jupyter-jaeger 1.0.0

Provides jupyter server proxy endpoints for launching Jaeger

jupyter-jaeger

This adds support for using the Jaeger distributed tracing tool with Jupyter. It facilitates the use case of tracking some process that starts in a kernel and is continued in a mime renderer.

We are using it to profile and debug ibis-vega-transform which goes back and forth between the kernel and the frontend to interactively render charts with Altair.

Installing this adds two Jupyter server extensions that start up the jaeger-all-in-one and jaeger-browser processes when you launch Jupyter. So to use it you must first instrument code in your kernel and/or in the frontend to record traces.

It also provis a NPM Typescript plugin you can use to access the client from inside a JupyterLab extension.

Usage

# Install jaeger-all-in-one binary
conda install -c conda-forge jaeger

# Install this package and teh REST API
pip install jupyter-jaeger jaeger_browser

# Optional, to see Jaeger icon in JupyterLab launcherr
jupyter labextension install jupyterlab-server-proxy-saulshanabrook

# Launch a Jupyter server
jupyter lab

# Open the jaeger UI
open http://localhost:8080/jaeger

Example

This repo also includes an example of starting a trace from a kernel and continueing it in a mimeredenr jupyterlab extension, in ./example. You can try this example by:

1. Launching the Binder for this repo.
2. Run the notebook.
3. Create a launcher and launch the Jaeger UI
4. Inspect the trace created by the notebook execution.