Open Stack

Ryan,

Can you describe your AMQP setup? Are you using RabbitMQ, QPID, or
ZeroMQ? What sort of HA do you have set up?

I noticed this behavior previously in Grizzly with incorrect federated
queue configuration in QPID where all schedulers would accept all nova
boot requests, then the last one marks the instance as a failure even
though it launched okay.

Best Regards,
Matthew Mosesohn

On Wed, Mar 12, 2014 at 2:12 AM, Aydelott, Ryan M. <ryade at mcs.anl.gov> wrote:
> We are running 3 nova schedulers against a rabbitmq cluster of 3, it would appear that only 1 of every three requests to boot a vm fail. Stopping two of the three schedulers brings us back to 100% spawn rate, whereas with all three it goes:
>
> fail
> fail
> success
> fail
> fail
> success
> ...
>
> Is this something that is a known good configuration to use, or should we be trying something else here (HA/VIP's).  Perhaps I'm missing something obvious as well?
>
>
> -ryan
>
> _______________________________________________
> OpenStack-operators mailing list
> OpenStack-operators at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>