Comment 2 for bug 1606827

Reviewed: https://review.openstack.org/347708
Committed: https://git.openstack.org/cgit/openstack/neutron/commit/?id=bb989be99db84a2789abe2849c786a075e3f5ab7
Submitter: Jenkins
Branch: master

commit bb989be99db84a2789abe2849c786a075e3f5ab7
Author: John Schwarz <email address hidden>
Date: Wed Jul 27 12:09:30 2016 +0300

    Don't use exponential back-off for report_state

    If an agent tries to report_state to the neutron-server and it fails
    because of a timeout (raising oslo_messaging.MessagingTimeout), then
    there is an exponential back-off effect, which causes the
    seemingly-simple report_state RPC call to take 60 seconds, then 120,
    then 240 and so on. This can happen if all the controllers are
    restarted simultaneously a number of time, as the bug report describes.

    Since the feature was intended for heavy RPC calls (like get_routers())
    and not for light calls such as report_state, it's safe to reduce the
    timeout to a constant 60 seconds interval.

    Closes-Bug: #1606827
    Change-Id: I15aeea9f8265b859bb1a8ee933b8b2ce1e64b695