Skip to main content
Sumo Logic

Python OpenTelemetry auto-instrumentation

OpenTelemetry Python instrumentation gives the possibility to capture telemetry data from the python written code applications. The best way to start it is to incorporate OpenTelemetry-Python. See a list of the supported libraries under OpenTelemetry Instrumentations.

How to instrument your app?

The best way to instrument Python applications is to use OpenTelemetry-Python automatic instrumentation. This approach is simple, easy, and does not require many code changes. Only a few Python packages have to be installed to successfully instrument the code of the application. Instruction below applies to OpenTelemetry-Python instrumentation in version 1.12.0rc2/0.32b0 with OpenTelemetry Protocol HTTP Exporter.

Installation of the package listed below is required to start the instrumentation and export telemetry data. Run the following pip commands:

$ pip install opentelemetry-distro==0.32b0
$ pip install opentelemetry-exporter-otlp-proto-http==1.12.0rc2

If requirements.txt is used to manage dependencies in the project add:


If pipfile, add these dependencies instead of using the pip commands. For example:

opentelemetry-distro = "0.32b0"
opentelemetry-exporter-otlp-proto-http = "1.12.0rc2"

Application specific packages installation

The next step is related to the installation of the instrumented packages used in the application. This step can be done by one simple command in the root directory of your Python application. Execution of the command below will install corresponding instrumented packages used by the application. This solution is recommended.

$ opentelemetry-bootstrap --action=install

It is possible to install instrumented packages manually but it will require from the user to find libraries used in the code and install their instrumented counterparts (e.g. Flask -> opentelemetry-instrumentation-flask).

After successful installation of the packages it is important to properly configure instrumentation. Configuration is set through environment variables.

  • OTEL_METRICS_EXPORTER=none - disables metrics exporter
  • OTEL_TRACES_EXPORTER=otlp_proto_http - sets the exporter to OTLP HTTP
  • OTEL_EXPORTER_OTLP_ENDPOINT=http://HOSTNAME:55681 - configures the endpoint where telemetry data will be sent.

    For Kubernetes environments see the available endpoints for a direct connection. For other environments see endpoints and protocols.
  • OTEL_SERVICE_NAME=SERVICE_NAME - configures the service name. Ensure the string value represents its business logic, such as "FinanceServiceCall". This will appear as a tracing service name in Sumo Logic.

  • OTEL_RESOURCE_ATTRIBUTES=application=APPLICATION_NAME - configure the application name. This will appear as a tracing application name in Sumo Logic. Additional attributes can be added here as comma separated key=value pairs.

When everything is configured it is very simple to run an instrumented application.

$ opentelemetry-instrument python3

Context propagation

By default OpenTelemetry uses W3C context propagation standard. If application instrumented by OpenTelemetry communicates with services where different context propagation is used then it is recommended to configure additional propagator to correlate events across services. This can be achieved by installation of propagators packages and OTEL_PROPAGATORS environment variable configuration.

  • B3 context propagation - common for service meshes like Istio, Linkerd and Zipkin instrumented applications.
    Propagator package installation: $ pip install opentelemetry-propagator-b3==1.12.0rc2
    Configuration: export OTEL_PROPAGATORS=b3,b3multi,tracecontext,baggage
  • AWS X-Ray - common for applications running in AWS Services
    Propagator package installation: $ pip install opentelemetry-propagator-aws-xray==1.0.1
    Configuration: export OTEL_PROPAGATORS=xray,tracecontext,baggage
  • Jaeger - common for Jaeger instrumented applications
    Propagator package installation: $ pip install opentelemetry-propagator-jaeger==1.12.0rc2
    Configuration: export OTEL_PROPAGATORS=jaeger,tracecontext,baggage
  • OpenTracing - common for OT instrumented applications
    Propagator package installation: $ pip install opentelemetry-ot-trace==0.32b0
    Configuration: export OTEL_PROPAGATORS=ottrace,tracecontext,baggage
Instrumentation notes
  • For Flask application instrumentation debug mode has to be disabled (debug=false).
  • In case of Django, application server has to be run with "--noreload" flag e.g. opentelemetry-instrument python3 runserver ---noreload