5.0.x environments should not deploy secondary controllers in parallel

[Vladimir Kuklin]

In 5.1 version we introduced parallel deployment of secondary controllers, but this cannot work with 5.0.x version.

[Vladimir Kuklin (vkuklin)]

This looks pretty much like versioning of serializers

[Mike Scherbakov (mihgen)]

We should consider logic for patching one by one, not only controllers, but also all other nodes.
Rollback is the same.

Actually what would happen if we try to do rollback on node which is in original state (no patching was performed)? It may happen in case if you do patching one by one, then click Stop button, so half nodes was patched and half was not. What if we now run patching again - or start rollback?

[Ihor Kalnytskyi (ikalnytskyi)]

Vladimir, could you please explain why it won't work for 5.0.x? From my experience, I can say that it works and OSTF tests are passes.

Mike, the puppet operation is idempotent. So if we apply rollback on the node which is already in original state and the config is the same, the puppet just ensures that all settings are applied and do nothing.

BTW, Mike, we don't have "Stop" button on UI for patching procedure, so you has to wait for either fail or success.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/118655

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/118655
Committed: https://git.openstack.org/cgit/stackforge/fuel-web/commit/?id=b36dccbc28584532ca1bdb679b096cc3f5f6f80a
Submitter: Jenkins
Branch: master

commit b36dccbc28584532ca1bdb679b096cc3f5f6f80a
Author: Igor Kalnitsky <email address hidden>
Date: Wed Sep 3 15:53:06 2014 +0300

    Use different serializers for different envs

    We should deploy secondary controllers using sequental logic
    for 5.0.x clusters.

    The commit implements a very basic approach for versioning serializers,
    so it may be revised in the future.

    Change-Id: Ia94bd16c3060dff76c2c79dca480784092331deb
    Closes-Bug: #1364519