Draft: Add a Docker entrypoint

• Default to Python 3.10, drop 3.8
• Add a Docker entrypoint
• Don’t install the config.

The Docker entrypoint saves effort for users: they can map a config inside a container based on the stock image without having to rebuild it.

Since the config is no longer necessary inside the image, don’t install it.

added 1 commit

• 1bca1c5b - Add a supervisord-managed entrypoint, drop config from the image

Compare with previous version

changed the description

Can you explain the reason of putting supervisord into the image itself?

Here, it’s mostly historical reasons — I’m just bringing in what we’ve been running for some time.

The benefit of running supervisord is that you can restart the service without killing the container itself, this is sometimes beneficial when you want to debug things while being inside.

For some other services it’s not as useful as it is in this case, but with the proxy the need to debug things live is, sadly, sometimes necessary.

I think this is a bit of anti-pattern of doing things; I'm also confused what the difference is between restarting the container as a whole vs. just restarting in the container as there is only one service running it it?

Restarting the container will also terminate all docker exec sessions running in it.

added 5 commits

• 1bca1c5b...39574ac1 - 4 commits from branch main
• 2f210360 - Add a supervisord-managed entrypoint, drop config from the image

Compare with previous version

added 2 commits

• 970af219 - Use ASGI scope instead of Host header to reliably detect server port
• cba18c6f - Add a supervisord-managed entrypoint, drop config from the image

Compare with previous version

added 2 commits

• 9b090b4e - Use ASGI scope instead of Host header to reliably detect server port
• 74adaec2 - Add a supervisord-managed entrypoint, drop config from the image

Compare with previous version

added 4 commits

• 81fea97b - 1 commit from branch main
• 736305ad - Add upload buffering, enabled by default on clients
• 2e7d1fd6 - Set Content-Type/Length headers for the payload
• 0ec174a3 - Add a supervisord-managed entrypoint, drop config from the image

Compare with previous version

added 2 commits

• 2c508705 - Log worker id on each JSON-RPC call
• d8c44cfa - Add a supervisord-managed entrypoint, drop config from the image

Compare with previous version

added 3 commits

• 7e286248 - Log worker id on each JSON-RPC call
• 2ea4785b - Add missing import of upload_data
• 372a7a67 - Add a supervisord-managed entrypoint, drop config from the image

Compare with previous version

added 30 commits

• 372a7a67...a68769d5 - 29 commits from branch main
• eb1efd3b - Add a supervisord-managed entrypoint, drop config from the image

Compare with previous version

added 1 commit

• 994afb05 - Ensure the websocket context is present in websocket subtasks

Compare with previous version

added 1 commit

• 655dced8 - Ensure the websocket context is present in websocket subtasks

Compare with previous version

added 1 commit

• ce063c3f - Increase httpx pool limits tenfold to avoid pool clogging

Compare with previous version

added 5 commits

• ce063c3f...fdc40110 - 4 commits from branch main
• 5664481a - Add a supervisord-managed entrypoint, drop config from the image

Compare with previous version

added 4 commits

• 5664481a...381cfd06 - 3 commits from branch main
• 6f231338 - Add a supervisord-managed entrypoint, drop config from the image

Compare with previous version

added 3 commits

• 6f231338...d07af15f - 2 commits from branch main
• 8a5020e9 - Add a supervisord-managed entrypoint, drop config from the image

Compare with previous version

added 4 commits

• 8a5020e9...d296c08a - 3 commits from branch main
• a2773f1e - Add a supervisord-managed entrypoint, drop config from the image

Compare with previous version

fwiw having an entrypoint would still make sense though that really could just be either the client or server being run from a respective client or server container. (can be the same layers otherwise)

Also for tls termination; a better way to do that should be figured out; For most setups i would expect some frontend will terminate tls rather then the proxy itself in the first place

marked this merge request as draft