Introduction of Granian in v42 introduce container startup delay/issue

[leplatrem]

an error occurs during startup with docker compose:

remote-settings ->        | urllib3.exceptions.MaxRetryError: HTTPConnectionPool(host='localhost', port=8888): Max retries exceeded with url: /v1/buckets/main-workspace/collections/quicksuggest (Caused by NewConnectionError("HTTPConnection(host='localhost', port=8888): Failed to establish a new connection: [Errno 111] Connection refused"))

/cc @luc-lisi

┆Issue is synchronized with this Jira Task

[leplatrem]

Root cause:

• uwsgi binds the listening socket in the master process before the app finishes importing
• Granian binds once the APP is loaded

For a few msecs, Granian is importing the python code before it spawns the worker.

We can see with this:

# slow.py
import time

print("importing...")
time.sleep(5)
print("done")

async def app(scope, receive, send):
    assert scope['type'] == 'http'

    await send({
        'type': 'http.response.start',
        'status': 200,
        'headers': [
            [b'content-type', b'text/plain'],
        ],
    })
    await send({
        'type': 'http.response.body',
        'body': b'Hello, world!',
    })

$ uv run granian --interface asgi slow:app

[INFO] Starting granian (main PID: 39575)
[INFO] Listening at: http://127.0.0.1:8000
[INFO] Spawning worker-1 with PID: 39579
importing...
done
[INFO] Started worker-1

And while "importing", connection is refused.
Granian does not seem to have any option/flag to change that.

Short term bandage will be to add a heathcheck in docker compose.

But maybe we could look into our code base to improve load time by using lazy imports (for example Sentry probes installed frameworks at import time)