Open Stack

Pardon the top-post here but what I'm seeing in this exchange is a
significant disconnect and a great example of why we're continuing to
struggle to define what an application is in the context of OpenStack.
I heard Jimmy asking for a PaaS, and Michael responding with how to
use an OpenStack environment to do some PaaS-like things.

OpenStack at it's heart is an infrastructure as a service - a way to
programatically provision resources (computers, network, storage).
Over time we've added lots of other projects on top that provide more
complicated services (load balancers, database services, etc.)  We've
also added some projects that provide a PaaS-like interface on top of
that infrastructure (murano, solum, maybe others?  Now with kubernetes
and some of the other container orchestration systems the lines are
getting more blurry.)

To bring it back to what I think Jimmy was talking about though - the
things he'd like to achieve (scaling, automatically attaching shared
storage, etc) are all things you get out of the box with most PaaS
approaches (openshift, cloud foundry, etc.)  For developers,
abstracted access to those constructs plus an ability to package and
deploy (to share) your application are the things that matter.

OpenStack doesn't provide those things in a consistent or
interoperable way.  If you're building your own cloud, you can
definitely choose which projects to bake in and maintain in order to
do this, but you can't count on that way being the same from cloud to
cloud.  I think Murano is about the only OpenStack-specific way to
handle these concerns *IF* we agree that this is what we mean when we
say "Application".  If there are other definitions of an application
we should focus on them and categorize them - I think it's the only
way we are going to get on the same page and start really reaching the
larger developer audience.

-Christopher





On Mon, Jun 27, 2016 at 7:47 AM, Michael Krotscheck
<krotscheck at gmail.com> wrote:
> Hello there, Jimmy-
>
> It really sounds like what you are looking for is a traditional support
> channel. i.e. an entity whom you are 'paying' to advocate to a community on
> your behalf, often with developers. Since you're currently running on
> Rackspace, you are covered under the CLA under which those resources have
> been donated to the Foundation. I recommend you reach out to Thierry to get
> the contact information for your account representative. I'm certain that
> Rackspace would be more than happy to assist you in scaling your
> applications.
>
> That's the official "dude, we're not a support channel" response. Here's the
> "I'm also an engineer" response that actually solves the problem you've
> listed:
>
> Firstly, simulated SAN failover feature (multiattach volumes) has been a
> coordinated discussion between Nova, Neutron, and Cinder for several cycles
> now, and it's been the subject of a whole 45 minute session at the Austin
> summit. Work is currently ongoing, but there are too many moving parts that
> need to land by feature freeze - in short, it's unlikely to land in Newton.
> From personal experience, Ocata's also a tossup; and that assumes the
> feature lands bug free with no performance implications.
>
> With that in mind, I personally would advise you against relying on a
> cloud-specific feature like this, because it's actually a limiting factor -
> you'd be restricted to a single cloud/region/zone for your DB clustering,
> plus it's a version and vendor lock-in; you'd only be able to replicate in
> on one zone on one particular OpenStack version.
>
> There's plenty of alternatives. Distributed filesystems are one
> (HDFS/Gluster), or the lower-level Distributed Replicated Block Device
> (drbd). Also, most databases have their own replication logic included, and
> while most of them were not built with cloud operations in mind, they
> usually don't require much tweaking to make them perform reliably.
>
> I hope this was helpful :)
>
> Michael Krotscheck
>
>
> On Thu, Jun 23, 2016 at 12:16 PM Jimmy Mcarthur <jimmy at tipit.net> wrote:
>>
>>
>>
>> Michael Krotscheck wrote:
>>
>> On Wed, Jun 22, 2016 at 11:50 AM Jimmy Mcarthur <jimmy at tipit.net> wrote:
>>>
>>> An example: you can't currently assign block storage to more than one VM
>>> at a time. This is something that I think is just sitting around as a patch
>>> to be approve in Neutron, but it's causing major problems for us as web
>>> application developers that are deploying on top of OpenStack. Basically, as
>>> a result of this and the lack of replication in Trove, we can't cluster.
>>>
>>>
>>> It's remarkably difficult to get integrated in IRC channels without
>>> knowing the lingo. Is there some suggestion from the user committee about
>>> where users like us could turn?
>>
>>
>> To address this specific issue: It sounds like you want to land a specific
>> feature in Neutron. The correct place to advocate for this is the weekly
>> neutron meeting. As someone who's recently landed a cross-project feature
>> (in 23 different projects), I can confidently say that every team is open to
>> - if occasionally grumpy about - unscheduled features that aren't on their
>> roadmap. It took me only a few questions, and quite a bit of humility, to be
>> given a primer on each teams' approval governance, approval process, and
>> roadmap feature selection.
>>
>> Maybe I wasn't clear about my role in OpenStack :) I'm not an OpenStack
>> developer. I'm a web and mobile application developer (more appropriately, a
>> project manager) that hosts our sites on OpenStack public cloud. I don't
>> have a patch to land in Neutron. I understand that it was already done and
>> is waiting for approval by that team.
>>
>>
>> OpenStack's governance empowers those who are willing to advocate for
>> themselves, as long as they are willing to back their requests with actual
>> code. I'm sure that Neutron would be very happy to address and shepherd any
>> patches you'd be willing to provide.
>>
>> Keep in mind that there is no place that I can currently advocate for my
>> team, which is why I'm raising the point :) I work for the Foundation
>> building web and mobile applications, but rely on OpenStack for
>> infrastructure. Specifically, we're running on the Rackspace cloud in the
>> same data center as Infra. The features I mention aren't within our skill
>> set to develop, but they're critical if OpenStack is to become a viable
>> option on which to host scalable web applications that need to share
>> data/resources.  Though I'm sure many could do it very ably, I don't expect
>> OpenStack developers to come and write PHP or javascript in order to use our
>> website. We're valid users of the software you all are doing such a great
>> job of building.
>>
>>
>> In regards to understanding the IRC 'lingo', I don't really know what that
>> could refer to. Could you clarify?
>>
>> Like any software product, there is common nomenclature that defines it.
>> Even reading the documentation can't possibly catch you up on the history of
>> the project and the people, especially since so much of it takes place in
>> IRC. If you're not out to become a full time OpenStack developer and simply
>> need something to work in a particular way, trying to integrate with that
>> project can be pretty tough.
>>
>> I certainly don't mean to start a great debate, but I would encourage you
>> to think of app developers that don't use OpenStack SDKs as well as those
>> that do. If we're not providing a place for those users to deliver feedback
>> and communicate, we could be missing out on lots of opportunities to study
>> how they are using the software. Companies (both large and small) don't
>> always have the resources to contribute back to OpenStack anymore than every
>> user of Ubuntu can contribute upstream.  There is a whole world of
>> application developers out there that have no need/ability to be involved at
>> that level.
>>
>> Cheers!
>> Jimmy
>>
>>
>> Michael Krotscheck
>>
>>
>
> _______________________________________________
> User-committee mailing list
> User-committee at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/user-committee
>