[rabbitmq-discuss] IPv6 issue?

[Simon MacMullen]

Hi Andrea.

On 16/11/11 15:48, Rosa, Andrea wrote:
> I had some communication problem with rabbitmq, from netstat command I
> was able to verify that the connection between client and server was
> ESTABLISHED, but form the client the connection was using TCP and from
> Rabbimq the connection was using TCP6.

This is rather OS-dependent. Are we talking about Linux?

On Linux (by default, because on Linux everything's configurable) IPv4 
is considered to be a tiny subrange of IPv6. Since we added IPv6 support 
in 2.3.1 we listen on "::" by default (i.e. all IPv6 addresses) and IPv4 
is included in that. This means that some OS tools might report 
connections as being IPv6 when they are "really" IPv4.

On other OSes (and Linux when it's been configured differently) we 
listen on both "::" and 0.0.0.0. We can't do that on default-Linux 
because binding to 0.0.0.0 fails.

> In this situation my client wasn’t able to publish messages.
>
> I changed my rabbitmq configuration file adding the following line:
>
> RABBITMQ_NODE_IP_ADDRESS = 0.0.0.0
>
> and I restarted the server and the problem has been fixed.
>
> Anyone has experienced with this similar issue before?
>
> Could you confirm that the change I did, can be used to enforce rabbitmq
> to use just IPv4?

Anyway, yes you can use RABBITMQ_NODE_IP_ADDRESS to force the server to 
use IPv4 only, or add something like

  {tcp_listeners, [{5672, "0.0.0.0"}]}

to rabbitmq.config.

I would like to figure out why this isn't working for you though - the 
TCP6 thing *should* be a red herring. Could this have been some 
transient problem that went away by coincidence? Or am I being optimistic?

> I have installed the rabbitmq 2.6.1 version, the client is Kombu python
> library using the py-amqplib.

Cheers, Simon

-- 
Simon MacMullen
RabbitMQ, VMware