openstack-ansible/doc/source/install-guide/configure-federation-mapping.rst

Home OpenStack-Ansible Installation Guide

Configure Identity Service (keystone) Domain-Project-Group-Role mappings

The following is an example service provider (SP) mapping configuration for an ADFS identity provider (IdP):

federated_identities:
  - domain: Default
    project: fedproject
    group: fedgroup
    role: _member_

Each IdP trusted by an SP must have the following configuration:

1. project: The project which federated users will have access to. If the project does not already exist then it is created in the domain with the name specified by domain.
2. group: The Identity (keystone) group to which the federated users will belong. If the group does not already exist then it is created in the domain with the name specified by domain.
3. role: The role which federated users will assume in that project. If the role does not already exist, it is created.
4. domain: The domain in which the project lives, and in which the role is assigned. If the domain does not already exist, it will be created.

With the above information, Ansible implements the equivalent of the following OpenStack CLI commands:

# if the domain does not already exist
openstack domain create Default

# if the group does not already exist
openstack group create fedgroup --domain Default

# if the role does not already exist
openstack role create _member_

# if the project does not already exist
openstack project create --domain Default fedproject

# map the role to the project and user group in the domain
openstack role add --project fedproject --group fedgroup _member_

If the deployer wants to add more mappings, additional options can be added to the list, for example:

federated_identities:
  - domain: Default
    project: fedproject
    group: fedgroup
    role: _member_
  - domain: Default
    project: fedproject2
    group: fedgroup2
    role: _member_

Identity Service federation attribute mapping

Attribute mapping adds a set of rules to map federation attributes to keystone users and/or groups. An IdP has exactly one mapping specified per protocol.

Mapping objects can be used multiple times by different combinations of IdP and protocol.

The details of how the mapping engine works, the schema and various rule examples are in the keystone developer documentation.

Consider an example SP attribute mapping configuration for an ADFS IdP:

mapping:
  name: adfs-IdP-mapping
  rules:
    - remote:
        - type: upn
      local:
        - group:
            name: fedgroup
            domain:
              name: Default
        - user:
            name: '{0}'
attributes:
  - name: 'http://schemas.xmlsoap.org/ws/2005/05/identity/claims/upn'
    id: upn

Each IdP for an SP needs to be set up with a mapping. This tells the SP how to interpret the attributes provided to the SP from the IdP.

In this particular case the IdP is publishing the upn attribute. As this is not in the standard Shibboleth attribute attribute map (see /etc/shibboleth/attribute-map.xml in the keystone containers), this IdP has been configured with the extra mapping through the attributes dictionary.

The mapping dictionary is a yaml representation very similar to the keystone mapping property which Ansible uploads. The above mapping produces the following in keystone.

root@aio1_keystone_container-783aa4c0:~# openstack mapping list
+------------------+
| ID               |
+------------------+
| adfs-IdP-mapping |
+------------------+

root@aio1_keystone_container-783aa4c0:~# openstack mapping show adfs-IdP-mapping
+-------+---------------------------------------------------------------------------------------------------------------------------------------+
| Field | Value                                                                                                                                 |
+-------+---------------------------------------------------------------------------------------------------------------------------------------+
| id    | adfs-IdP-mapping                                                                                                                      |
| rules | [{"remote": [{"type": "upn"}], "local": [{"group": {"domain": {"name": "Default"}, "name": "fedgroup"}}, {"user": {"name": "{0}"}}]}] |
+-------+---------------------------------------------------------------------------------------------------------------------------------------+

root@aio1_keystone_container-783aa4c0:~# openstack mapping show adfs-IdP-mapping | awk -F\| '/rules/ {print $3}' | python -mjson.tool
[
    {
        "remote": [
            {
                "type": "upn"
            }
        ],
        "local": [
            {
                "group": {
                    "domain": {
                        "name": "Default"
                    },
                    "name": "fedgroup"
                }
            },
            {
                "user": {
                    "name": "{0}"
                }
            }
        ]
    }
]

The interpretation of the above mapping rule is that any federated user authenticated by the IdP is mapped to an ephemeral (non-existant) user in keystone. The user is a member of a group named fedgroup, which in turn is in a domain called Default. The user's ID and Name (federation always uses the same value for both properties) for all OpenStack services will be the value of upn.

---