Comment 0 for bug 1476957

Steps to reproduce:

1/ .. create and configure a new cluster via Fuel UI (roles assignement, network config, plugins settings ..) BUT not "deploy changes"
2/ # fuel --env 1 deployment default
3/ .. update IP addresses for some/all nodes (sections nodes and endpoints) ..
4/ # fuel --env 1 deployment upload
5/ # fuel deploy-changes (or via Fuel UI)

Expected results:
After the OSt deployment the "hiera('nodes')" from puppet manifests returns correct information (with ip addresses configured)

Actual results:
after the OSt deployment the "hiera('nodes')" from puppet manifests returns outdated data (with ip addresses by default).

Impacts:
All Fuel plugins using hiera('nodes') will fail to configure its stuff with bad IP addresses.
Since Fuel allows to update IP addresses before deployment, (potentially) users will do it. Fuel must update the same way it does for astute.yaml the nodes.yaml with the modified data.

Environement:
Fuel 6.1 / Ubuntu (surely the same result with CentOS)
Plugins LMA collector 0.7.2

Other informations / pre diagnostic :
* OpenStack and others services (rabbitmq, haproxy ..) are correctly configured with good IPs (those updated before deployment), fortunately because the /etc/hiera/nodes.yaml is generated at post_deployment stage.
* the hiera hierarchy cannot be updated (ie move down 'nodes' after 'astute') because this will break a whole lot of stuff

Workaround:
on the plugin manifests side, instead using "hiera('nodes')", use the old school way 'parseyaml($astute_settings_yaml)'
but this workaround is not accesptable.