[Openstack] Messaging reliability/durability expectations
sandy.walsh at RACKSPACE.COM
Tue Oct 14 15:39:35 UTC 2014
Openstack RPC-over-AMQP (oslo.messaging) automatically ack()'s all messages that are received. So, it becomes the responsibility of the sender to retry. For example, the scheduler in Nova does this. However, if the client fails before getting the message, AMQP will automatically requeue it and send it to another client. But that's an unlikely occurrence as it's queue-per-host so there wouldn't be a second client. Therefore, queues should be configured to be durable.
Hope it helps!
From: Aaron Knister [aaron.knister at gmail.com]
Sent: Tuesday, October 14, 2014 12:00 PM
To: Raghu Vadapalli
Cc: <openstack at lists.openstack.org>
Subject: Re: [Openstack] Messaging reliability/durability expectations
Thanks Raghu. I think I might not be asking the right questions. Part of my ignorance here comes from not understanding AMQP. I think really what I'm trying to figure out is whether openstack expects durable queues. It sounds like the answer is no but confirmation of this would be great. Even if it doesn't expect durable queues is there a noticeable advantage to openstack to have them?
Sent from my iPhone
On Oct 14, 2014, at 10:12 AM, Raghu Vadapalli <rvatspacket at gmail.com<mailto:rvatspacket at gmail.com>> wrote:
On Tuesday, Oct 14, 2014 at 7:49 AM, Aaron Knister <aaron.knister at gmail.com<mailto:aaron.knister at gmail.com>>, wrote:
Thanks Remo, could you elaborate a little? Which is part of RabbitMQ? The HA layer or the message retransmission? I'm currently using qpid. Also, just so I'm clear, is it the openstack code or the low-level messaging drivers (rabbit, zmq, qpid) that retransmit on message delivery failure?
That should be the job messaging infrastructure provided by rabbit or qpid
Sent from my iPhone
On Oct 13, 2014, at 11:59 PM, Remo Mattei <remo at italy1.com<mailto:remo at italy1.com>> wrote:
That is part of RabbitMQ and yes it will resend the msg.
On Oct 13, 2014, at 22:31, Aaron Knister <aaron.knister at gmail.com<mailto:aaron.knister at gmail.com>> wrote:
I'm building a production-grade cloud where HA is a requirement. I'm currently working on implementing HA of the messaging layer and am not clear on what the expectations/assumptions are of the messaging layer regarding message durability and reliability. I've seen documentation scattered about that suggests OpenStack makes no assumptions that messages will be delivered or will need to persist for any period of time (https://openstack.redhat.com/Highly_Available_Qpid_for_OpenStack).
What's the consequence of RPC message loss within openstack? Is it true that it will resend messages?
Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to : openstack at lists.openstack.org<mailto:openstack at lists.openstack.org>
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Openstack