[openstack-dev] [Openstack-operators] [all] Consistent policy names

Lance Bragstad lbragstad at gmail.com
Fri Sep 21 14:13:02 UTC 2018


On Fri, Sep 21, 2018 at 2:10 AM Ghanshyam Mann <gmann at ghanshyammann.com>
wrote:

>  ---- On Thu, 20 Sep 2018 18:43:00 +0900 John Garbutt <
> john at johngarbutt.com> wrote ----
>  > tl;dr+1 consistent names
>  > I would make the names mirror the API... because the Operator setting
> them knows the API, not the codeIgnore the crazy names in Nova, I certainly
> hate them
>
> Big +1 on consistent naming  which will help operator as well as developer
> to maintain those.
>
>  >
>  > Lance Bragstad <lbragstad at gmail.com> wrote:
>  > > I'm curious if anyone has context on the "os-" part of the format?
>  >
>  > My memory of the Nova policy mess...* Nova's policy rules traditionally
> followed the patterns of the code
>  > ** Yes, horrible, but it happened.* The code used to have the OpenStack
> API and the EC2 API, hence the "os"* API used to expand with extensions, so
> the policy name is often based on extensions** note most of the extension
> code has now gone, including lots of related policies* Policy in code was
> focused on getting us to a place where we could rename policy** Whoop whoop
> by the way, it feels like we are really close to something sensible now!
>  > Lance Bragstad <lbragstad at gmail.com> wrote:
>  > Thoughts on using create, list, update, and delete as opposed to post,
> get, put, patch, and delete in the naming convention?
>  > I could go either way as I think about "list servers" in the API.But my
> preference is for the URL stub and POST, GET, etc.
>  >  On Sun, Sep 16, 2018 at 9:47 PM Lance Bragstad <lbragstad at gmail.com>
> wrote:If we consider dropping "os", should we entertain dropping "api",
> too? Do we have a good reason to keep "api"?I wouldn't be opposed to simple
> service types (e.g "compute" or "loadbalancer").
>  > +1The API is known as "compute" in api-ref, so the policy should be for
> "compute", etc.
>
> Agree on mapping the policy name with api-ref as much as possible. Other
> than policy name having 'os-', we have 'os-' in resource name also in nova
> API url like /os-agents, /os-aggregates etc (almost every resource except
> servers , flavors).  As we cannot get rid of those from API url, we need to
> keep the same in policy naming too? or we can have policy name like
> compute:agents:create/post but that mismatch from api-ref where agents
> resource url is os-agents.
>

Good question. I think this depends on how the service does policy
enforcement.

I know we did something like this in keystone, which required policy names
and method names to be the same:

  "identity:list_users": "..."

Because the initial implementation of policy enforcement used a decorator
like this:

  from keystone import controller

  @controller.protected
  def list_users(self):
      ...

Having the policy name the same as the method name made it easier for the
decorator implementation to resolve the policy needed to protect the API
because it just looked at the name of the wrapped method. The advantage was
that it was easy to implement new APIs because you only needed to add a
policy, implement the method, and make sure you decorate the implementation.

While this worked, we are moving away from it entirely. The decorator
implementation was ridiculously complicated. Only a handful of keystone
developers understood it. With the addition of system-scope, it would have
only become more convoluted. It also enables a much more copy-paste pattern
(e.g., so long as I wrap my method with this decorator implementation,
things should work right?). Instead, we're calling enforcement within the
controller implementation to ensure things are easier to understand. It
requires developers to be cognizant of how different token types affect the
resources within an API. That said, coupling the policy name to the method
name is no longer a requirement for keystone.

Hopefully, that helps explain why we needed them to match.


>
> Also we have action API (i know from nova not sure from other services)
> like POST /servers/{server_id}/action {addSecurityGroup} and their current
> policy name is all inconsistent.  few have policy name including their
> resource name like "os_compute_api:os-flavor-access:add_tenant_access", few
> has 'action' in policy name like
> "os_compute_api:os-admin-actions:reset_state" and few has direct action
> name like "os_compute_api:os-console-output"
>

Since the actions API relies on the request body and uses a single HTTP
method, does it make sense to have the HTTP method in the policy name? It
feels redundant, and we might be able to establish a convention that's more
meaningful for things like action APIs. It looks like cinder has a similar
pattern [0].

[0]
https://developer.openstack.org/api-ref/block-storage/v3/index.html#volume-actions-volumes-action


>
> May be we can make them consistent with
> <service-type>:<resource>:<action_with_snake_case> or any better opinion.
>
>  > From: Lance Bragstad <lbragstad at gmail.com>> The topic of having
> consistent policy names has popped up a few times this week.
>  >
>  > I would love to have this nailed down before we go through all the
> policy rules again. In my head I hope in Nova we can go through each policy
> rule and do the following:
>  > * move to new consistent policy name, deprecate existing name* hardcode
> scope check to project, system or user** (user, yes... keypairs, yuck, but
> its how they work)** deprecate in rule scope checks, which are largely
> bogus in Nova anyway* make read/write/admin distinction** therefore adding
> the "noop" role, amount other things
>
> + policy granularity.
>
> It is good idea to make the policy improvement all together and for all
> rules as you mentioned. But my worries is how much load it will be on
> operator side to migrate all policy rules at same time? What will be the
> deprecation period etc which i think we can discuss on proposed spec -
> https://review.openstack.org/#/c/547850


Yeah, that's another valid concern. I know at least one operator has
weighed in already. I'm curious if operators have specific input here.

It ultimately depends on if they override existing policies or not. If a
deployment doesn't have any overrides, it should be a relatively simple
change for operators to consume.


>
>
> -gmann
>
>  > Thanks,John
> __________________________________________________________________________
>  > OpenStack Development Mailing List (not for usage questions)
>  > Unsubscribe:
> OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
>  > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>  >
>
>
>
> __________________________________________________________________________
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20180921/98012c92/attachment.html>


More information about the OpenStack-dev mailing list