Nova Network crash when instances are created

[ArKam]

Using fedora 19 Schrodïnger's Cat and OpenStack Grizzly (2013.1.3) has a host for nova-compute / nova-network, it is impossible to use nova-network has network manager.

AMQP Broker is Qpidd C++ Broker

As reported to this topic on ask.openstack.org and openstack.redhat.com:

https://ask.openstack.org/en/question/5288/nova-network-is-crashing-hard/

http://openstack.redhat.com/forum/discussion/560/launch-instance-error-with-fedora-19-image-in-openstack-grizzly/p1?post#Form_Body

Nova-Network seems to use a bad formatted message which result in a crash exception on SQLAlchemy 0.8.2 used by Fedora 19.

Here is the error message we receive when we try to create a new instance:

http://paste.openstack.org/show/n3XI87B6Ss8s94C85VqL/

has we can read it from the python stack from line 49 to 55 SQLAlchemy seems to receive a bad formatted message from the nova.openstack.common.rpc.common.amqp module.

So it seems that Nova Network is incorrectly formatting its messages.
This can't be a qpid missconfiguration as any other services (Glance/Cinder) are able to communicate and create volumes and images correctly, plus, if we disable network module, OS is able to create correctly the instances without network so RPC are passing through AMQP correctly.

Here is the Nova.conf file use:

http://paste.openstack.org/show/nQyKLdxGjkb4VEgik5dv/

Our Compute node is the CLUNODE02 here, and here is its IP Configuration:

http://paste.openstack.org/show/FjivZQkAfXhDeaOLiRXF/

Here is the result of a nova-manage network list:

http://paste.openstack.org/show/Vn7WoLtBLDI0oWGOLZ3D/

As indicate on the question on ask Q&A Website, we cannot perform a SQLAlchemy regression on FE19 due to OS Dependencies which then are broken.

[Joe Gordon (jogo)]

Grizzly has been end of lifed, is this bug still valid?