nova boot ends in error state when using nova-network VlanManager

[Christoph Bachhuber-Haller]

Steps to reproduce:
- Setup a devstack from scratch using nova-network
- delete the default network
  # nova-manage network delete 10.0.0.0/24
- change nova.conf to use VlanManager:
network_manager = nova.network.manager.VlanManager
- restart nova-network
- create a new network with a vlan id:
nova-manage network create --label=network --fixed_range_v4 10.0.1.0/24 --vlan 42

- boot a vm on the cirros image:
 nova --debug boot --flavor 1 --image 0b969819-2d85-4f7f-af76-125c5bb5789f test

Expected behavior: The new VM goes to Active state
Actual behavior: The new VM goes to Error state, also nova-network log has this exception:
a7-abaf-78db50a4b62c] network allocations from (pid=13676) allocate_for_instance /opt/stack/nova/nova/network/manager.py:494
2014-04-07 15:32:02.137 ERROR nova.network [req-87a65a9e-9196-4203-9de2-f6911d2aef4b admin demo] No db access allowed in nova-network: File "/usr/local/lib/python2.7/dist-packages/eventlet/greenthread.py", line 194, in main
    result = function(*args, **kwargs)
  File "/usr/local/lib/python2.7/dist-packages/oslo/messaging/rpc/dispatcher.py", line 128, in <lambda>
    yield lambda: self._dispatch_and_reply(incoming)
  File "/usr/local/lib/python2.7/dist-packages/oslo/messaging/rpc/dispatcher.py", line 133, in _dispatch_and_reply
    incoming.message))
  File "/usr/local/lib/python2.7/dist-packages/oslo/messaging/rpc/dispatcher.py", line 176, in _dispatch
    return self._do_dispatch(endpoint, method, ctxt, args)
  File "/usr/local/lib/python2.7/dist-packages/oslo/messaging/rpc/dispatcher.py", line 122, in _do_dispatch
    result = getattr(endpoint, method)(ctxt, **new_args)
  File "/opt/stack/nova/nova/network/floating_ips.py", line 119, in allocate_for_instance
    **kwargs)
  File "/opt/stack/nova/nova/network/manager.py", line 497, in allocate_for_instance
    requested_networks=requested_networks)
  File "/opt/stack/nova/nova/network/manager.py", line 1837, in _get_networks_for_instance
    networks = self.db.project_get_networks(context, project_id)
  File "/opt/stack/nova/nova/db/api.py", line 1370, in project_get_networks
    return IMPL.project_get_networks(context, project_id, associate)
  File "/opt/stack/nova/nova/cmd/network.py", line 47, in __call__
    stacktrace = "".join(traceback.format_stack())

I think the exception was introduced by this patch that disables direct database access from nova-network: https://review.openstack.org/#/c/79716/

However, VlanManager still relies on database access for the given scenario, and there are 3 other places in manager.py that rely on direct db access:
devuser@ubuntu:/opt/stack/nova$ grep self.db nova/network/manager.py -n
1389: vifs = self.db.virtual_interface_get_by_instance(context,
1446: vif = self.db.virtual_interface_get_by_address(context,
1837: networks = self.db.project_get_networks(context, project_id)
1914: not self.db.network_in_use_on_host(context, network['id'],

Therefore, I cannot currently use conductor with nova-network VlanManager, which is a regression from Havana.