Guest agent still depends on trove models

[Robert Myers]

There is still a couple spots where the guest agent is trying to import models used in trove. Namely the backup models and the agent heartbeat. None of the code in the guest actually uses the database models but all the logic gets loaded for the sqlalcemy models and causes the guest binary to bloat on the host by about 20 megs of RSS.

Moving the backup state to a separate module and removing an used function lowers the memory usage by about 15 - 20 Megs.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/124777

[Nikhil Manchanda (slicknik)]

Moving to kilo, since this is not blocking Juno RC1 release.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/124777
Committed: https://git.openstack.org/cgit/openstack/trove/commit/?id=0007a5076aa795fc354b122ae591980c9413f054
Submitter: Jenkins
Branch: master

commit 0007a5076aa795fc354b122ae591980c9413f054
Author: Robert Myers <email address hidden>
Date: Sat Sep 27 19:25:59 2014 -0500

    Removing dependency on trove models in the guest agent

    Reasons:
     - The guest agent is importing backup models and agent
       heartbeat, this triggers the all of the trove database
       setup logic which bloats the guest process on the host.

     - Moving the state to its own module and removing an unused
       function decreases the memory usage by about 15 - 20 megs.

    Closes-Bug: #1375311
    Change-Id: I126c0b89c170b325d85b3f09afca399b4f5de9e8