Silent Data Loss in Celery + RabbitMQ on Azure Container Apps: A Queue Registration Post-Mortem

Short summary

A detailed post-mortem of a production incident where Celery tasks on Azure Container Apps silently dropped data because a stale cloud resource field caused the worker to not consume specific queues. The API returned 202 Accepted, RabbitMQ accepted messages, but no worker was listening — and nothing in the stack raised an error. Root cause was a mismatch between declared task queues and the -Q flag in the worker entrypoint.

• •Silent data loss occurred because worker -Q flag didn't include all task queues
• •Every layer returned success while messages went unconsumed
• •Root cause was a stale field on a cloud platform resource, not application code

Generated with AI, which can make mistakes.