RoleValidator doesn't take into account volumes metadata from plugins

[Sergey Kreys]

"fuel role --update" is failed if role metadata contains custom volume defined by some fuel plugin:

# fuel role --rel 2 --role aic-swift-storage --update --file ./aic-swift-storage.yaml
400 Client Error: Bad Request for url: http://1.0.35.2:8000/api/v1/releases/2/roles/aic-swift-storage/ (Wrong data in volumes_roles_mapping. Volumes with ids [u'swift-storage'] are not in the list of allowed volumes [u'os', u'logs', u'mongo', u'ceph', u'cephjournal', u'mysql', u'vm', u'image', u'horizon', u'cinder', u'cinder-block-device'])

Fuel plugin can provide custom volume metadata. Refer to method get_volumes_metadata in nailgun/objects/cluster.py - as an example how to get these volumes.

Steps to reproduce:
1) Install fuel plugin that provides custom volume (new volume with id 'swift-storage' - in our case);
2) assume that at a post-install step roles that use that custom volume from plugin are created; *
3) Update plugin to a newer version;
4) assume that after plugin update roles are going to be updated from updated yaml files. *

*Steps 2,4 can actually be done manually, without losing generality.

Expected results:
Roles are updated according to a metadata provided with updated fuel plugin.

Actual result:
"fuel role --update" is failed for roles with custom volumes in volumes_roles_mapping.

Reproducibility: always

Workaround: no

Impact: User is not able to update roles with new metadata.
Thus, if new version of fuel plugin contains change for allocate_size in volumes_roles_mapping, it will not be applied to existing environment.

[Vitalii Kulanov (vitaliy-t)]

Can you, please, specify fuel version?

[Sergey Kreys (skreys)]

[root@nailgun ~]# fuel -v
9.0.0

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/mitaka
Review: https://review.openstack.org/381324

[Evgeniy L (rustyrobot)]

`fuel role --update` command tries to update `releases` model, plugin roles are attached to cluster, not release.

There are multiple ways to deal with it:

1. The easiest way is to update plugin roles using plugin mechanism, which is a. change plugin, b. update it.

2. Do refactoring of the way we store roles, and create a full copy of roles metadata into cluster related model, after that reimplement roles update handler and cli, to be able to update roles for cluster, not for release, using this mechanism we will be able to accumulate per-cluster roles.

[Vladimir Kuklin (vkuklin)]

The issue is that you are using release roles instead of plugin-provided roles. You need to refactor the plugin to use plugin-provided roles.

[OpenStack Infra (hudson-openstack)]

Change abandoned by Vladimir Kuklin (<email address hidden>) on branch: stable/mitaka
Review: https://review.openstack.org/381324