xenapi: windows agent unreliable due to reboots

[John Garbutt]

The windows nova-agent now can trigger a gust reboot during resetnetwork, so the hostname is correctly updated.

Also there was always a reboot during the first stages of polling for the agent version that can cause the need to wait for a call to timeout, rather than detecting a reboot.

Either way, we need to take more care to detect reboots while talking to the agent.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/122100
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=d65ca49990f1241a5a8afdc3ad0e1a99c878275a
Submitter: Jenkins
Branch: master

commit d65ca49990f1241a5a8afdc3ad0e1a99c878275a
Author: John Garbutt <email address hidden>
Date: Mon Jun 30 14:09:24 2014 +0100

    xenapi: deal with reboots while talking to agent

    The latest version of the agent has started changing the hostname
    through the official windows APIs, when resetnetwork is called. This
    means a reboot might happen after resetnetwork.

    A reboot happens when polling for the version. Until now we wait for the
    call timeout, before fetching the new dom id. This change ensures that
    if we spot a reboot, the plugin exits early rather than keeping polling
    the wrong dom id.

    Turns out its best to wait for the dom_id to change, before trying
    to poll the agent again. Once the dom_id in xenapi has been updated,
    the xenstore keys are always in place.

    Trying too early leads to lots of reboot detections because we are
    retrying with the old dom_id. XenServer continues to return the old
    dom_id for a little while after the reboot.

    Closes-Bug: #1370999

    Change-Id: Id0bf5b64f2b271d162db5bbce50167ab1f665c87