[rabbitmq-discuss] Rabbit Cluster "hangs" on consumer disconnects

[Simon MacMullen]

On 01/11/2013 5:45PM, Cordell, Ron wrote:
>             ** Reason for termination ==
>             ** {{shutdown,
>                      {gen_server,call,
>                          [rabbit_auth_backend_ldap,
>                           {check_resource,
>                               [{username,<<"SJINRMQVORTEXWRITE">>},
>                                {user_dn,
>                                    "CN=SJINRMQ
>             VORTEXWRITE,OU=INRMQ,OU=INTEGRATION,OU=Environments,OU=Servers,DC=RHB,DC=AD"},
>                                {vhost,<<"vortex">>},
>                                {resource,exchange},

So that looks like a bug in the LDAP plugin (the plugin itself shuts 
down before the channels which depend on it). We'll get that fixed, but 
I doubt it's causing the problems you're seeing.

> I’ve verified that the application code is shutting down connections
> cleanly. We also don’t see this behavior with the same application in
> environments where no rabbit cluster is setup, just a single instance.
> This appears to be something that happens to Rabbit itself – the Erlang
> portion of things seems to be working OK.
>
> I’m attempting to replicate this behavior in other environments so I can
> try to isolate it but haven’t had a chance to do so thus far, but it
> does happen in every environment where we have a rabbit cluster.

What happens if you attempt to shut down the service by issuing 
"rabbitmqctl stop" rather than stopping the Windows service?

Also, we fixed several bugs that could cause deadlocks during shutdown 
in 3.2.0, so it might be worth trying an upgrade.

Cheers, Simon

-- 
Simon MacMullen
RabbitMQ, Pivotal