Open naemono opened 6 years ago
I believe I have run into the same issue, or a similar one.
I think I'm getting a similar issue in both kombu 4.1.0 and 4.2.0 in that the connection appears to degrade and then not be able to recover: https://github.com/celery/kombu/issues/857 I've not managed to get any responses yet though so I'm not sure if anyone has been able to identify the source. Have you tried downgrading to kombu 4.0.2 since this medium post indicates that it may prove effective for some instances:- https://medium.com/squad-engineering/two-years-with-celery-in-production-bug-fix-edition-22238669601d
can anyone tell the update with kombu 4.6.x?
From what I'm seeing, a kombu consumer using amqp doesn't gracefully recover from an issue such as a rabbitmq restart, or any recoverable error without explicitly calling
consumer.consume()
before anydrain_events
calls.Example showing issue
Consumer
Producer
Run the above scripts in parallel, and then restart rabbitmq, and watch the queue fill up with producer messages while the consumer throws no additional errors after
recovery
completes.If you uncomment the
consumer.consume()
line in the consumer, things will automatically recover.Is this expected behavior, or is this an underlying bug within kombu's
Connection.ensure
method where the reconnection actions happen?