Description
Why?
After upgrading RabbitMQ to 3.8.17 we noticed that, once in a while, messages were not consumed anymore. The consumers alocated to it died and were not replaced.
This results in emails / updates being delayed.
Possible explanation
For any reason (GC, network failure) some timeout was exceeded and the connection deemed unusable.
Upon connection replacement, no attempt is made to resume the consumer.
Current fix
Restart James.
Expected behaviour
We expect James, upon connection replacement, to happily restart its consumers.
How?
Reconnection handler should also resume consummers.
Attachments
Issue Links
- links to