Code review comment for lp:~stub/charms/trusty/postgresql/rewrite

Looking at http://data.vapour.ws/charm-test/charm-bundle-test-azure-651-all-machines-log,
it seems debugging output has been turned up to 11, and I think this
breaks 'juju wait'. Without 'juju wait', it is impossible to get the
tests running reliably

Its hard to tell through all the noise, but I think here is where the
initial deploy actually terminates:

unit-postgresql-0[21563]: 2015-09-17 01:01:47 INFO
unit.postgresql/0.juju-log server.go:254 db-admin:2: *** End
'db-admin-relation-changed' hook

After this, we keep seeing noise about leadership leases, and other
stuff including the juju run commands that are checking that the logs
are quiet (thus shooting itself in the foot, because the only way to
check if the logs are quiet adds noise to the logs...).

This never got picked up locally, as I don't even know how to turn on
this output and don't get it here (using juju stable).

Also see Bug #1496130, where we are fixing things so the OpenStack
mojo tests can use the wait algorithm.

What is confusing me though is that the Cassandra tests last passed
Sep 4th, and I would have expected them to have started failing much
earlier (it too relies on juju wait). What is even more confusing, the
last successful run at
http://reports.vapour.ws/charm-test-details/charm-bundle-test-parent-703
leads me to the AWS log at
http://data.vapour.ws/charm-test/charm-bundle-test-aws-591-all-machines-log
which contains a whole heap of logging information from other tests,
in addition to the Cassandra logs, which means test isolation was
busted and the results bogus.