[one-users] Survey - About Feature #189

[Tino Vazquez]

Hi Ettore,

We agree with you that this functionality is desirable. On the other
hand, we were always extremely concern about the separation between
providing infrastructure for running VMs and the needs of the services
running inside the VM. The ability to reference other machines in one
VM template is one example of a feature that kind of lies in the
frontier between the two.

Here are our reasons to support the dropping of the cross reference
feature (we will keep the ability to reference variables within the
same template):

* Cleaner and more robust code. Dropping this feature will reduce
considerably the probabilities of a deadlock between OpenNebula
components.

* We have in our roadmap the definition and implementation of a
service layer [1] on top of OpenNebula that will handle services at an
upper level. A service will be able to be specified using a DAG like
notation, and it will include the ability to perform cross references
with another VMs.

Any feedback on this will be welcome.

Regards,

-Tino

--
Constantino Vázquez, Grid & Virtualization Technology
Engineer/Researcher: http://www.dsa-research.org/tinova
DSA Research Group: http://dsa-research.org
Globus GridWay Metascheduler: http://www.GridWay.org
OpenNebula Virtual Infrastructure Engine: http://www.OpenNebula.org



On Tue, Mar 2, 2010 at 3:12 PM, Ettore Enrico Delfino Ligorio
<ettoreusp at gmail.com> wrote:
> Hi Tino,
>
> I think that this is desirable feature because it allows a richer use
> of OpenNebula api by a application (webservice or website) deploying
> complex context systems. For me, complex contexts are services made by
> more than one virtual machine.
>
> For example, some service may deploy a system with: 02 webserver, 01
> database server and 01 ldap server.
>
> At some point, opennebula contextualization script of webserver may
> use a information from OpenNebula instance of the database server.
>
> Although, i think that the main requisite is cross reference to the VM
> own template.
>
> Best regards.
> --
> Ettore Enrico Delfino Ligorio
> +55-11-3091-6480
>
>
> On Mon, Mar 1, 2010 at 2:16 PM, Tino Vazquez <tinova79 at gmail.com> wrote:
>> Dear OpenNebula users,
>>
>> Feature #189 [1] is posing us a design question, and we would like to
>> ask the community about it.
>>
>> When resolving contextualization substitutions, first the template is
>> written in the database and then a parsing is applied to resolve the
>> $variables present in the contextualization section. This design stems
>> from the cross references that can be done with others VMs, i.e,
>>
>> master_public_ip=  "$83.$NIC[IP, NETWORK=\"Public\"]"
>>
>> will be getting the IP of network interface attached to network
>> "Public" of the machine with id 83.
>>
>> It is the case that, removing this feature (cross references to other
>> VMs in the contextualization section, not the references to the VM own
>> template), we can save one DB write to the database, and thus increase
>> the throughput of machines that can be sent to OpenNebula.
>>
>> We would like to survey members of the community to find users that
>> use this feature, and, if possible, to what extend and purpose.
>>
>> [1] http://dev.opennebula.org/issues/189
>>
>> --
>> Constantino Vázquez, Grid & Virtualization Technology
>> Engineer/Researcher: http://www.dsa-research.org/tinova
>> DSA Research Group: http://dsa-research.org
>> Globus GridWay Metascheduler: http://www.GridWay.org
>> OpenNebula Virtual Infrastructure Engine: http://www.OpenNebula.org
>> _______________________________________________
>> Users mailing list
>> Users at lists.opennebula.org
>> http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
>>
>