Refactor instance usage notifications for compute manager

This patch makes compute manager call notify_usage_exists() via the
conductor instead of directly, avoiding the direct database access
required to fetch bandwidth information. I initially just added
a notify_bw_usage() method to conductor, which just handled the
bandwidth part (which is the database hit). However, the metadata
payload part may be depended upon by custom stuff for periodic
notifications, so copying all that stuff as well made it a bunch
of duplicated stuff. Thus, I decided to just make it a direct proxy
for the compute_utils method.

Note that this is a bit of a departure from the previous pattern of
just proxying db interfaces in conductor, but I think the simplicity
gain is worth it.

Related to blueprint no-db-compute

Change-Id: I09f2652d4903f4d6036c6e3d74db063952341856

0 files changed, 0 insertions, 0 deletions