RE: [tosca] Artifacts questions

Hi Luca,

I agree that this is an area where more work is needed on the spec. Specifically:

1. The operation definition (A.5.12) only talks about implementation artifacts, not about deployment artifacts. I’m sure this is not intentional, but this should get corrected in the text nevertheless.

2. We need more “normative” artifact type definitions. For example:

a. We should have types for Chef Recipes, Puppet manifests, etc.

b. We also need types for VM and Docker images.

c. We should have mechanisms to specify whether artifacts need to be unzipped, untarred, etc.

3. For implementation artifacts specifically, we need more clarity to specify where these artifacts need to be executed. If my understanding is correct, the general assumption is that implementation artfacts are scripts that are run on the Compute node that is the “container” for the HostedOn relationship for software components. This raises the following questions:

a. What if the implementation artifact isn’t a script that can be executed directly (e.g. what if the implementation artifact is a Chef recipe)?

b. What if the “container” for the HostedOn relationship isn’t a Compute node (e.g. the Software Component is a Database that is “hosted on” a DBMS)?

c. What is the context within which implementation artifacts are executed (e.g. working directory, user id, etc?)

Feel free to help drive the process by contributing your specific use cases and for offering suggestions on how to best support your use cases.

Thanks,

Chris

From: tosca@lists.oasis-open.org [mailto:tosca@lists.oasis-open.org] On Behalf Of Luca Gioppo
Sent: Friday, June 05, 2015 12:57 AM
To: TOSCA
Subject: [tosca] Artifacts questions

On C.6.3.2 is stated that:

"TOSCA orchestrators expect node templates to provide either a deployment artifact or an implementation artifact of a defined artifact type that it is able to process."

In the examples it seems that there is a mix of deployment concepts and "implementation" concepts:

- if I understood well the deployment concept reading few lines below "... when encountering a deployment artifact on the create operation; will automatically attempt to deploy the artifact based upon its artifact type."

This is, I hope, clear enough orchestrator find deployment artifact of type zip and copy? unzip it? in any case does something with it (as you can see there is no clear and unique behaviour even here)

But then I find this example: B.8.1.3.3 Example: Retrieving artifact in a specified location

and I start to get confused. Who copy the implementation script "wordpress_install.sh" and where?

where is written that the the zip artifact is a deployment and if it is why do we need a create implementation to move it?

All this part is really difficult to understand also there is no simple example of the two type of artifacts and how they have to be used.

There should not be any interpretation of how to manage a type X artifact the operation linked to it should "copy" "unzip" "copy and unzip" but at the moment I have not found.

Luca

tosca message