[openamq-dev] round-robin queues

[Martin Sustrik]

Hi Terry,
> Martin> 1. that messages would be lost when a service instance fails
> Martin> 2. that messages would be round-robined to a non-responsive service 
> Martin> instances
>
> I'm wondering if you noted my 1 above (use auto-delete queues)? If so, then
> can you tell me how your 1 and 2 occur?  I guess it could be something with
> how quickly the TCP/IP protocol allows you to handle a remote machine going
> offline?
>   
1. The auto-delete queues are exactly the problem. When the client 
fails, the queue is deleted and so are the messages in the queue.

2. If the service is non-responsive, but the TCP connection is still 
alive, the queue won't be destroyed (at least until missing heartbeats 
cause it to be torn down) therefore the messages would still be 
round-robined to it.

Martin