Description
Describe the bug
Hi Team,
We have an RPC setup (similar to https://github.com/rabbitmq/rabbitmq-tutorials/blob/main/java/RPCClient.java) where we create a new reply queue (classic, not server-named, non-durable, exclusive, auto-delete) for every request to our server.
Recently we've tried to setup federation on those queues. Functionally it appears to work very well, but we've noticed that the connection count (and memory usage) keeps growing. The queues do get removed at the end of the request as expected.
You can clearly see from the graphs below where we enabled the federation
Is this an issue with our setup or could it be an issue with federation?
We're using RabbitMQ 4.1.1 Erlang 27.3.4.1 on two 3-node clusters. Happy to provide additional info.
Reproduction steps
To be honest, I haven't had a chance yet to try it in a more controlled environment. Will try it ASAP.
- Setup queue federation between two clusters
- Have an application continuous create and queues matching the federation policy
- Observe connection count and memory usage
Expected behavior
Expecting additional resource usage because of federation to be proportional to the number of active federated queues/links
Additional context
No response