Open Stack

On 26/11/13 03:26, Keith Bray wrote:
>
>
> On 11/25/13 5:46 PM, "Clint Byrum" <clint at fewbar.com> wrote:
>
>> Excerpts from Tim Schnell's message of 2013-11-25 14:51:39 -0800:
>>> Hi Steve,
>>>
>>> As one of the UI developers driving the requirements behind these new
>>> blueprints I wanted to take a moment to assure you and the rest of the
>>> Openstack community that the primary purpose of pushing these
>>> requirements
>>> out to the community is to help improve the User Experience for Heat for
>>> everyone. Every major UI feature that I have implemented for Heat has
>>> been
>>> included in Horizon, see the Heat Topology, and these requirements
>>> should
>>> improve the value of Heat, regardless of the UI.
>>>
>>>
>>> Stack/template metadata
>>> We have a fundamental need to have the ability to reference some
>>> additional metadata about a template that Heat does not care about.
>>> There
>>> are many possible use cases for this need but the primary point is that
>>> we
>>> need a place in the template where we can iterate on the schema of the
>>> metadata without going through a lengthy design review. As far as I
>>> know,
>>> we are the only team attempting to actually productize Heat at the
>>> moment
>>> and this means that we are encountering requirements and requests that
>>> do
>>> not affect Heat directly but simply require Heat to allow a little
>>> wiggle
>>> room to flesh out a great user experience.
>>>
>>
>> Wiggle room is indeed provided. But reviewers need to understand your
>> motivations, which is usually what blueprints are used for. If you're
>> getting push back, it is likely because your blueprints to not make the
>> use cases and long term vision obvious.
>
> Clint, can you be more specific on what is not clear about the use case?

The part where it's used for something. What I'm hearing in this thread 
is "We need this for a template catalog, but we're not going to talk 
about that until later. Trust us, though, we really need it."

That's a problem, because none of this stuff makes sense except in the 
context of the template catalog. What does the template catalog do? Who 
writes the templates in the catalog? Why is the template catalog not 
backed by git? Who can say which parts of the solution belong in the 
template catalog and which in Heat when we don't even know what it's a 
solution to?

> What I am seeing is that the use case of meta data is not what is being

I'm contesting it:

https://review.openstack.org/#/c/56703/7/doc/source/template_guide/hot_spec.rst

The problem here isn't that the process wasn't followed, it's that all 
of these things - one-shot outputs, operator-specific structured 
metadata sections in templates, APIs to report versions of randomly 
selected daemons - are fundamentally bad ideas. And they're all bad in 
exactly the same way: they demand that Heat implement something that it 
can only do in a half-baked manner, because it's slightly easier (or 
seems slightly easier) than implementing something in some external tool 
that can work.

It's difficult to escape the conclusion that these features have been 
decided upon behind closed doors with the community excluded, by people 
who are not especially familiar with Heat, with a view to minimising the 
effort required to implement some other (proprietary?) thing and not 
necessarily to coming up with the best implementation. If this is your 
development process then you are Doing Open Source Wrong. At a very 
minimum you need to work about 10x harder to explain all of these design 
decisions - in particular the reasons why the obvious alternatives that 
don't appear to have been considered are insufficient - to the 
community. It's much, much easier, though, to just do your design in the 
open with community involvement.

Take this metadata section idea. (Please!) Let's step back and examine 
the very last decision in the chain - assume for a moment that we accept 
that this metadata needs to be included in the template. There are three 
obvious ways to do that:

  1) Include them as YAML comments
  2) Include them in the body of the template
  3) Include them as a separate YAML document in the same file

How many of those were actually considered? My bet is only #2. It 
doesn't really matter though, because there is not one word of design 
documentation that would help me understand the reasons for rejecting #1 
& #3 (neither of which require any changes to Heat or the template 
format). How can I have confidence that #2 is the correct solution to 
the problem? Especially if I don't agree (I don't)?

The number of options not considered grows (and my confidence in the 
solution shrinks) exponentially as you step all the way back through the 
chain of undocumented decisions to the actual thing that drove this idea.

> contested, but that the Blueprint of where meta data should go is being
> contested by only a few (but not all) of the core devs.  The Blueprint for

I contested that first because without knowing what the feature is even 
for, there's nothing else to contest.

> in-template metadata was already approved for Icehouse, but now that work
> has been delivered on the implementation of that blueprint, the blueprint
> itself is being contested:
>     https://blueprints.launchpad.net/heat/+spec/namespace-stack-metadata

It seems clear that we have to communicate better to new contributors 
what the expectations are around getting features in. But I would be 
really sad if we have to jump straight in to the sort of legalistic 
processes that some of the larger OpenStack programs follow. This is 
still a relatively small project; we should be able to handle these 
questions by talking to each other.

> I'd like to propose that the blueprint that has been accepted go forth
> with the code that exactly implements it, and if there are alternative
> proposals and appropriate reasons for the community to come to consensus
> on a different approach, that we then iterate and move the data (deprecate
> the older feature if necessary, e.g. If that decision comes after
> Icehouse, else of a different/better implementation comes before Icehouse,
> then no harm done).

-1

>>
>>> There is precedence for an optional metadata section that can contain
>>> any

As Clint said, you're comparing APIs to templates. Those are very different.

>>> end-user data in other Openstack projects and it is necessary in order
>>> to
>>> iterate quickly and provide value to Heat.

Non sequitur.

>>
>> Nobody has said you can't have meta-data on stacks, which is what other
>> projects use.
>>
>>> There are many use cases that can be discussed here, but I wanted to
>>> reiterate an initial discussion point that, by definition,
>>> "stack/template_metadata" does not have any hard requirements in terms
>>> of
>>> schema or what does or does not belong in it.
>>>
>>> One of the initial use cases is to allow template authors to categorize
>>> the template as a specific "type".
>>>
>>>      template_metadata:
>>>          short_description: Wordpress

First of all, "short_description" is a dreadful name for something that 
is basically tags.

Secondly, these are tags in the template catalog. So why are they 
incorporated in the template, and not stored in the template catalog?

Finally, if you really want this it can be trivially implemented using 
comments.

>>
>> Interesting. Would you support adding a "category" keyword to python so
>> we don't have to put it in setup.cfg and so that the egg format doesn't
>> need that section? Pypi can just parse the python to categorize the apps
>> when they're uploaded. We could also have a file on disk for qcow2 images
>> that we upload to glance that will define the meta-data.
>>
>> To be more direct, I don't think the templates themselves are where this
>> meta-data belongs. A template is self-aware by definition, it doesn't
>> need the global metadata section to tell it that it is WordPress. For
>> anything else that needs to be globally referenced there are parameters.
>> Having less defined inside the template means that you get _more_ wiggle
>> room for your template repository.
>
> Clint, you are correct that the Template does not need to know what it is.
>   It's every other service (and users of those services) that a Template
> passes through or to that would care to know what it is. We are suggesting
> we put that meta data in the template file and expressly ignore it for
> purposes of parsing the template language in the Heat engine, so we agree
> it not a necessary part of the template.  Sure, we could encode the
> metadata info in a separate catalog...  but, take the template out of the
> catalog and now all that useful associated data is lost or would need to
> be recreated by someone or some service.  That does not make the template
> portable, and that is a key aspect of what we are trying to achieve (all
> user-facing clients, like Horizon, or humans reading the file, can take
> advantage). We don't entirely know yet what is most useful in portability

What's most useful in portability is placing the minimum requirements 
(especially undocumented and/or non-standard requirements) on the 
template author, since you can't control the template author.

> and what isn't, so meta data in-template provides the "wiggle room"
> innovation space to suss that out.  We already know of some specific use
> cases of data that we feel are important, which Tim identified one
> specific example.. As specific metadata items become popular or prove to
> be useful to rely on by the larger community or service operators (public
> and private) of Heat, we as a community can drive that information back
> into the schema for the template or some portable format mechanism.
>
>>
>> I 100% support having a template catalog. IMO it should be glance,
>> which is our catalog service in OpenStack. Who cares if nova or heat are
>> consuming images or templates. It is just sharable blobs of data and
>> meta-data in a highly scalable service. It already has the concept of
>> global and tenant-scope. It just needs an image type of 'hot' and then
>> heat can start consuming templates from glance. And the template authors
>> should maintain some packaging meta-data in glance to communicate to
>> users that this is "Wordpress" and "Single-Node". If Glance's meta-data
>> is too limiting, expand it! I'm sure image authors and consumers would
>> appreciate that.
>
> This is definitely interesting... And takes the long view IMO.  Let me
> explain:  I don't anticipate Heat catalog'ing in Glance is something that
> has a high chance of getting implemented in the Icehouse timeframe (at
> least, not more so than in-template metadata), do you?  From a SOA service
> deployer perspective, I'm sure you can appreciate that rolling out new
> functionality in Glance to support an Orchestration project use case is
> not simple, and requires strong business justification and coordination as
> an operator of a cloud.. One worth exploring for sure, but not the go-to
> default strategy. I view the metadata change as very minor with little to
> no disruption to any service, including Heat (Heat just ignores the

I view it as an open invitation to fragment the template format and the 
ecosystem.

> metadata, completely).. This fits very well in an iterative development
> model.  New blueprints could be raised, as you suggested, to move metadata
> and catalog features into Glance.  My concern is that if we go the Glance
> route now, we are encouraging a precedent that we aren't iterative (due to
> abandoning an already accepted blueprint in favor of a more complex and
> time intensive solution) and that we won't get this implemented within the
> current release cycle.
>
>>
>>> This would let the client of the Heat API group the templates by type
>>> which would create a better user experience when selecting or managing
>>> templates. The the end-user could select "Wordpress" and drill down
>>> further to select templates with different options, "single node", "2
>>> web
>>> nodes", etc...
>>>
>>
>> That is all api stuff, not language stuff.
>
> If this were done solely at the API, it would have to be maintained 1-to-1
> with a template (in which case there is an implicit and explicit
> association), and exported with the template in order to port the
> template.

I don't understand all this stuff about 'portability'. From your 
previous descriptions of the template catalog, it sounded like it's for 
templates hand-picked by the operator. Why would one operator want to 
import the tags defined by another operator anyway? And Why wouldn't 
that just be a git clone of the backing repository? And how does any of 
this make the template more "portable"???

>>
>>> Once a feature has consistently proven that it adds value to Heat or
>>> Horizon, then I would suggest that we can discuss the schema for that
>>> feature and codify it then.
>>>
>>> In order to keep the discussion simple, I am only responding to the need
>>> for stack/template metadata at the moment but I'm sure discussions on
>>> the
>>> management api and template catalog will follow.
>>>
>>
>> Your example puts the template catalog in front of this feature, and I
>> think that exposes this feature as misguided.
>
> I'd like template author and creation date portable so that it goes
> wherever the template goes in a portable format that consumers (any
> client) of the Heat service can understand.  I'd also like this

Put it in a comment.

> information to be available for a deployed stack in Heat, so it is need
> not be catalog specific.  If author/creation-date information were only

If it needs to be available for a deployed stack (you haven't made an 
argument for this yet) then add it into the template-proper. If Heat 
needs to read it then some sort of wild-west schemaless metadata area is 
definitely not the place to put it.

Alternately, maybe we need to look at storing templates in such a way 
that the comments are preserved. I would definitely look favourably on a 
blueprint to that effect.

> available in the catalog, then we would have to entirely wrap the Heat API
> in another service or risk drift in the data between Heat and the separate
> non-wrapping catalog service. Instead of raising a blueprint for every
> piece of data that may be useful to deploying this service (some of which

Please don't raise a blueprint for every patch. e.g. a feature and its 
documentation 
(https://blueprints.launchpad.net/heat/+spec/update-stack-metadata-spec) 
are not two blueprints.

> certain folks may not care about [e.g. Many of the folks arguing against
> metadata all together], the in-template metadata is, IMO, a suitable
> approach to experiment and then we can drive back experiential use-cases
> into future improvements (e.g. in-language schema changes or catalog
> changes).  At Rackspace we have experience running a nearly identical
> service, so we know this data, in this particular place, is a valid use
> case for consumers of an Orchestration service.
> Example:
> (1) Cloud customer of service operator deploys a template (may or may not
> have come from the catalog)

If it didn't come from the catalog provided by the operator, what are 
the chances that it contains this information anyway? How is the average 
template author going to know about the schema for the metadata read by 
your proprietary tool?

> (2) Customer encounters problem with the stack and calls support
> (3) Support specialist [not the service operator/developer] is trying to
> figure out who to contact for additional help given failures seen with the
> stack. The dev team of the service is not the expert in application foo
> and did not develop the template. It's not a Heat issue, it's a
> template/application foo issue.

This sounds like a good argument for Heat preserving comments.

> Having this detail come from Glance, as you pointed out, would be great.
> But, my hope is that many folks will develop and share templates, and the
> folks developing templates won't necessarily be running glance or have
> access to upload their own images into glance (that is a service operator
> choice).  If we force this basic info to glance, we are limiting our
> portability (and therefore adoption story) of Heat.

What's really going to limit portability is having multiple overlapping 
undocumented proprietary schemas for parts of the template. I actually 
think this has the potential to damage the Heat ecosystem, and that's 
one of the reasons I'm pushing back on it.

regards,
Zane.