1
0
mirror of https://github.com/ansible/awx.git synced 2024-11-01 16:51:11 +03:00
awx/docs/resource_copy.md
Ryan Petrello ff1e8cc356
replace celery task decorators with a kombu-based publisher
this commit implements the bulk of `awx-manage run_dispatcher`, a new
command that binds to RabbitMQ via kombu and balances messages across
a pool of workers that are similar to celeryd workers in spirit.
Specifically, this includes:

- a new decorator, `awx.main.dispatch.task`, which can be used to
  decorate functions or classes so that they can be designated as
  "Tasks"
- support for fanout/broadcast tasks (at this point in time, only
  `conf.Setting` memcached flushes use this functionality)
- support for job reaping
- support for success/failure hooks for job runs (i.e.,
  `handle_work_success` and `handle_work_error`)
- support for auto scaling worker pool that scale processes up and down
  on demand
- minimal support for RPC, such as status checks and pool recycle/reload
2018-10-11 10:53:30 -04:00

9.6 KiB

Starting from Tower 3.3 and API v2, user are able to copy some existing resource objects to quickly create new resource objects via POSTing to corresponding /copy/ endpoint. A new CopyAPIView class is introduced as the base view class for /copy/ endpoints. It mimics the process of manually fetching fields from the existing object to create a new object, plus the ability to automatically detect sub structures of existing objects and make a background task-based deep copy when necessary.

Usage

If an AWX resource is copiable, all of its object detail API views will have a related URL field "copy", which has form /api/<version>/<resource name>/<object pk>/copy/. GET to this endpoint will return can_copy, which is a boolean indicating whether the current user can execute a copy operation; POST to this endpoint actually copies the resource object. One field name is required which will later be used as the name of the created copy. Upon success, 201 will be returned, along with the created copy.

For some resources like credential, the copy process is not time-consuming, thus the entire copy process will take place in the request-response cycle, and the created object copy is returned as POST response.

For some other resources like inventory, the copy process can take longer, depending on the number of sub-objects to copy (will explain later). Thus, although the created copy will be returned, the copy process is not finished yet. All sub-objects (like all hosts and groups of an inventory) will not be created until after the background copy task is finished in success.

Currently the available list of copiable resources are:

  • job templates
  • projects
  • inventories
  • workflows
  • credentials
  • notifications
  • inventory scripts

For most of the resources above, only the object to be copied itself will be copied; For some resources like inventories, however, sub resources belonging to the resource will also be copied to maintain the full functionality of the copied new resource. In specific:

  • When an inventory is copied, all its hosts, groups and inventory sources are copied.
  • When a workflow job template is copied, all its workflow job template nodes are copied.

How to add a copy end-point for a resource

The copy behavior of different resources largely follow the same pattern, therefore a unified way of enabling copy capability for resources is available for developers:

Firstly, create a /copy/ url endpoint for the target resource.

Secondly, create a view class as handler to /copy/ endpoint. This view class should be subclassed from awx.api.generics.CopyAPIView. Here is an example:

class JobTemplateCopy(CopyAPIView):

    model = JobTemplate
    copy_return_serializer_class = JobTemplateSerializer

Note the above example declares a custom class attribute copy_return_serializer_class. This attribute is used by CopyAPIView to render the created copy in POST response, so in most cases the value should be the same as serializer_class of corresponding resource detail view, like here the value is the serializer_class of JobTemplateDetail.

Thirdly, for the underlying model of the resource, Add 2 macros, FIELDS_TO_PRESERVE_AT_COPY and FIELDS_TO_DISCARD_AT_COPY, as needed. Here is an example:

class JobTemplate(UnifiedJobTemplate, JobOptions, SurveyJobTemplateMixin, ResourceMixin):
    '''
    A job template is a reusable job definition for applying a project (with
    playbook) to an inventory source with a given credential.
    '''
    FIELDS_TO_PRESERVE_AT_COPY = [
        'labels', 'instance_groups', 'credentials', 'survey_spec'
    ]
    FIELDS_TO_DISCARD_AT_COPY = ['vault_credential', 'credential']

When copying a resource object, basically all fields necessary for creating a new resource (fields composing a valid POST body for creating new resources) are extracted from the original object and used to create the copy.

However, sometimes we need more fields to be copied, like credentials of a job template, which cannot be provided during creation. In this case we list such fields in FIELDS_TO_PRESERVE_AT_COPY so that these fields won't be missed.

On the other hand, sometimes we do not want to include some fields provided in create POST body, like vault_credential and credential fields used for creating a job template, which do not have tangible field correspondence in JobTemplate model. In this case we list such fields in FIELDS_TO_DISCARD_AT_COPY so that those fields won't be included.

For models that will be part of a deep copy, like hosts and workflow job template nodes, the related POST body for creating a new object is not available. Therefore all necessary fields for creating a new resource should also be included in FIELDS_TO_PRESERVE_AT_COPY.

Lastly, unit test copy behavior of the new endpoint in /awx/main/tests/functional/test_copy.py and update docs (like this doc).

Fields in FIELDS_TO_PRESERVE_AT_COPY must be solid model fields, while fields in FIELDS_TO_DISCARD_AT_COPY do not need to be. Note there are hidden fields not visible from model definition, namely reverse relationships and fields inherited from super classes or mix-ins. A help script tools/scripts/list_fields.py is available to inspect a model and list details of all its available fields.

# In shell_plus
>>> from list_fields import pretty_print_model_fields
>>> pretty_print_model_fields(JobTemplate)

CopyAPIView will automatically detect sub objects of an object, and do a deep copy of all sub objects as a background task. There are sometimes permission issues with sub object copy. For example, when copying nodes of a workflow job template, there are cases where the user performing copy has no use permission of related credential and inventory of some nodes, and it is desired those fields will be None. In order to do that, developer should provide a static method deep_copy_permission_check_func under corresponding specific copy view. Like

class WorkflowJobTemplateCopy(WorkflowsEnforcementMixin, CopyAPIView):

    model = WorkflowJobTemplate
    copy_return_serializer_class = WorkflowJobTemplateSerializer

    # Other code

    @staticmethod
    def deep_copy_permission_check_func(user, new_objs):
        # method body

    # Other code

Static method deep_copy_permission_check_func must have and only have two arguments: user, the user performing the copy; new_objs, a list of all sub objects of the created copy. Sub objects in new_objs are initially populated disregarding any permission constraints, developer shall check user's permission against these new sub objects and react like unlink related objects or sending warning logs. deep_copy_permission_check_func should not return anything.

Lastly, macro REENCRYPTION_BLACKLIST_AT_COPY is available as part of a model definition. It is a list of field names which will escape re-encryption during copy. For example, extra_data field of workflow job template nodes.

Acceptance Criteria

  • Credentials should be able to copy themselves. The behavior of copying credential A shall be exactly the same as creating a credential B with all needed fields for creation coming from credential A.
  • Inventories should be able to copy themselves. The behavior of copying inventory A shall be exactly the same as creating an inventory B with all needed fields for creation coming from inventory A. Other than that, inventory B should inherit A's instance_groups, and have exactly the same host and group structures as A.
  • Inventory scripts should be able to copy themselves. The behavior of copying inventory script A shall be exactly the same as creating an inventory script B with all needed fields for creation coming from inventory script A.
  • Job templates should be able to copy themselves. The behavior of copying job template A shall be exactly the same as creating a job template B with all needed fields for creation coming from job template A. Other than that, job template B should inherit A's labels, instance_groups, credentials and survey_spec.
  • Notification templates should be able to copy themselves. The behavior of copying notification template A shall be exactly the same as creating a notification template B with all needed fields for creation coming from notification template A.
  • Projects should be able to copy themselves. The behavior of copying project A shall be the same as creating a project B with all needed fields for creation coming from project A, except for local_path, which will be populated by triggered project update. Other than that, project B should inherit A's labels, instance_groups and credentials.
  • Workflow Job templates should be able to copy themselves. The behavior of copying workflow job template A shall be exactly the same as creating a workflow job template B with all needed fields for creation coming from workflow job template A. Other than that, workflow job template B should inherit A's labels, instance_groups, credentials and survey_spec, and have exactly the same workflow job template node structure as A.
  • In all copy processes, name field of the created copy of the original object should be able to customize in the POST body.
  • The permission for a user to make a copy for an existing resource object should be the same as the permission for a user to create a brand new resource object using fields from the existing object.
  • The RBAC behavior of original workflow job template /copy/ should be pertained. That is, if the user has no necessary permission to the related project and credential of a workflow job template node, the copied workflow job template node should have those fields empty.