Closed jrouly closed 3 years ago
After attaching to the live container and poking through the code there some, it looks like it might be out of date with latest master
in this repo. Is the publication schedule for rhasspy/rhasspy:latest
out of sync with merges?
Hi @jrouly, thanks for looking into this. I've had a lot of weird cases with Quart's websockets, and this is the first time I'm seeing them triggering an exception because the server is interpreting them as a request for a static file. I don't see this at all when I run outside a container on my machine.
Is the publication schedule for
rhasspy/rhasspy:latest
out of sync with merges?
Nothing automated, unfortunately. Myself and other Rhasspy developers have tried getting the build process onto Github, but it's pretty ugly and long. It's a docker buildx
process that simultaneously builds images for amd64
, arm/v7
, arm64
, and arm/v6
with a lot of Python dependencies and pre-compiled binaries. And now I'm trying to shove PyTorch in there :smile:
P.S. You can fight some of the log spam by passing some additional command-line arguments:
$ docker run ... rhasspy/rhasspy ... -- --log-level info
Note the --
separator that ensures everything after is passed directly to rhasspy-server-hermes
. Additionally, you can edit the supervisord.conf
file in your profile and adjust logging there; it's re-generated each time you save your profile or restart, so pass --nogenerate-conf
alongside --profile
, etc. to avoid overwriting your changes.
Environment: running
rhasspy:latest
(be3bbfdf4e23
) on docker.Symptom: whenever the browser is open to the Rhasspy web server, websocket connections are made to:
These websocket requests appear to fail with
Unexpected response code 500
, but that's beside the point. Nothing appears the matter in the browser for the most part.The thing that's really problematic is the logspam in the container output, which makes it very difficult to track what's going on:
I was able to narrow down the websocket requests as the problem by attaching to the running container and creating the directories that the
FileNotFoundError
was complaining about (static directories inside the virtual environment), which resulted in these newer errors:and then creating those subsequent directories resolved the second level of exceptions. But I'm still seeing a ton of
NotFound(404)
error messages in container output, presumably for each of those websocket calls.