This is a known issue in Dify v1.9.2 Docker deployments. The most common cause is that the Celery worker processes are not subscribed to both the pipeline and priority_pipeline queues simultaneously, causing uploaded files to remain stuck in the queue without any activity appearing in the worker logs. To resolve this issue, update the worker configuration in entrypoint.sh or docker-compose.yaml to ensure the workers listen to both queues. For example:
-Q ${CELERY_QUEUES:-dataset,priority_pipeline,pipeline,mail,ops_trace,app_deletion,plugin,workflow_storage,conversation} \
Alternatively, add a dedicated priority-worker service in your docker-compose.yaml to specifically listen to the priority_pipeline queue:
priority-worker:
image: langgenius/dify-api:1.9.2
restart: always
environment:
<<: *shared-api-worker-env
MODE: worker
CELERY_QUEUES: priority_pipeline
LOG_LEVEL: INFO
depends_on:
db:
condition: service_healthy
redis:
condition: service_started
volumes:
- ./volumes/app/storage:/app/api/storage
networks:
- default
After modifying the configuration, restart all containers and check the Celery worker logs to confirm they are listening to both queues. If the issue persists, stop all containers, remove old images/volumes, and restart from scratch to clear any outdated configurations. Additionally, ensure all required services (Redis, Vector DB, embedding provider) are running and healthy, and verify that your embedding model API key is valid and supported by the pipeline knowledge base. If you are using Weaviate v4.x, ensure your collection schema includes a named vector called <vector_name> with the name default; otherwise, the pipeline will fail to process data and remain queued. It is recommended to upgrade to the latest Dify version, as newer versions improve error handling and pipeline reliability.