acceptance docs for [AUDIC] Arbitrary User-Defined Inventory & Credential

This commit is contained in:
Ryan Petrello 2017-04-24 17:03:49 -04:00
parent 3243110496
commit 4a438e044e

View File

@ -0,0 +1,316 @@
Custom Credential Types Overview
================================
Prior to Tower 3.2, Tower included bundled credential types, such as
"Machine", "Network", or "Amazon Web Services". In 3.2, we have added support
for custom types so that customers can extend Tower with support for
third-party credential mechanisms.
Important Changes
-----------------
* Tower has a new top-level resource, ``Credential Type``, which can fall into
one of several categories, or "kinds":
- SSH
- Vault
- Source Control
- Network
- Cloud
``Credential Types`` are composed of a set of field ``inputs`` (for example,
``"username"`` - which is a required string - and ``"password"`` - which is
a required string which should be encrypted at storage time) and custom
``injectors`` which define how the inputs are applied to the environment when
a job is run (for example, the value for ``"username"`` should be injected
into an environment variable named ``"MY_USERNAME"``).
By utilizing these custom ``Credential Types``, customers have the ability to
define custom "Cloud" and "Network" ``Credential Types`` which
modify environment variables, extra vars, and generate file-based
credentials (such as file-based certificates or .ini files) at
`ansible-playbook` runtime.
* Multiple ``Credentials`` can now be assigned to a ``Job Template`` as long as
the ``Credential Types`` are unique. For example, you can now create a ``Job
Template`` that uses one SSH, one Vault, one EC2, and one Google Compute
Engine credential. You cannot, however, create a ``Job Template`` that uses
two OpenStack credentials.
* Custom inventory sources can now utilize a ``Credential``; you
can store third-party credentials encrypted within Tower and use their
values from within your custom inventory script (by - for example - reading
an environment variable or a file's contents).
API Interaction for Credential Management
-----------------------------------------
``HTTP GET /api/v2/credential_types`` provides a listing of all supported
``Credential Types``, including several read-only types that Tower provides
support for out of the box (SSH, Vault, SCM, Network, Amazon Web Services,
etc...)
Superusers have the ability to extend Tower by creating, updating, and deleting
new "custom" ``Credential Types``:
HTTP POST /api/v2/credential_types/
{
"name": "Third Party Cloud",
"description": "Integration with Third Party Cloud",
"kind": "cloud",
"inputs": {
"fields": [{
"id": "api_token",
"label": "API Token",
"type": "string",
"secret": True,
}]
},
"injectors": {
"env": {
"THIRD_PARTY_CLOUD_API_TOKEN": "{{api_token}}"
}
}
}
In Tower 3.2, when users create or modify ``Credentials``, they specify the
``credential_type``, and the inputs they pass in are dictated by the
defined ``inputs`` for that ``Credential Type``:
HTTP POST /api/v2/credentials/
{
"name": "Joe's Third Party Cloud API Token",
"description": "",
"organization": <pk>,
"user": null,
"team": null,
"credential_type": <pk>,
"inputs": {
"api_token": "f239248b-97d0-431b-ae2f-091d80c3452e"
}
}
HTTP GET /api/v2/credentials/N
{
"name": "Joe's Third Party Cloud API Token",
"description": "",
"organization": <pk>,
"user": null,
"team": null,
"credential_type": <pk>,
"inputs": {
"api_token": "$encrypted$"
}
}
Defining Custom Credential Type Inputs
--------------------------------------
A ``Credential Type`` specifies an ``inputs`` schema which defines a set of
ordered fields for that type:
"inputs": {
"fields": [{
"id": "api_token", # required - a unique name used to
# reference the field value
"label": "API Token", # required - a unique label for the
# field
"help_text": "User-facing short text describing the field.",
"type": ("string" | "number" | "ssh_private_key") # required,
"secret": true, # if true, the field will be treated
# as sensitive and stored encrypted
"multiline": false # if true, the field should be rendered
# as multi-line for input entry
},{
# field 2...
},{
# field 3...
}]
"required": ["api_token"] # optional; one or more fields can be marked as required
},
As an alternative to static types, fields can also specify multiple choice
strings:
"inputs": {
"fields": [{
"id": "api_token", # required - a unique name used to reference the field value
"label": "API Token", # required - a unique label for the field
"choices": ["A", "B", "C"]
}]
},
Defining Custom Credential Type Injectors
-----------------------------------------
A ``Credential Type`` can inject ``Credential`` values through the use
of the Jinja templating language (which should be familiar to users of Ansible):
"injectors": {
"env": {
"THIRD_PARTY_CLOUD_API_TOKEN": "{{api_token}}"
},
"extra_vars": {
"some_extra_var": "{{username}}:{{password}"
}
}
``Credential Types`` can also generate temporary files to support .ini files or
certificate/key data:
"injectors": {
"file": {
"template": "[mycloud]\ntoken={{api_token}}"
},
"env": {
"MY_CLOUD_INI_FILE": "{{tower.filename}"
}
}
Job and Job Template Credential Assignment
------------------------------------------
In Tower 3.2, ``Jobs`` and ``Job Templates`` have a new many-to-many
relationship with ``Credential`` that allows selection of multiple
network/cloud credentials. As such, the ``Job`` and ``JobTemplate``
API resources in `/api/v2/` now have two credential related fields:
HTTP GET /api/v2/job_templates/N/
{
...
'credential': <integer-or-null>
'vault_credential': <integer-or-null>
...
}
...and a new endpoint for fetching all "extra" credentials:
HTTP GET /api/v2/job_templates/N/extra_credentials/
{
'count': N,
'results': [{
'name': 'My Credential',
'credential_type': <pk>,
'inputs': {...},
...
}]
}
Similar to other list attachment/detachment API views, cloud and network
credentials can be created and attached via an `HTTP POST` at this new
endpoint:
HTTP POST /api/v2/job_templates/N/extra_credentials/
{
'id': <cloud_credential_primary_key>,
'associate': True,
}
HTTP POST /api/v2/job_templates/N/extra_credentials/
{
'id': <network_credential_primary_key>,
'disassociate': True,
}
HTTP POST /api/v2/job_templates/N/extra_credentials/
{
'name': 'My Credential',
'credential_type': <primary_key>,
'inputs': {...},
...
}
API Backwards Compatability
---------------------------
`/api/v1/credentials/` still exists in Tower 3.2, and it transparently works as
before with minimal surprises by attempting to translate `/api/v1/` requests to
the new ``Credential`` and ``Credential Type`` models.
* When creating or modifying a ``Job Template`` through `v1` of the API,
old-style credential assignment will transparently map to the new model. For
example, the following `POST`'ed payload:
{
credential: <pk>,
vault_credential: <pk>,
cloud_credential: <pk>,
network_credential: <pk>,
}
...would transparently update ``JobTemplate.extra_credentials`` to a list
containing both the cloud and network ``Credentials``.
Similarly, an `HTTP GET /api/v1/job_credentials/N/` will populate
`cloud_credential`, and `network_credential` with the *most recently applied*
matching credential in the list.
* Custom ``Credentials`` will not be returned in the ``v1`` API; if a user
defines their own ``Credential Type``, its credentials won't show up in the
``v1`` API.
* ``HTTP POST`` requests to ``/api/v1/credentials/`` will transparently map
old-style attributes (i.e., ``username``, ``password``, ``ssh_key_data``) to
the appropriate new-style model. Similarly, ``HTTP GET
/api/v1/credentials/N/`` requests will continue to contain old-style
key-value mappings in their payloads.
* Vault credentials are a new first-level type of credential in Tower 3.2.
As such, any ``Credentials`` pre-Tower 3.2 that contain *both* SSH and Vault
parameters will be migrated to separate distinct ``Credentials``
post-migration.
For example, if your Tower 3.1 installation has one ``Credential`` with
a defined ``username``, ``password``, and ``vault_password``, after migration
*two* ``Credentials`` will exist (one which contains the ``username`` and
``password``, and another which contains only the ``vault_password``).
Additional Criteria
-------------------
* Rackspace is being removed from official support in Tower 3.2. Pre-existing
Rackspace Cloud credentials should be automatically migrated to "custom"
credentials. If a customer has never created or used Rackspace Cloud
credentials, the only change they should notice in Tower 3.2 is that
Rackspace is no longer an option provided by Tower when creating/modifying
a Credential.
Acceptance Criteria
-------------------
When verifying acceptance we should ensure the following statements are true:
* `Credential` injection for playbook runs, SCM updates, inventory updates, and
ad-hoc runs should continue to function as they did prior to Tower 3.2 for the
`Credential Types` provided by Tower.
* It should be possible to create and modify every type of `Credential` supported
prior to Tower 3.2 (SSH, SCM, EC2, etc..., with the exception of Rackspace).
* Superusers (and only superusers) should be able to define custom `Credential
Types`. They should properly inject environment variables, extra vars, and
files for playbook runs, SCM updates, inventory updates, and ad-hoc runs.
* The default `Credential Types` included with Tower in 3.2 should be
non-editable/readonly and cannot be deleted by any user.
* Stored `Credential` values for _all_ types should be consistent before and
after Tower 3.2 migration/upgrade.
* `Job Templates` should be able to specify multiple extra `Credentials` as
defined in the constraints in this document.
* Custom inventory sources should be able to specify a cloud/network
`Credential` and they should properly update the environment (environment
variables, extra vars, written files) when an inventory source update runs.
* If a `Credential Type` is being used by one or more `Credentials`, the fields
defined in its ``inputs`` should be read-only.
* `Credential Types` should support activity stream history for basic object
modification.