Comment 20 for bug 1559136

Reviewed: https://review.openstack.org/324646
Committed: https://git.openstack.org/cgit/openstack/fuel-library/commit/?id=fee92985a15fbad8dedaa3dc29b66ddbb6b89182
Submitter: Jenkins
Branch: stable/mitaka

commit fee92985a15fbad8dedaa3dc29b66ddbb6b89182
Author: Dmitry Mescheryakov <email address hidden>
Date: Thu Apr 14 20:12:49 2016 +0300

    Rank master score based on start time

    Right now we assign 1000 to the oldest nodes and 1 to others. That
    creates a problem when Master restarts and no node is promoted until
    that node starts back. In that case the returned node will have score
    of 1, like all other slaves and Pacemaker will select to promote it
    again. The node is clean empty and afterwards other slaves join to
    it, wiping their data as well. As a result, we loose all the messages.

    The new algorithm actually ranks nodes, not just selects the oldest
    one. It also maintains the invariant that if node A started later
    than node B, then node A score must be smaller than that of
    node B. As a result, freshly started node has no chance of being
    selected in preference to older node. If several nodes start
    simultaneously, among them an older node might temporarily receive
    lower score than a younger one, but that is neglectable.

    Also remove any action on demote or demote notification - all of
    these duplicate actions done in stop or stop notification. With these
    removed, changing master on a running cluster does not affect RabbitMQ
    cluster in any way - we just declare another node master and that is
    it. It is important for the current change because master score might
    change after initial cluster start up causing master migration from
    one node to another.

    Partial-Bug: #1561894
    Partial-Bug: #1559136
    Change-Id: I2ee629eb07a6145647211da4dbe186aa9c45f61d