CRITICAL: `GetLegendGraphicFilter` plugin cannot be run in multithreading mode, skipping.

[Raruto]

Checklist

• I've searched through the existing issues and this bug has never been reported before

Subject of the issue

I'm not sure if this is really the real source of the problem (as you can see in the attached trace log I also see many other warnings), but it seems that by changing the styles of a categorized layer on the g3w-client map makes the server application fatally crash (although this doesn't seem to happen directly right away after the first occurrence of the CRITICAL warning message).

Steps to reproduce

1. Upload new qgis project based on G3WSUITE_3.5.zip
2. Set legend position rendering In a separate TAB

3. Repeatedly change the layer building style between categorized and graduated and after a while the application should crash

In the image above the pending request url is the following:

http://localhost:3000/ows/7/qdjango/10/?SERVICE=WMS&VERSION=1.3.0&REQUEST=GetLegendGraphic&SLD_VERSION=1.1.0&FORMAT=application/json&TRANSPARENT=true&ITEMFONTCOLOR=white&LAYERFONTCOLOR=white&LAYERTITLE=true&ITEMFONTSIZE=10&CRS=EPSG:3857&BBOX=1251272.6926508327,5432351.667251072,1252946.7991937594,5434187.096788135&SYMBOLWIDTH=16&SYMBOLHEIGHT=16&STYLES=Categorized&LAYER=buildings_2f43dc1d_6725_42d2_a09b_dd446220104a

Environment

• g3w-admin: 3.5.dev-20230203102702
• g3w-client: v3.8.0-alpha.1 (ref: Show layer categories also when legend is show on separate TAB g3w-client#319)
• g3w-suite-docker: 3.5.dev
• browser: Firefox
• operating system: Windows 10 (desktop)

Link to your project

https://gis3w-snc.sentry.io/issues/3915018006/?project=109745

Additional info

• docker-log-2023-02-06.txt
• installed-apps.txt
• G3WSUITE_3.5.zip

[Raruto]

It seems that the problem is related to the built-in development server (runserver) which is not able to handle some multithreaded processes correctly:

• g3w-suite-docker/scripts/docker-entrypoint-dev.sh

# Start Django server
python3 manage.py runserver 0.0.0.0:8000

Updating the dev entrypoint script like the following seems to work fine (or at least it doesn't seem to crash so frequently):

• g3w-suite-docker/scripts/docker-entrypoint.sh#L31-L38

gunicorn base.wsgi:application \
    --limit-request-fields 0 \
    --error-logfile - \
    --log-level=debug \
    --timeout ${G3WSUITE_GUNICORN_TIMEOUT:-120} \
    --workers=${G3WSUITE_GUNICORN_NUM_WORKERS:-8} \
    --max-requests=${G3WSUITE_GUNICORN_MAX_REQUESTS:-200} \
    -b 0.0.0.0:8000

I'm closing here because it's more of a problem related to client development, in the future, these kind of inconsistencies between development and production will probably be fixed after g3w-suite/g3w-suite-docker#79 is merged.

👋 Raruto