Pulp

# Pulp 3.0.0 Minimum Viable Product (MVP)

<span style="color:red;">Lines highlighted in red need more attention.</span>

## Overall Guarantees

  - This is not a direct replacement for Pulp 2.

  - This is the minimum required for a 3.0.0 beta and GA.

  - All REST API calls will update the DB using transactions as necessary to ensure data integrity.

## Authentication

As an authenticated user I can manage user(s).

  - Add a user

  - View user(s)

  - Update any user detail

  - Delete a user

As an API user, I can have documentation to generate a JSON Web Token (JWT) without the server being online.

As an administrator, I can disable JWT token expiration. This configuration is in the settings file and is system-wide.

As an administrator, I can configure the JWT tokens to expire after a configurable amount of time. This configuration is in the settings file and is system-wide.

The JWT shall have a username identifier

<span style="color:red;">As an API user, I can authenticate any API call (except to request a JWT) with a JWT. (not certain if this should be the behavior)</span>

As an API user, I can invalidate all existing JWT tokens for a given user.

As an authenticated user, when deleting a user 'foo', all of user 'foo's existing JWTs are invalidated.

As an autheticated user, I can invalidate a user's JWTs in the same operation as updating the password.

As an un-authenticated user, I can obtain a JWT token by using a username and password.

## Repositories

As an authenticated user, I can list all repos.

  - All fields are included

  - Pagination is supported

  - <span style="color:red;">Filtering support</span>

As an authenticated user, I can CRUD a repository

  - Create a repo

  - Read a repo

  - Update all mutable repo fields

  - Delete a repo (asynchronous)

As an authenticated user, I can list a repository's associated importers and publishers

  - All fields are included

  - Pagination is supported

  - <span style="color:red;">Filtering support</span>

<span style="color:red;">As an authenticated user, I can summarize content in a repo (including counts)</span>

## Importers

note: Importer attributes will commonly be available on importers, but aren't guaranteed to be used by all importers.

As an authenticated user, I can CRUD an importer

  - Create an importer

  - Read an importer

  - Update all mutable importer fields

  - Delete an importer (asynchronous)

As an authenticated user I can configure the following attributes on an Importer:

  - validate (bool) \[optional: defaults to True\]: If true, the plugin will validate imported content.

  - ssl_ca_certificate (str) \[optional\] String containing a PEM encoded CA certificate used to validate the server certificate presented by the external source.

  - ssl_client_certificate (str) \[optional\] Contains a PEM encoded client certificate used for authentication.

  - ssl_client_key (str) \[optional\] Contains a PEM encoded private key used for authentication.

  - ssl_validation (bool) \[optional: defaults to True\]: If true, SSL peer validation must be performed.

  - proxy_url (str) \[optional\] Contains the proxy URL. Format: scheme://user:password@host:port.

  - username (str) \[optional\] Contains the username to be used for authentication when syncing.

  - password (str) \[optional\] Contains the password to be used for authentication when syncing.

  - download_policy (str) \[optional\] Contains the downloading policy name. This is a choice of three options:  

    \- immediate (default) - Downloading occurs during sync. The sync task does not complete until downloading is done.  

    \- background - Downloading is started by the sync but occurs in the background. The sync task completes before downloading is complete.  

    \- on-demand - The sync task records everything that would be downloaded but does not download content. Downloading occurs on demand as driven by client requests for content.

  - feed_url (str) \[optional\] Contains the URL of an external content source. This is optional.

  - sync mode: (str) \[optional\]. It has two choices 'additive' and 'mirror':  

    \- additive (default) - all remote content is added to the local repository on sync. During sync no content is removed from the local repository.  

    \- mirror - the local content will mirror the remote content exactly, removing local content if not also present in the remote content.

  - name (str) \[required\] Contains the name.

  - last_updated (datetime) \[read-only\] Contains the datetime of the last importer update.

  - last_synced (datetime) \[read-only\] Contains the datetime of the last importer sync.

  - <span style="color:red;">max_count (int) - ???</span>

## Publishers

note: Publisher attributes will commonly be available on publishers, but aren't guaranteed to be used by all publishers.

As an authenticated user, I can CRUD a publisher

  - Create a publisher

  - Read a publisher

  - Update all mutable publisher fields

  - Delete a publisher (asynchronous)

As an authenticated user I can configure the following attributes on a Publisher:

  - relative_path (str) \[optional\] The (relative) path component of the published url.

  - name - (str) \[required\] contains the name.

  - last_published (datetime) \[read-only\] When the last successful publish occurred.

  - last_updated (datetime) \[read-only\] The datetime of the last publisher update.

  - <span style="color:red;">auto_publish(bool) - ??? consider adding auto-publish feature to MVP</span>

## Content Manipulation

As an authenticated user, I can trigger an importer to sync.

  - I can follow the progress of all syncs. (Syncs are asynchronous.)

  - I cannot pass "sync" options.

  - Auto-publish is not included as an importer property.

As an authenticated user, I can trigger a publisher to publish.

  - I can follow the progress of all publishes. (Publishes are asynchronous.)

  - I cannot pass "publish" options.

## Upload & Copy

#### Getting bits from the client to Pulp

As an authenticated user, I can request a file ID from the server to upload a file with

As an authenticated user, I can upload a file with the server provided file ID, an optional chunk size, and an optional offset.

As an authenticated user, I can rely on Pulp to auto-delete uploaded files after a configurable time. (Eg: 6 hours).

As a user, I can delete an uploaded file by file ID

#### Creating Artifacts and Content Units

As a user, I can reference a file ID at Artifact creation time.

As an authenticated user, I can create a content unit by providing the content type, its Artifacts using file upload IDs for each Artifact, and the metadata supplied in the POST body. This call is atomic, either all Artifacts and the content unit are created in the database and on the filesystem or none are.

As an authenticated user, I can reuse a file ID to create multiple Artifacts without uploading the file twice.

#### Unit Management / Copy

As an authenticated user, I can add and remove one or more units to and from a destination repo.

  - <span style="color:red;">Filtering support for specifying the unit(s)</span>

  - I can follow the progress. (adding and removing are asynchronous).

## Versioned Repositories

As an authenticated user, I can list the content in a particular repository version

  - All fields are included

  - Pagination is supported

  - <span style="color:red;">Filtering support</span>

As an authenticated user, I can discover a URL to the latest version of a repository  

<span class="resource repository the on attributes or endpoint, API dedicated a through \^ Is" style="color:red;"></span>

As an authenticated user, I can run a publisher without a repository version and have it default to the latest version.

<span style="color:red;">As an authenticated user, I can delete a repository version by specifying the version</span>

<span style="color:red;">As an authenticated user, I can upload multiple content(s?) and add create a single new version that adds all of them.</span>

## Orphans

<span style="color:red;">As an authenticated user, I can clean up orphaned content units</span>  

<span style="color:red;">\* I can follow the progress of all cleanups. (Cleanups are asynchronous.)</span>

<span style="color:red;">As an authenticated user, I can delete a specific content unit</span>  

<span style="color:red;">\* If the content unit is still in at least one repository the delete fails with a listing of all repositories the unit is part of.</span>  

<span style="color:red;">\* Artifacts and associated files from the deleted unit are cleaned up</span>

<span style="color:red;">As an authenticated user, I can delete multiple content units with filtering</span>  

<span style="color:red;">\* If a content unit is still in at least one repository the delete fails with a listing of all repositories the unit is part of.</span>  

<span style="color:red;">\* Artifacts and associated files from deleted units are cleaned up</span>

<span style="color:red;">As an authenticated user, I see all (orphans) units that are not in any repositories</span>

## Filter

<span style="color:red;">I can filter all nouns *(What is the meaning of "filter?" What is a noun?)*</span>

## Task Management

As an authenticated user, I can list all tasks

  - Filtering support on \['state', 'id', 'group'\]

  - This does not include associated progress reports

As an authenticated user, I can see a detail view for a specific task

  - all attributes of a task

  - all associated progress reports

As an authenticated user, I can cancel a task

  - don't dare to use the DELETE verb!

## Task Group

I can view a summary of the status of all tasks in a group

## Event Listener Notifier

I can receive serialized task info via AMQP on each task save

*Can this be restated in more pedantic terms? Does this mean that an arbitrary host can attach itself to Pulp's AMQP message bus and get updates on the progress of tasks?*

## Status

As an unauthenticated user I can view the status of Pulp workers, resource managers, and celerybeats.

As an unauthenticated user I can view the status of httpd's connection to the database and message broker.

<span style="color:red;">As an administrator, the WSGI app will not start if all migrations have not been applied</span>

## Plugin API

As a plugin writer, I have a plugin API that is semantically versioned at 0.x separate from the REST API

As a plugin writer, I can report progress with a message and state

As a plugin writer, I can report progress with an optional suffix

As a plugin writer, I can report progress with a total count of things to do an the current count of things done

As a plugin writer, non-fatal exceptions a on the Task and are included in the Task detail. non_fatal exceptions do not cause the Task to be marked as failed, but may be interpreted by the user as not fully successful.

As a plugin writer, the working directory is set before Task work is done and cleaned up afterwards. I should not need to interact with the file system outside of the working dir.

As a plugin writer, I can provide a subclassed Importer. The importer's responsibility is to synchronize the content of a Pulp repository with the content of a remote repository.

As a plugin writer, I can provide a subclassed Publisher. The publisher's responsibility is to publish content.

As a plugin writer, I can define unit types by subclassing Content models to provide concrete content unit types to be manged by the platform.

As a plugin writer, I can interact with and create Artifacts

As a plugin writer, my app will be discovered by Pulp's app via an entry point provided by the plugin writer

As a plugin writer, I can use the plugin API to query content units/artifacts associated with a repository.

As a plugin writer, I can add and remove content units to and from a repository.

## CLI

<span style="color:red;">We will port what is there with as little effort as possible *(Does this mean that porting will be easy for developers, or that switching from the Pulp 2-3 CLI will be easy for users? If the former, isn't this an implementation detail that doesn't belong in an MVP document? If the latter, does this mean that we're going to carry forward the issues with pulp-admin, like a lack of status codes?)*</span>

<span style="color:red;">repo CRUD</span>  

<span style="color:red;">CRUD for importers</span>  

<span style="color:red;">CRUD for publishers</span>  

<span style="color:red;">trigger syncs</span>  

<span style="color:red;">trigger publish</span>  

<span style="color:red;">list content in a repo</span>  

<span style="color:red;">upload</span>  

<span style="color:red;">server status</span>  

<span style="color:red;">list and cancel tasks</span>  

<span style="color:red;">authn via basic auth</span>  

<span style="color:red;">\_(Should the supported set of operations be stated in terms of "The capabilities listed in the 'Authenctication,' 'Repositories,' and 'Filter' sections will be supported by the CLI."?)\_</span>

## Download API

As a plugin writer, I can download files via

  - http://

  - https://

  - file://

As a plugin writer, I can configure a downloader with:

  - Basic Auth

  - SSL Cert Client Auth

  - Custom CAs will be configured via a "trust store" either on the system or similar. Pulp will not do anything to read/load/manage CAs directly.

As a plugin writer, I can provide arbitrary behaviors for customized downloaders

  - For example token authentication in the docker plugin

As a plugin writer, I can have connection pooling/reuse

As a plugin writer, I have proxy settings

  - proxy url (containing basic auth info)

As a plugin writer, I can have great logs

As a user, I have documentation about how to use something for bandwidth limiting

As a plugin writer, I can configure the validation mechanisms used at download time

  - checksum validation - minimum (md5, sha1, sha256, sha512)

  - size validation

<span style="color:red;">As a plugin writer, I expect units that are missing from the remote repository to not be created in Pulp when using the immediate download policy.</span>

<span style="color:red;">As a plugin writer, I expect units that are missing from the remote repository to be created in Pulp when using background or on_demand download policies.</span>

As a plugin writer I can configure mirror lists and rotate between the mirrors

  - round robin

  - nearest mirror support

As a plugin writer, the plugin API provides tooling whereby I can provide the content to be added and removed from the repository. This tooling supports both immediate and deferred downloading.

As a plugin writer I can manage the catalog by using ChangeSets

As a plugin writer, the plugin can participate in adding content for cases where the decision to add additional content is based content that has been downloaded.

As a plugin writer, I can fetch content myself (but I am not encouraged to do so) with code I write

As a plugin writer, I can CRUD content units

## {color:red} Consumer Applicability

<span style="color:red;">Using consumer profiles and repo bindings I can compute applicability with 2.y parity  

Performance needs to be awesome</span>

<span style="color:red;">\_(Is the Pulp Consumer going away in Pulp 3? If so, is this section still appropriate?)\_</span>

## Plugin compatibility

rpm will work with platform  

puppet will work with platform  

ostree will work with platform  

python will work with platform  

file_plugin will work with platform  

docker will work with platform

## Migrations

users can run an executable similar to pulp-manage-db that is not named pulp-manage-db *(Why the change in name?)*

<span style="color:red;">What about migrating fields that we don't use in 3.0 but will use in 3.1+. For example the auto-publish feature?</span>

## Glossary

Repository - A named collection of content.

Artifact - A file associated with one content (unit). Artifacts are not shared between content (units). Create a content unit using an uploaded file ID as the source for its metadata. Create Artifacts associated with the content unit using an uploaded file ID for each; commit as a single transaction.

Content (unit) - A single piece of content manged by Pulp. Each file associated with a content (unit) is called an Artifact. Each content (unit) may have zero or many Artifacts.