[rabbitmq-discuss] Restarting/re-establising failed federation connections

Thu Jul 19 16:04:18 BST 2012

I am getting similar failures to reconnect.

My Virginia cluster is upstream from my California cluster. And there is a
link the other way, making the federation bi-directional.

After some days of idleness (still in testing mode) both links went down.

Restarting the California rabbit reestablished the link to Virginia and I
could send federated messages from Virginia to California.

The California to Virginia link was still down, however. I ran an AMQP
client on the Virginia machine and could connect to California.

This actually happened previously. In response I upgraded all my rabbits
to 2.8.4 and Erlang R14B04.

Now that it has happened again I will restart everything and capture clean
logs etc.

In fact, the situation appears to be consistent as Virginia is the center
of a star federation network that includes California, Brazil, Dublin, and
Tokyo, all configured identically.

All links have failed in exactly the same manner.

Michael

On 7/19/12 6:41 AM, "Matthias Radestock" <matthias at rabbitmq.com> wrote:

>David,
>
>On 18/07/12 01:56, David Gillies wrote:
>> I've posted up the logs from the downstream server from around the time
>> that one of the upstream servers lose connectivity and the downstream
>> server is never able to connect again here:
>> https://gist.github.com/85da45443b92e69e29a0
>
>Please use the correct terminology, it gets really confusing otherwise.
>In federation, messages flow from the upstream to the downstream (makes
>sense, doesn't it?), and connections are established from the downstream
>to the upstream.
>
>Anyway, There is nothing unusual in the logs...the downstreams (using
>the correct terminology) lose connectivity and try to re-establish it.
>They *should* keep trying indefinitely. Is there any evidence that they
>don't?
>
>There is nothing special about the way federation establishes
>connections, so if at some point you think that a downstream should be
>able to connect to the upstream but doesn't, then run an ordinary amqp
>client on the downstream machine and see whether it can connect to the
>upstream.
>
>Regards,
>
>Matthias.
>_______________________________________________
>rabbitmq-discuss mailing list
>rabbitmq-discuss at lists.rabbitmq.com
>https://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss