Debugging dockerized Python apps in VSCode

Despite using VSCode as my primary editor, I never really bothered to set up the native debugger to step through application code running inside Docker containers. Configuring the debugger to work with individual files, libraries, or natively running servers is trivial¹. So, I use it in those cases and just resort back to my terminal for debugging containerized apps running locally. However, after seeing a colleague’s workflow in a pair-programming session, I wanted to configure the debugger to cover this scenario too.

I’m documenting this to save my future self from banging his head against the wall trying to figure it out again.

Desiderata

I want to start a web app with docker compose up and connect the VSCode debugger to it from the UI. For this to work, along with the webserver, we’ll need to expose a debug server from the app container which the debugger can connect to.

App layout

For demonstration, I’ll go with a simple containerized starlette² app served with uvicorn³. However, the strategy will be similar for any web app. Here’s the app’s directory structure:

src
├── __init__.py
├── main.py
├── requirements.txt
├── Dockerfile
└── docker-compose.yml

In main.py, we’re exposing an endpoint as follows:

from starlette.applications import Starlette
from starlette.responses import JSONResponse
from starlette.routing import Route


async def homepage(request) -> JSONResponse:
    return JSONResponse({"hello": "world"})


app = Starlette(debug=True, routes=[Route("/", homepage)])

The requirement.txt lists out the runtime dependencies:

starlette
uvicorn

Then the Dockerfile builds the application:

FROM python:3.12-slim-bookworm
WORKDIR /usr/src/app
COPY . /usr/src/app
RUN pip install --no-cache-dir -r requirements.txt
EXPOSE 8000
CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8000"]

Finally, we orchestrate the app in a docker-compose.yml file:

# docker-compose.yml
version: "3.9"

services:
  web:
    build:
      context: .
      dockerfile: ./Dockerfile

    # This overrides the CMD in the Dockerfile
    command: uvicorn main:app --host 0.0.0.0 --port 8000
    ports:
      - 8000:8000

Add launch.json

Now, in the .vscode folder of the project’s root directory, add a file named launch.json. Create the folder if it doesn’t exist. You can also do this part manually; to do so:

• Click on the debugger button and then click on create a launch.json file.
• Select the Python debugger.
• Finally, select the Remote Attach debug config.

However, if you dislike clicking around, here’s the full content of launch.json for you to copy and paste (into $PWD/.vscode/launch.json):

{
   "version":"0.2.0",
   "configurations":[
      {
         "name":"Python: Remote Attach",
         "type":"python",
         "request":"attach",
         "connect":{
            "host":"localhost",
            "port":5678
         },
         "pathMappings":[
            {
               "localRoot":"${workspaceFolder}",
               "remoteRoot":"."
            }
         ],
         "justMyCode":true
      }
   ]
}

This instructs the VSCode debugger to attach to a debug server running on localhost through port 5678. The next section will elaborate on how to run the debug server in a container.

Launch configuration will vary depending on your project and each project needs to be set up individually. The official doc⁴ lists out all the supported application types with example configurations. To avoid having to reconfigure the same app repetitively, tracking the entire .vscode directory via source control is probably a good idea.

Add docker-compose.debug.yml

Next up, we’ll need to update the command section of services.web in the docker-compose.yml to expose the debug server. The debugpy⁵ tool from Microsoft does that for us.

However, instead of changing the docker-compose.yml file for debugging, we can add a separate file for it named docker-compose.debug.yml. Here’s the content of it:

# docker-compose.debug.yml
version: "3.9"

services:
  web:
    build:
      context: .
      dockerfile: ./Dockerfile
    command:
      - "sh"
      - "-c"
      - |
        pip install debugpy -t /tmp \
        && python /tmp/debugpy --wait-for-client --listen 0.0.0.0:5678 \
        -m uvicorn main:app --host 0.0.0.0 --port 8000        
    ports:
      - 8000:8000
      - 5678:5678

Here:

• sh -c: Selects the shell inside the Docker container.
• pip install debugpy -t /tmp: Installs the debugpy tool into the /tmp directory of the container.
• python /tmp/debugpy --wait-for-client --listen 0.0.0.0:5678: Runs debugpy, sets it to wait for a client connection and listen on all network interfaces at port 5678.
• -m uvicorn main:app --host 0.0.0.0 --port 8000: Starts an uvicorn server hosting the application defined in main:app, making it accessible on all network interfaces at port 8000.

Start the debugger

Before starting the VScode debugger, go to the project root and run:

docker compose -f docker-compose.debug.yml up

Now click on the debugger button and select the Python: Remote attach profile to start debugging. Hack away!

P.S.: An altruist on Reddit brought to my attention that a more elaborate version of this, with prettier screenshots, can be found in the official documentation⁶.

Recent posts