[Openstack] Messaging reliability/durability expectations

Aaron Knister aaron.knister at gmail.com
Tue Oct 14 21:48:18 UTC 2014


Thanks Chris, Noel for your helpful replies! The fixes to all 3 of these
issues seem to be patches to the rabbit driver for oslo. Are the other
drivers (e.g. qpid) any more robust or are they just not heavily used so
more bugs may be lurking there?

I'd really like to use zeromq but the lack of any message security there
makes me a little hesitant.

-Aaron

On Tue, Oct 14, 2014 at 5:36 PM, Noel Burton-Krahn <noel at pistoncloud.com>
wrote:

> Hi Aaron,
>
> Unfortunately, durable queues don't fix the case where rabbit dies and
> restarts on a new host (and loses its durable queue store)
>
> There's a fix here, but it's been waiting a while for a merge
>
> https://review.openstack.org/#/c/109373/
>
> --
> Noel
>
>
>
>
> On Tue, Oct 14, 2014 at 1:51 PM, Aaron Knister <aaron.knister at gmail.com>
> wrote:
>
>> (Tested in icehouse)
>>
>> On Tue, Oct 14, 2014 at 4:50 PM, Aaron Knister <aaron.knister at gmail.com>
>> wrote:
>>
>>> For those of you following alone at home--  I just discovered that
>>> durable queues are particularly nice for nova scheduler. Without them an
>>> outage of either the MQ daemon (qpid in my case) or the scheduler itself
>>> can cause the scheduling requests to get dropped on the floor. With
>>> durability the instance gets scheduled when the broker or the scheduler
>>> come back to life.
>>>
>>> On Tue, Oct 14, 2014 at 3:32 PM, Gordon Sim <gsim at redhat.com> wrote:
>>>
>>>> On 10/14/2014 07:36 PM, Aaron Knister wrote:
>>>>
>>>>> With RabbitMQ if a message is silently dropped by the broker will a
>>>>> timeout still occur/exception be raised because no reply/ack was
>>>>> received?
>>>>>
>>>>> With the QPID driver the automatic ack()'s Sandy mentioned don't occur?
>>>>> Will the sender eventually become aware that a message was lost? I'll
>>>>> take a timeout over a silent message drop any day.
>>>>>
>>>>
>>>> Yes, for an RPC 'call' (i.e. not a cast, which has no reply) the call
>>>> will block until the response is received or the timeout expires. In the
>>>> latter case an exception is raised. This is independent of which driver is
>>>> used.
>>>>
>>>>
>>>
>>
>> _______________________________________________
>> Mailing list:
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
>> Post to     : openstack at lists.openstack.org
>> Unsubscribe :
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20141014/1565004e/attachment.html>


More information about the Openstack mailing list