High server load after upgrading to 5.13.3

[mathilde-cryptpad]

Hello,

We've recently upgraded our Weblate instance and have been since then facing a high server load which results in service disruptions. We are using Docker and granian seems to be putting a lot of pressure on the server. With two processes running at almost max CPU all the time.

Here is our docker-compose.yml:

version: '3'
services:
  weblate:
    image: weblate/weblate
    tmpfs:
      - /run
      - /tmp
    ports:
      - "127.0.0.1:8080:8080"
    volumes:
      - weblate-data:/app/data
    env_file:
      - ./environment
    read_only: true
    restart: always
    depends_on:
      - database
      - cache
  database:
    image: postgres:17-alpine
    env_file:
      - ./environment
    volumes:
      - postgres-data:/var/lib/postgresql/data
    restart: always
  cache:
    image: redis:7-alpine
    read_only: true
    restart: always
    command: ["redis-server", --save, '60', '1']
    volumes:
      - redis-data:/data
volumes:
  weblate-data: {}
  postgres-data: {}
  redis-data: {}

Maybe it's tied to the great amount of requests we are receiving. Probably a lot of AI crawlers and other data-mining leeches, hammering our Nginx reverse proxy. We didn't had the issue before 5.13, is Granian is more sensitive to these requests? More resources consuming?

Thanks in advance for your help!

[nijel]

Granian has slightly less overhead than gunicorn, so it should not make the situation worse. Maybe the AI crawler's traffic has increased? You might want to integrate Anubis to deal with the AI traffic, see https://docs.weblate.org/en/latest/admin/install/docker.html#anubis-docker-container-integration

[nijel]

Moreover, we were also tweaking granian configuration in point releases. Make sure you use the 5.13.3.5 Docker tag.