Open Stack

On Tue, Nov 20, 2012 at 5:32 PM, Mike Wilson <geekinutah at gmail.com> wrote:

> A direct exchange instead of a topic exchange, sorry, I'm still feeling my
> way around AMPQ.
>

I don't think changing the exchange type will solve the problem. I'm also
not sure it's a good idea to raise an exception immediately in the sender
because that eliminates the benefit of the loose coupling between sender
and receiver that the message bus is bringing in the first place (by which
I mean you won't be able to restart a compute agent if something is trying
to talk to it). The sender should, eventually, time out if it does not
receive a response.

Making the queue durable should cause the message bus to hold onto the
messages and allow the receiver to be restarted without losing messages.

Watch out for upgrade issues with changing the default setting for durable.
IIRC, all subscribers to a queue must use the same value or they get an
error trying to attach to the queue.

Doug


>
> I will also test just setting the durable property. I have a feeling that
> I read something somewhere that indicated to me that I just had to get off
> of the topic exchange but we will see. Thx.
>
> Mike Wilson
> Bluehost.com
>
>
> On Tue, Nov 20, 2012 at 2:58 PM, Russell Bryant <rbryant at redhat.com>wrote:
>
>> On 11/20/2012 04:46 PM, Mike Wilson wrote:
>> > The topic exchange is what causes the behavior I noted. If no one is
>> > listening on a topic then it doesn't have a route, therefore throw it
>> > away with no error. I'm not familiar with notifier, but if it addresses
>> > a node topic then it should work just as well with a node queue.
>>
>> Notifications are definitely topic oriented.
>>
>> > I will do some more research into what expectations we have around our
>> > communication, but I think at least I am going to try to convert our
>> > installation to use queues for node communication and see how it goes.
>> > I'll probably have something interesting to say about it in a few days
>> :-).
>>
>> Well, we're using queues, it's just queues on a topic exchange.
>>
>> I think Doug Hellman had a good point in his message.  Maybe we just
>> need to make all of the queues durable by default?
>>
>> --
>> Russell Bryant
>>
>> _______________________________________________
>> OpenStack-dev mailing list
>> OpenStack-dev at lists.openstack.org
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>
>
>
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20121120/49f5f7a2/attachment.html>