[rhos-list] Optimizing IT: Red Hat Virtual Event

Daniel Dumitriu daniel at dumdan.com
Thu Dec 6 05:26:51 UTC 2012


Hey, Angus,

It's really good to see somebody with a "Heat bias". The heavier, the
better!

Your last point is really interesting, because it expresses what I meant
by "Heat managing/orchestrating" the aggregation of profiles...

I wouldn't call the generic Heat target "an application" but, rather, a
"Heat system" or "Heat environment" or... any other name for
"Instantiated Infrastructure Architecture Patterns".

In any case, if I understood you correctly, Heat is, already, headed in
that direction.

(Note to self: Must look into Heat...)

I'm sure I shouldn't be speaking for Ohad, but I worked a bit with
Foreman (struggled would be more accurate, partly due to my
stubbornness) and I could, at least, say that no, Foreman is, definitely
not, focused on "bare metal".
At least in the most recent "incarnations" Foreman's goal is to manage
complex Puppet-based configurations. The bare-metal is there, but pretty
well defined (read: almost a plugin) and used as a means to the goal. I
do like it because of that, and the fact that it has a rather uniform
approach to provisioning. 
I ran into Foreman because of my attempts to parametrize and script
RHOS: I needed an API-enabled, maybe script-able, solution to deploy the
hardware OpenStack components. It looked like Foreman could, eventually,
fit such a bill.

There's one other thing that haunts me: too many people are reluctant to
take my (or anyone else's) word that virtual environments will provide
the same performance or security as the bare-metal. So, over time, I
created my own definition for "hybrid cloud": a cloud comprised of both
virtual and physical machines.
This is completely different from the "mixed" or "heterogeneous" cloud,
which spans private and public cloud deployments.

As for Image Factory - I know nothing of it and should, maybe, look
"into it".
However, whatever framework is used to create the images, we may reach a
point where we actually want to provide a dedicated, kvm-capable,
machine(s?) that can be shared between tenants, with some kind of
stateless image generating sessions. More or less, what Perry was
suggesting (but without specific scheduling).

Ok, enough, for now.
(And apologies to everyone of a different persuasion - in the last year
or so, *for me*, if it's virtual, it has to be KVM)


Daniel


-- 
___________
Daniel Dumitriu
daniel at dumdan.com
Telephone: +1-416-626-9345
Mobile:    +1-416-318-2487


-----Original Message-----
From: Angus Salkeld <asalkeld at redhat.com>
To: Daniel Dumitriu <daniel at dumdan.com>
Cc: Perry Myers <pmyers at redhat.com>, Heron, Rodrique (CTR) WDC31
<Rodrique.Heron at cigna.com>, rhos-list at redhat.com <rhos-list at redhat.com>,
Ohad Levy <ohadlevy at redhat.com>, Ian McLeod <imcleod at redhat.com>, Steve
Loranz <sloranz at redhat.com>, John R. Dunning <jrd at redhat.com>, Steven
Hardy <shardy at redhat.com>, Zane Bitter <zbitter at redhat.com>
Subject: Re: [rhos-list] Optimizing IT: Red Hat Virtual Event
Date: Thu, 6 Dec 2012 10:47:52 +1100

On 05/12/12 17:24 -0500, Daniel Dumitriu wrote:
>Hi,... everyone!
>
>Well, all this sounds very interesting...
>
>Because I am, actually, using Oz to create my images for Nova... Of
>course, that doesn't say much because it's, still, a manual process with
>no integration.
>
>Perry, this is how things may happen, eventually - and I could be wrong
>(I'm known to do that, a lot):
>
>- somebody (Oz ?) creates images for VMs AND kickstarts for physicals;
>- somebody (Foreman ?) provisions physicals;
>- somebody (Foreman ? Heat ?) manages server profiles (templates ?);
>- somebody (Nova with Glance) provisions VMs;
>- somebody (Heat?) manages/orchestrates the aggregation of
>  profiles (templates) and pushes them as schedules to Nova or
>  Foreman...
>
>Except... there's a lot of overlapping, which it is understandable,
>given the origins of each project.


Yea, there is a heap of overlap. All are trying to focus on something
a bit different.
- I believe Foreman is focused on baremetal (I am not up-to-date on the project tho')
- ImageFactory and Aeolus is focused on being cloud neutral
- Heat is focused on tight integration with OpenStack.

I have a bias (Heat developer:)
- With heat you _could_ (if you used bleading edge OpenStack) use OpenStack baremetal
   feature to deploy directly to baremetal (not vm's). (Some of our users are trying
   to get this to work).
- You definitely can use Puppet with CloudFormation templates:
https://s3.amazonaws.com/cloudformation-templates-us-east-1/puppet-master-configuration.template
https://s3.amazonaws.com/cloudformation-templates-us-east-1/wordpress-via-puppet-client.template
https://s3.amazonaws.com/cloudformation-templates-us-east-1/wordpress-via-puppet-client-building-block.template

- One thing to add to this is Heat will do more than just provision
   the vm. You can setup a true Virtual application with networking,
   objectstore, volume, loadbalancer, autoscaling (soon DNSaas) resources.
   Current resources:
   https://github.com/openstack/heat/tree/master/heat/engine/resources

-Angus

>
>By the way, I keep insisting on Foreman because (despite the trouble I
>have with it) it holds the promise of a "real cloud": one that's not
>exclusively virtualized.
>Also it looks like it is the only one that's focused on managing Puppet
>configurations (though Heat might, as well).
>Data center operations it would be benefiting a lot from the
>streamlining of such an approach (IMHO).
>
>(And I am sooo sorry I cannot dedicate more time and effort to pushing
>this stuff forward - my day job, you know...)
>
>Thanks,
>
>Daniel
>
>-- 
>___________
>Daniel Dumitriu
>daniel at dumdan.com
>Telephone: +1-416-626-9345
>Mobile:    +1-416-318-2487
>
>
>-----Original Message-----
>From: Perry Myers <pmyers at redhat.com>
>To: Heron, Rodrique (CTR) WDC31 <Rodrique.Heron at cigna.com>
>Cc: daniel at dumdan.com <daniel at dumdan.com>, rhos-list at redhat.com
><rhos-list at redhat.com>, Ohad Levy <ohadlevy at redhat.com>, Ian McLeod
><imcleod at redhat.com>, Steve Loranz <sloranz at redhat.com>, John R. Dunning
><jrd at redhat.com>, Angus Salkeld <asalkeld at redhat.com>, Steven Hardy
><shardy at redhat.com>, Zane Bitter <zbitter at redhat.com>
>Subject: Re: [rhos-list] Optimizing IT: Red Hat Virtual Event
>Date: Wed, 05 Dec 2012 15:28:16 -0500
>
>Hi Rodrique and Daniel,
>
>On 12/05/2012 02:08 PM, Heron, Rodrique (CTR) WDC31 wrote:
>> I missed the event.
>
>It's actually just recorded presentations, and they are available for 90
>days I believe.  So you may still be able to register for the event and
>listen to the presentations for a while.
>
>> I am using Foreman, and it would be nice to use Foreman to build images and push them to glance.
>
>Right, so today Foreman just takes images that already exist and what
>you're looking for is a feature enhancement to Foreman that would
>integrate with something like Image Factory or Oz to build new VM images
>from source media and then push those images into Glance.  Do I have
>that right?
>
>> -
>> Rodrique Heron
>>
>> -----Original Message-----
>> From: rhos-list-bounces at redhat.com [mailto:rhos-list-bounces at redhat.com] On Behalf Of Daniel Dumitriu
>> Sent: Wednesday, December 05, 2012 1:54 PM
>> To: Perry Myers
>> Cc: rhos-list at redhat.com
>> Subject: Re: [rhos-list] Optimizing IT: Red Hat Virtual Event
>>
>> This is just a repeat of the comment/question I left in "The Lounge" of the Event:
>
>Apologies for not being in the lounge today to take the questions live.
> I had another commitment and couldn't attend.
>
>> ==
>> Thanks, Perry, for the OpenStack presentation.
>> Inciting, to say the least.
>> Question: At various points, you did mention "Heat" as an upstream
>> project. Could you be more specific on the provisioning side of
>> things ?
>
>It's a service similar to Cloud Formations in Amazon AWS.  It provides
>the ability to define a template that describes multiple VMs working in
>concert.  So you would define a template with 1 or more VM definitions,
>and then tell Heat to launch that template.  Heat then uses the
>OpenStack Nova APIs to launch the VMs in the proper order, waiting for
>things like dependencies in one VM before launching the 2nd.
>
>I've included some of the Heat folks on this thread.  I'm sure they can
>do better justice to your question than I could :)
>
>> It would be interesting, for example, to see what could happen if
>> "Foreman" would be used to "provision the Cloud". Especially for a
>> "private hybrid cloud" that includes both VMs and hardware servers.
>
>Foreman could definitely be used for both bare metal server provisioning
>for OpenStack core services and Compute Nodes while also being used to
>provision the VMs that will be in the cloud.  I thought that Ohad
>(Foreman creator/maintainer) had done some integration work (or maybe it
>was just planned) to better integrate Foreman w/ OpenStack from a guest
>provisioning perspective (using Fog as an abstraction layer I think)
>
>Ohad, any additional thoughts here?
>
>Perry
>




More information about the rhos-list mailing list