Some log levels not correctly setup from flags (e.g. nova.utils)

Some loggers are created before the flags are read, so are not set to DEBUG even when running with the --verbose flag set.

I added a logging.warn statement to log._get_level_from_flags to see/verify this, and these loggers are created before the FLAG variables are dumped (which I presume is immediately after they are set): nova.root, nova.exception, sqlalchemy, nova.utils. amqplib, nova.fakerabbit, nova.rpc.

This is for the compute node (I haven't checked if other daemons exhibit the same problem or if they have a different set of flags)

Running against trunk.

WARNING:root:LOG nova.root => INFO
2011-02-03 17:16:03,818 WARNING nova.root [-] LOG nova.exception => INFO
2011-02-03 17:16:03,873 WARNING nova.root [-] LOG sqlalchemy => WARN
2011-02-03 17:16:03,913 WARNING nova.root [-] LOG nova.utils => INFO
2011-02-03 17:16:04,006 WARNING nova.root [-] LOG amqplib => WARN
2011-02-03 17:16:04,056 WARNING nova.root [-] LOG nova.fakerabbit => INFO
2011-02-03 17:16:04,057 WARNING nova.root [-] LOG nova.rpc => INFO

2011-02-03 17:16:04,096 AUDIT nova.root [-] Starting compute node (version 2011.1-LOCALBRANCH:LOCALREVISION)
2011-02-03 17:16:04,096 DEBUG nova.root [-] Serving nova-compute from MainProcess (pid=29867) serve /home/justinsb/lefthand/nova/nova/nova/service.py:224
2011-02-03 17:16:04,097 DEBUG nova.root [-] Full set of FLAGS: from MainProcess (pid=29867) serve /home/justinsb/lefthand/nova/nova/nova/service.py:225
2011-02-03 17:16:04,097 DEBUG nova.root [-] rabbit_retry_interval : 10 from MainProcess (pid=29867) serve /home/justinsb/lefthand/nova/nova/nova/service.py:228
...<snip>...
2011-02-03 17:16:04,101 DEBUG nova.root [-] verbose : True from MainProcess (pid=29867) serve /home/justinsb/lefthand/nova/nova/nova/service.py:228