Make service liveness checks more consistent / correct

[Stanislaw Pitucha]

The scheduler and nova-manage check for service being alive in different ways. Additionally there are issues when the host checking the liveness has clock set N seconds in the future and the checked service just updated the timestamp. Negative time differences (-1 day, +many seconds) were not accounted for properly.

[Thierry Carrez (ttx)]

Note that synchronized clocks is a basic requirement.

[Stanislaw Pitucha (stanislaw-pitucha)]

Fix proposed in https://review.openstack.org/#change,791

[Openstack Gerrit (openstack-gerrit)]

Reviewed: https://review.openstack.org/791
Committed: http://github.com/openstack/nova/commit/ad74424768463cd6ad02dd9d6fdf64fc7b3bbafb
Submitter: Jenkins
Branch: master

 status fixcommitted
 done

commit ad74424768463cd6ad02dd9d6fdf64fc7b3bbafb
Author: David Subiros <email address hidden>
Date: Fri Aug 5 16:35:56 2011 +0100

    Improve the liveness checking for services

    With this modification both nova-manage and scheduler use the flag
    service_down_time and check positive and negative values in a
    correct way.
    Fixes bug: 867674.

    Change-Id: I15c48d80cafa2089cd228c09c61b0a1e513730e8