🚨 Queue Job tasks stuck in "pending" state in Docker Swarm with multiple nodes (Odoo 17)

Module

queue_job

Describe the bug

In a Docker Swarm environment with three nodes, one Odoo instance, Pgpool, and Traefik, asynchronous tasks remain stuck in the pending state in the queue. Even after a requeue, they are not processed. The only way to restart task processing is to restart the Odoo service, which sometimes results in task duplication.

This issue does not occur in a Docker Compose environment with a single host, where tasks are processed normally after configuring the workers.

To Reproduce

Affected versions: Odoo 17, queue_job v17

Steps to reproduce the issue:

1. Deploy Odoo 17 with queue_job in a Docker Swarm environment with three nodes, a Pgpool instance, and a Traefik reverse proxy.
2. Configure Odoo workers and queue_job with defined channels.
3. Trigger an asynchronous task.
4. Observe that the task remains in the pending state and is never processed, even after a requeue.
5. Restart the Odoo service. At this point, some tasks are eventually executed, but sometimes duplicated.

Expected behavior

Pending (pending) tasks should be processed by the workers without requiring an Odoo restart and without the risk of duplication.

Additional context

Docker Compose Environment (Working Fine):

• Infrastructure: 1 host
• Resources: 4 CPU, 32 GB RAM
• Workers: workers=6
• Queue Job Channels: ODOO_QUEUE_JOB_CHANNELS: root:2,enrollment:2,notification:2

Docker Swarm Environment (Problematic):

• Infrastructure: 3 nodes (each with 4 CPU, 32 GB RAM)
• Workers: workers=10
• Queue Job Channels: ODOO_QUEUE_JOB_CHANNELS: root:4,enrollment:4,notification:2
• Load Balancing & Proxy: Using Pgpool and Traefik

Logs & Debugging:

• Logs show that tasks are correctly added to the queue, but no worker processes them.
• Enabling log-handler=queue_job:DEBUG did not provide any conclusive insights.

Any help or debugging suggestions would be greatly appreciated! 🚀