You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ariatosca.apache.org by "Ran Ziv (JIRA)" <ji...@apache.org> on 2017/08/07 21:00:00 UTC

[jira] [Commented] (ARIA-334) Policy for operation execution configuration

    [ https://issues.apache.org/jira/browse/ARIA-334?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16117254#comment-16117254 ] 

Ran Ziv commented on ARIA-334:
------------------------------

Sounds interesting. Is there a reason for having the policy type as {{aria.Operation}} and then have a single {{configuration}} property though, rather than having the policy type be {{aria.OperationConfig}} and then have the properties themselves be the configuration?

> Policy for operation execution configuration
> --------------------------------------------
>
>                 Key: ARIA-334
>                 URL: https://issues.apache.org/jira/browse/ARIA-334
>             Project: AriaTosca
>          Issue Type: Story
>            Reporter: Tal Liron
>            Assignee: Tal Liron
>
> Our support for [using dependencies to configure operation execution|https://cwiki.apache.org/confluence/display/ARIATOSCA/Execution+Configuration] solves the technical problem, but it is very cumbersome to use in large templates.
> One solution is to use YAML macros (in the {{dsl_definitions}} section of the TOSCA template). But this is still awkward and ugly to have to insert the macro for every operation.
> The TOSCA way to do this is by policy. I suggest we support a new policy for operation execution, while *also* lettings user override those values locally per operation.
> Here's how it can be defined in the ARIA profile:
> {code}
> policy_types:
>   aria.Operation:
>     derived_from: tosca.policies.Root
>     targets: [ tosca.nodes.Root, tosca.groups.Root ]
>     properties:
>       configuration:
>         type: map
>         entry_schema: string
> {code}
> You could then use it like so:
> {code}
> topology_template:
>   policies:
>     ssh:
>       type: aria.Operation
>       targets: [ bono, sprout, ralf, homer, homestead, vellum ]
>       properties:
>         configuration:
>           ssh.user: { get_property: [ HOST, host, ssh.user ] }
>           ssh.password: { get_property: [ HOST, host, ssh.password ] }
>           ssh.address: { get_attribute: [ HOST, public_address ] }
> {code}
> What the above means is that for *all* operations on *all* interfaces of the target nodes (or groups) the above configuration parameters would be automatically applied (though local use of {{dependencies}} could override these values).
> The nice thing about this is that 1) it just needs to be defined once, and 2) it avoids our hacky {{dependencies}} notation. 
> It might also make sense to add more properties to the policy for applying the policy only to specific interfaces or even specific operations and the target nodes. But that could possibly wait for a followup JIRA task.
> Note that the above example would also require us to fix intrinsic functions. As it stands, the parser would fail the above, because using the {{HOST}} keyword can only be used for properties inside a node or relationship. We would thus also need to allow for policies to use {{HOST}}, and assume it applies to the target node.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)