rpc.register_opts race in service.py causes NoneType exception

[Dan Prince]

Using the latest Nova Folsom. I'm seeing a race condition where sometimes rpc.register_opts isn't called quick enough on startup causing the following error:

2012-05-06 02:29:24 INFO nova.wsgi [-] Started S3 Objectstore on 172.19.0.3:3333
2012-05-06 02:29:25 AUDIT nova.service [-] SIGTERM received
2012-05-06 02:29:25 INFO nova.wsgi [-] WSGI server has stopped.
2012-05-06 02:29:25 CRITICAL nova [-] 'NoneType' object has no attribute 'rpc_backend'

----

The issue seems to be that 'wait()' gets called while service.start is still running. The issue *only* occurs if you try to use the RPC defaults (as in not specifying rpc_backend in the nova.conf file).

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/7168

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/7168
Committed: http://github.com/openstack/nova/commit/9a9a28e80ae043bdf7476651ea8d92aedf0ba36c
Submitter: Jenkins
Branch: master

commit 9a9a28e80ae043bdf7476651ea8d92aedf0ba36c
Author: Dan Prince <email address hidden>
Date: Sat May 5 22:42:36 2012 -0400

    Fix a race with rpc.register_opts in service.py.

    Fixes an issue in service.py where rpc.register_opts may not
    get called before wait() accesses FLAGS... which can cause
    NoneType exceptions with some versions of nova.conf.

    Fixes LP Bug #995328.

    Change-Id: Ib177c2a082a4cf1c22935c1e6fdf3ba714a33aca