[one-users] [one-ecosystem] Custom monitoring attributes in VM TEMPLATE

[Simon Boulet]

Hi Ruben

I like the idea of placing the monitoring information under a MONITOR
section.

I'm looking at scaling OpenNebula and adding some caching, having the
TEMPLATE section to be static  where the information would change only ex.
when a VM resize operation is done (or at least, not being updated each
time the VM is monitored). One could identify each sections "lifetime" ex. :

USER_TEMPLATE = updated on vm.update or when a VM action fails (and the
ERROR attribute is updated)
TEMPLATE = updated on vm.resize, attach, detach, etc. operation
MONITOR = updated on each poll (typically LASTPOLL + MONITORING_INTERVAL in
oned.conf.. perhaps add a NEXTPOLL attribute to make caching lifetime
easier)
HISTORY_RECORDS = updated on VM state changes (one.vm.action) etc.

Then perhaps vm.vmpool could return only the base <VM> attributes (nothing
below the USER_TEMPLATE, TEMPLATE, MONITOR) and one would have to call
one.vm.info to retrieve the full VM XML.

I understand that would break backward compatibility. But I think it would
be a huge improvement to ease caching of the OpenNebula information.

> I'd say that comments to issues in redmine is a good place to link the
discussions to specific features, new things and general aspects should go
to the mailing list.

Users or ecosystem list?

Simon


On Thu, Apr 25, 2013 at 6:23 AM, Ruben S. Montero
<rsmontero at opennebula.org>wrote:

> Hi Simon
>
> The criteria is that every thing to be consumed by OpenNebula should be in
> the Template, so the user can not modify it. Otherwise it may lead to
> OpenNebula incorrectly manage the object (e.g. remove or modify an
> attribute from the DISK section) or a security threat (e.g. a user
> modifying the SOURCE attribute to access someone else's image ) That is the
> main reason to place the system generated information in the TEMPLATE
> section, including monitoring info (you may bypass a billing system based
> on a probe).
>
> You are right that ERROR should be moved to USER TEMPLATE. The user may
> want to delete the error once she is notified... I've opened an issue for
> this [1]
>
> Now the other probes. This is for historical reasons. First versions of
> OpenNebula 1.0 (4 years ago!) did not have the functionality that we have
> now, and that lead to some inconsistencies in the data model. Probably CPU,
> NETRX... and the other probe results should be placed probably under a
> MONITOR section. During the development of the project we've opted to keep
> backwards compatibility over a more consistent data model. Probably we are
> wrong, but trying to keep the ecosystem and other adaption working from one
> release to other has been a priority for us.
>
> You've raised other interesting issue. First, we are sorry if we failed to
> engage in any discussion about the implementation of new issues. This is by
> no means because of a lack of interest but for the trillions of things we
> have to deal with. Your feedback is very valuable for us and always
> considered internally even if we fail to answer it.
>
> I'd say that comments to issues in redmine is a good place to link the
> discussions to specific features, new things and general aspects should go
> to the mailing list.
>
> However, this is probably not as fluent and agile as it should. We've done
> in the past IRC meetings that we may restore plan a regular hangout for
> example. We'll work on this and send a proposal to the mailing list.
>
> Thank you very much for feedback!
>
> [1] http://dev.opennebula.org/issues/1955
>
>
>
>
> On Wed, Apr 24, 2013 at 4:24 PM, Simon Boulet <simon at nostalgeek.com>wrote:
>
>> Hi,
>>
>> Custom monitoring attributes (extra attributes returned by the IM MONITOR
>> or VMM POLL) are currently being added under the TEMPLATE section of the VM
>> definition (from vm/VirtualMachine.cc:3165)
>>
>> Perhaps I'm wrong, but I am seeing the TEMPLATE section a block that
>> containt the "static" definition of the VM (number of CPU, maximum memory,
>> attached NICs, Disks, etc.). Ex. one could cache the information found in
>> the TEMPLATE section on their frontend for a certain period of time. And it
>> starts to make even more sense now that we have a USER_TEMPLATE section
>> (since OpenNebula 4.0).
>>
>> Also errors are added under the TEMPLATE section.
>>
>> Wouldn't it make more sense if the custom monitoring attributes (and
>> errors) were being reported directly under the VM section (just like the
>> other monitoring attributes NETRX, NETTX, etc. are being reported).
>>
>> PS. Not sure if the ecosystem list is the right place for this. In the
>> past I used the users list as well as the bug tracker to submit technical
>> questions and discussions. And I find it difficult to be informed, sharing
>> the vision and debating some implementations of upcoming features in
>> OpenNebula.
>>
>> Simon
>>
>> _______________________________________________
>> Ecosystem mailing list
>> Ecosystem at lists.opennebula.org
>> http://lists.opennebula.org/listinfo.cgi/ecosystem-opennebula.org
>>
>>
>
>
> --
> Ruben S. Montero, PhD
> Project co-Lead and Chief Architect
> OpenNebula - The Open Source Solution for Data Center Virtualization
> www.OpenNebula.org | rsmontero at opennebula.org | @OpenNebula
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opennebula.org/pipermail/users-opennebula.org/attachments/20130425/c49510b6/attachment-0001.htm>