetry-sdk (underscore-prefixed), not the public opentelemetry.opamp namespace provided by opentelemetry-opamp-client. These are two different things.

The package's own pinned SDK version (opentelemetry-sdk==1.41.1) does not contain opentelemetry._opamp, making this a self-contradictory release: the distro cannot load against the exact SDK version it requires.

Workaround: Pin to elastic-opentelemetry==1.11.0, which does not have this import and works correctly with the same application and Python version.

• Distro or OpenTelemetry config options

  Click to expand

  OTEL_EXPORTER_OTLP_ENDPOINT=https://<redacted>.ingest.us-central1.gcp.elastic-cloud.com:443
  OTEL_EXPORTER_OTLP_HEADERS=Authorization=ApiKey <redacted>
  OTEL_SERVICE_NAME=<redacted>
  OTEL_RESOURCE_ATTRIBUTES=deployment.environment=staging,service.version=1.1.0
  OTEL_EXPORTER_OTLP_PROTOCOL=http/protobuf
  

• requirements.txt:

  Click to expand

  elastic-opentelemetry==1.13.0   # broken; 1.11.0 is the workaround
  

  Installed transitive versions confirmed via pip show:

  • opentelemetry-sdk: 1.41.1
  • opentelemetry-api: 1.41.1
  • opentelemetry-opamp-client: 0.2b0
  • opentelemetry-semantic-conventions: 0.62b1

Description written by Claude and reviewed by Nikki.

[xrmx]

opentelemetry._opamp.agent is exported by opentelemetry-opamp-client, opentelemetry-sdk never contained any OpAMP code. Since it was previously contained by elastic-opentelemetry maybe you override it by downgrading it? try to reinstall opentelemetry-opamp-client or create a new environment.

Cannot reproduce here:

opentelemetry-instrument --metrics_exporter=console --traces_exporter=console uvicorn examples.fastapi.app:app
INFO:     Started server process [1348511]
INFO:     Waiting for application startup.
INFO:     Application startup complete.
INFO:     Uvicorn running on http://127.0.0.1:8000 (Press CTRL+C to quit)
^CINFO:     Shutting down
INFO:     Waiting for application shutdown.
INFO:     Application shutdown complete.
INFO:     Finished server process [1348511]
{
    "resource_metrics": [
        {
            "resource": {
....

With pip freeze returning:

annotated-doc==0.0.4
annotated-types==0.7.0
anyio==4.13.0
blinker==1.9.0
certifi==2026.5.20
charset-normalizer==3.4.7
click==8.4.1
elastic-opentelemetry==1.13.0
fastapi==0.136.3
Flask==3.1.3
googleapis-common-protos==1.75.0
grpcio==1.80.0
h11==0.16.0
idna==3.16
importlib_metadata==8.7.1
itsdangerous==2.2.0
Jinja2==3.1.6
MarkupSafe==3.0.3
opentelemetry-api==1.41.1
opentelemetry-exporter-otlp==1.41.1
opentelemetry-exporter-otlp-proto-common==1.41.1
opentelemetry-exporter-otlp-proto-grpc==1.41.1
opentelemetry-exporter-otlp-proto-http==1.41.1
opentelemetry-instrumentation==0.62b1
opentelemetry-instrumentation-system-metrics==0.62b1
opentelemetry-opamp-client==0.2b0
opentelemetry-proto==1.41.1
opentelemetry-resource-detector-azure==0.1.5
opentelemetry-resource-detector-containerid==0.62b1
opentelemetry-resourcedetector-gcp==1.12.0a0
opentelemetry-sdk==1.41.1
opentelemetry-sdk-extension-aws==2.1.0
opentelemetry-semantic-conventions==0.62b1
packaging==26.2
protobuf==6.33.6
psutil==7.2.2
pydantic==2.13.4
pydantic_core==2.46.4
requests==2.34.2
starlette==1.1.0
typing-inspection==0.4.2
typing_extensions==4.15.0
urllib3==2.7.0
uuid_utils==0.16.0
uvicorn==0.48.0
Werkzeug==3.1.8
wrapt==2.2.1
zipp==4.1.0

[xrmx]

Interesting, using python -m uvicorn instead of uvicorn breaks stuff here as well, but I get a completely different import error. Let me check if this is a regression from upstream.

[xrmx]

Uhm my issue seemed to be more about having the source code of this repository on my python path, without that both plain upstream 1.41.1 and elastic-opentelemetry 1.13.0 seems to run opentelemetry-instrument --metrics_exporter=console --traces_exporter=console python -m uvicorn app:app just fine.