stable/grizzly patches are failing jenkins in check-tempest-devstack-vm-neutron

There are a lot of sys.exit's in the neutron code so I'm wondering if that's part of the apparent "silent" fail here, the code hits a problem and does a sys.exit which kills the process before the error can be written to the log?

Hi,
I have addressed this with https://review.openstack.org/#/c/49483/.
Thanks
Gary

This is dying somewhere in the middle of a devstack setup while running commands via the quantumclient compat wrapper, and is only affecting stable as far as we've seen, so I'm pretty confident the issue is not on the infrastructure itself. Probably quantumclient, neutron or at worst devstack... so marking this invalid for the openstack-ci project.

Status update from Akihiro Motoki http://lists.openstack.org/pipermail/openstack-dev/2013-October/015991.html

quantum-debug command in stable/grizzly quantum code refers QuantumShell.client_manager.quantum.
client_manager object comes from neutrtonclient/common/clientmanager.py.
Thus the current proxy mechanism from quantumclient to neutronclient does not work.

The easiest fix seems to define QuantumShell class based on NeutronShell to extend authenticate_user() method. It ensures self.client_manager.quantum variable: http://paste.openstack.org/show/48005/

The essential problem is a drect reference of class internal variable and it should be fixed ideally. However it requires changes in several components (neutronclient, grizzly quantum and havana neutron). quantumclient proxy already refers to internal methods in neutronclient, and I think the above solution does not matter much.

I will propose a patch to quantumclient branch from now.

In bug 1234550, the fix in neutronclient was merged.
I confirmed stable/grizzly blocking issue has been addressed.