Why are you calling this attribute "optional"?

ARIA always exposes it, and also *should* fill in its value. This might
depend on the actual plugin doing the provisioning work... we would need to
verify this on a case by case basis. As to which value, again it depends on
the plugin.

Which IP address should it use? TOSCA doesn't say. I think everyone agrees
that networking in TOSCA is very poorly designed.

On Tue, Nov 14, 2017 at 4:41 PM, Steve Baillargeon <
steve.baillarg...@ericsson.com> wrote:

> As an example, does ARIA always expose the "optional" private_address for
> a Compute node instance?
> If so, what method is used by ARIA to select the primary OAM IP address
> when multiple IPs are associated with the Compute node instance?
>
> -Steve
>
> -----Original Message-----
> From: Tal Liron [mailto:t...@cloudify.co]
> Sent: Tuesday, November 14, 2017 5:27 PM
> To: dev@ariatosca.incubator.apache.org
> Subject: Re: Attributes
>
> ARIA exposes all defined attributes, even if they have no default value
> (in which case they will be null).
>
> Note that three attributes (tosca_id, tosca_name, state) are automatically
> filled by ARIA, if they are available. Any node type inheriting from
> tosca.nodes.Root will have them.
>
> I'm not sure what you mean by "optional" or "mandatory": there is no way
> to define attributes that way. See section 3.5.10.2 in the spec.
>
> On Tue, Nov 14, 2017 at 4:05 PM, Steve Baillargeon <
> steve.baillarg...@ericsson.com> wrote:
>
> > Hi
> > The YAML spec is confusing about when the service or node instance
> > actually exposes the value of an attribute.
> > Does ARIA always implement and expose the values of all attribute
> > definitions (optional and mandatory) defined in normative types?
> >
> > Related observation: it is very strange that the YAML spec does not
> > support a REQUIRED keyname for the attribute definition. Then how do
> > we go about declaring an optional vs mandatory attribute definition in
> > custom type definitions? For now, I am almost ready to conclude that
> > all attributes SHOULD be exposed by the TOSCA orchestrator regardless
> > if it is required or not.
> >
> > -Steve
> >
>

Reply via email to