Bytewax is instrumented to offer observability of your dataflow.

The default configuration logs anything at the log level ERROR to standard output. You can control the log level by passing the log_level parameter to the setup_tracing function. If you want to see all the messages Bytewax emits, set the level to TRACE.

The TRACE level includes everything that would be sent to an opentelemetry compatible backend, like Jaeger, or the Opentelemetry Collector. It is really verbose, and your stdoutput will be flooded with logs, so use it carefully.

Try it#

Let’s try to see what jaeger can show us about a dataflow. We will make bytewax talk to the opentelemetry collector integrated in a jaeger instance. We will use the example as a reference, since it doesn’t require any other setup. You will need docker and docker-compose to run this example.

Create a folder where you’ll keep the dataflow and two more files we’ll need to run everything.

$ mkdir bytewax-tracing
$ cd bytewax-tracing

Then create a docker compose file to run jaeger with the opentelemetry collector:

# file: docker-compose.yml
version: "3"
    image: jaegertracing/all-in-one:latest
      - "16686:16686"
      - "4317:4317"
      - "4318:4318"

Now run docker compose up and everything should be up and running.

Now we need the dataflow. Download the example in this folder:

$ curl \

To instrument your dataflow, call bytewax.tracing.setup_tracing with the config object you want, and keep the returned object around (if you don’t assign to the tracer variable, tracing would not work)

# file:
from bytewax.tracing import OtlpTracingConfig, setup_tracing

tracer = setup_tracing(
# of the file

Create a virtual environment and install the needed dependencies:

$ python3 -m venv .venv
$ source .venv/bin/activate # Or on fish shell
(.venv) $ pip install bytewax sseclient-py urllib3 aiohttp_sse_client

Now you can run it with:

(.venv) $ python -m dataflow

Open your browser at and take a look at traces coming into Jaeger’s UI.

Adding custom metrics to your dataflow#

Bytewax integrates with the Prometheus Python Client to provide the ability to define custom metrics in your Python code.

Metrics that are created in your dataflow will be included in the Prometheus compatible endpoint (by default: http://localhost:3030/metrics) that is exposed when the BYTEWAX_DATAFLOW_API_ENABLED=1 environment variable is set.

For more information on creating and configuring custom metrics, see the documentation for the Prometheus Python Client.

Join our community Slack channel

Need some help? Join our community!

If you have any trouble with the process or have ideas about how to improve this document, come talk to us in the #questions-answered Slack channel!

Join now