VMware Carvel Community Meetings, 1 Mar 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Carvel Community Meeting - March 1, 2021

Description

Carvel Community Meeting - March 1, 2021

Topics discussed include: Announcement of office hours, project roadmap for March, and backlogged items.

Details and Agenda here: https://hackmd.io/F7g3RT2hR3OcIh-Iznk2hw

A

All right welcome to uh this week's edition of the carville community meeting, just a reminder it's being recorded, so please adhere to the code of conduct when um engaging in this meeting.

A

One of the things I wanted to point out before I hand it off to helen is this project is so robust with the many different repos attached to the tool suite that we decided that a weekly community meeting just wasn't enough to go over all the items that we wanted to go over and discuss.

A

You know anything that was pressing, so we are implementing the carnival office hours and the first one will be march 11th, which is next week next thursday, and it starts at 11 30 a.m. Pacific time we're gonna be doing these every second and fourth thursday of the month.

A

um All of this is listed out in the agenda and we'll be posting it to our website and the repos. With all of that information, the same zoom link will be used.

A

um The one that we use for the carville community meeting will also be used for the office hours and the way that it'll work is the first 30 minutes. We dedicated to any discussion items that we didn't get to for the community meeting or anything that we need to discuss at that time and then the last 30 minutes more ever.

A

However, much time we have left we'll open it up to anything that you from the community needs to discuss, if there's anything that you're curious about with the projector you're having any issues on the maintainers are healed here to help guide you through that, and with that, I will kick it off to helen to share a screen and we'll go over uh the rest of the items on the agenda.

B

Cool, um so what you're looking at is the carnival roadmap for the next month or two, as you can so for march and april, and as you can see, I kind of grayed out the april part because uh roadmaps for the future types of work is, is still being flushed out and as we go, we would learn more of what we'll be planning uh work. What we'll be delivering in april so just wanted to call that out.

B

So you see march and april, and then we have some of our um tools highlighted on the left, ytt image package, cap k, build and cap controller and also wanted to provide an disclaimer there. We also have a lot of other experimental tools that you might be aware of, such as kwt terraform provider or vendor. And such um you you the reason why you're not seeing them it's not because um we it's it's not because we don't care about them. It's because you know within our bandwidth. Within our capacity.

B

There are some of the things that we prioritize, um but please, if you have, um you know, need improvements or defect, fixes that you would like for other tools that are not reflected on this. Please please, please communicate that through kubernetes slack or you can. um You know, feature vote on github issues by showing adding thumbs up or various emojis, then we will take that input and prioritize them accordingly. So don't read this as oh.

B

We forgot about all the other tools, so I wanted to call that out first, um so um the way it's kind of also represented here, as you can see, some are you know, some boxes are filled with colors and some blocks are just outlined, so the boxes that are filled like this one, for example, yt schema type checking these are um actively being developed at the moment. So those are delivery, work or development.

B

Work versus outlined items are things that we are working on coming up with uh the solutions for it designing of it, or even on this trying to understand the problem space better, so we're kind of distinguishing that here so art our team. Does you know that dual agile track work, where we're doing both delivery and discovery at the same time?

B

And that's why you see those all kind of represented here and we wanted to be transparent on you know not just what we're building day to day week by week, but we are what we are also thinking of building in the future. So some of these items that we are doing discovery work on uh will come in in the future delivery work so yeah with those um kind of like things shared for ytt, um currently we're actively working on the type checking piece. So it's currently noted as experimental toll.

B

So our goal is um to remove that experimental tool flag and um making sure the schema work. Is you know solid that you can use it for your production, not just as an experiment. So that's what we're working on and then in parallel, the discovery work. What we're doing is abstracting away data values. What I mean by that is for users of cad controller who are managing their packaging, that they don't need to know under the hood of cad controller.

B

It's using ytt for schema validation because technically it could be any kind of any other tool that does schema validation. So that's something that we're going to start um trying to understand the problem, space more and then coming up with the solution. So we will.

B

We will also push out those design docs to y'all as well, so that you can also provide feedback if there's any so that um discovery work, we're gonna work on and after that, we'll continue to do more discovery. Work on ytt schema work. How can we generate open api compatible schema from ytt schema is the next track of discovery. What we'll be doing last week, we already kicked off this discovery work. um So this is a docs improvement guides and adding guides and code examples to ytt. We've heard that ydt is hard to ramp up.

B

It has a lot of good reference materials on what features are available, but maybe the next step of how to use those features might be lacking. So we heard on your feedback- and we want to you know, provide those how-to guidelines um with um providing a code example, so that, ideally, you could maybe just copy paste some of those and get started and coming up with the pocs proof of content much easier.

B

So that's what we're working on and for image package we're currently actively working on performance improvement of image package, the next we are working on recursive bundles, so we've already worked on um both designing and I think we we've already kicked off the first story for recursive bundles, so how to bundle a bundle, um that's what we're working on and in parallel, we also have a pair actively engaged in um trying to come up with a solution for image signing.

B

So we are actively engaged with notary, v2, open source project in shaping and kind of understanding what the image signing and verification look look like, and also starting to get involved with google's cosign open source initiative there, so so that we could provide image signing through image package in the future.

B

I'm gonna jump real, quick to cap controller lan, so we have also packaging api work. That is currently in progress. The alpha version is available, so you can check it out from the carbo.dev cap controller site you'll be able to access the documentation, so we will continue to work on building out the packaging api for cap controller.

B

Cap and k build, so we don't have um like other tools like big chunks of feature, work that we have identified, but we will continue to work on future improvements like small feature, improvements and bug fixes along the way, um and we actively have. We usually have about two or three tracks of work um per day that um based on our capacity that's available, so based on that capacity, we will be able to pull in these prioritized github issues for a cap and cable.

B

So you know there's we know, there's a lot more coming a lot more still in the pipeline, so we'll continue to prioritize them and again, if there are specific issues that you want us to prioritize to work on, um please let us know and yeah. We would like to either get those work started within the team, or we would like to help you to possibly contribute to working on those github issues as well to involve you in more of the collaboration there.

B

So this is the high level roadmap for carville this month and then next month, um during the community meeting I'll share more about what's coming up in april and may thank.

C

You thank you helen.

D

Can I see a question about that diagram? Real quick or this is my.

C

D

Yeah yeah, I'm just wondering for the rectangles that are colored in I.e, the ones that are signifying feature, work that can be delivered. uh Suppose, there's a user. Does it communicate when it's going to be released in a version. So if someone is really waiting for schema type checking is this kind of communicating that mid-march there might be a ytt version that has schema type checking available to use.

B

At least that's what we're aiming for um it's as any romance, it's um there's room for change, but that is what we're aiming for.

D

Okay, that's cool just wanted to clarify that thanks.

B

E

B

An interesting.

E

A

Go ahead, nancy go ahead. John.

E

um Just to clarify that one actually specific example, because there's a bunch of features that we're collecting under uh this a a feature flag. So there's an experimental schema enable flag on ytt, although that feature might be available, might still be considered experimental as we work out the full set as well so anyway. Sorry, I wanted to add that clarification thanks nancy.

A

Yeah, uh my question is just about um the accessibility of the roadmap. Is it on github or anywhere that the community can also uh view the roadmap.

B

um We will make it available. um I think we'll decide what will be the best way: okay to be accessible, yeah, okay, great, not a secret. So all right, great thanks.

C

Cool any other questions from folks.

C

All right, let's dive into some stories, then um so we'll start with recursive bundles today, um right now we do uh have a story that uh dennis just picked up this morning. I believe, um but we'll talk about a couple of these stories that we haven't discussed yet so, starting with this one, so copy recursive, bundles to tar and import the tar to registry.

C

um So the use case here is, as a software packager, I want to create a bundle to distribute the multiple applications that are part of deployment, so that application consumers can retrieve all needed software from a single place.

C

And then so, there's a couple of pre-requirements in here, um and these are just some- I guess examples that will be used repeatedly throughout this epic, so just just to provide some grounding here.

C

There's this bundle one and it has an images yaml that looks like this has a a single image it's pointing to and then a bundle bundle 2 that it's pointing to so bundle 1 has a recursive bundle and then looking at bundle 2 its images lock just has an image, so you can think of it as a a simpler, a simple bundle that doesn't recurse, um and then these are just kind of listed as as the two images being used.

C

So now, if we step into the first scenario, so user creates tar from recursive bundle.

C

So, given this images lock, which has both bundles that it's referring to.

C

When we run push dash b, the recursive bundle with the experimental flag, then we should see that it succeeds.

C

And then, when we run copy this uh from from the registry to atar, then we should see that it writes something locally here to a purple, local tarball, um and then this last step is really just to show that you know. If you were to extract it, uh it would be what you expect. The the individual layers here.

C

Are any questions on this first scenario.

E

I have a kind of fundamental question: um I'm trying to remember where this landed. Do we know how do I know whether a particular image is a bundle or not.

C

C

I think it will, at a minimum, have a certain structure to it, but dennis I see you reaching for your mic. If you wanna answer this.

D

There is a structure, I believe, there's an annotation or something we also stamp on these images as well.

E

D

E

D

File, oh, when we push it, I should say so. When we pull it, we can check the annotation that that the I guess it's in the manifest or it's in or in the on the image. Config there's a labels section, and we have a key that signifies it's a bundle, but definitely also the directory structure inside is if we have a dot image package, it's bundled directly. I can't remember what it is, but this is also our.

D

C

See any other questions on this scenario or in general,.

C

Okay, are you okay to go forward.

C

Cool, so the second scenario here is uh user imports from recursive bundles tar. So this is copying from atar to a repo.

C

um I talked with sojourn wrote these stories he's just not here today, so I'm just kind of representing the track. um We talked about this one a little bit and we took off the um experimental recursive flag because the deduping really should have happened from like this uh sorry. Well, this whole scenario, the first scenario, so we felt that we didn't want to introduce any.

C

I don't know, maybe like temporary logic there. By by introducing this, you know feature flag, because under the covers, it doesn't seem like anything, would actually change. Does that make sense to folks as to why we we left the experimental flag here off.

A

She's, pointing.

E

Turned on sorry, steven turned on for scenario, one it's required for scenario, one in order to get that behavior but scenario; two, it wouldn't.

C

Right right, just the pre-existing copy functionality should already work. Oh, I.

E

E

Okay got it. Okay, do you have a question stephen yeah.

F

Yeah I wanted to ask: can we get that noted down as part of the story, because, while I'm here, I it all makes sense and I agree with it and then I'll pick it up in a week and go? Why doesn't this have.

C

That flag cool I'll take a note to add this right after this meeting.

D

So what does it say on the right of this table? This, like the text, should basically this we get this for free. It sounds like so like there's no changing in the output of anything. It's just almost a no-op scenario to. We think we think.

E

C

Okay, but we we did want to include it as a scenario to make sure that we do explicitly test to test this out.

D

Yeah I mean just reading the output, we say importing five images. Let's say like this is a ux question. Is there any value in differentiating saying imported two bundles three images: they're all images at the end of the day. So that's why we said importing five images.

C

Yeah, I don't know any any any thoughts on there in general. I know we've. I think, we've just been staying consistent so far by saying images but open open to thoughts.

E

E

I think in terms of helping the user envision, what's going on that, like the bundle sort of leads its flock of images along, it can be very helpful that the output um kind of helps match that sort of my two sense that, like um they, they don't just look like one among them, but they sort of the bundle, images kind of lead the rest of them and that whatever was contained in a particular bundle. We could see oh yeah, it's still still in there.

E

um I guess maybe another way I'm thinking about is like, what's going on in my user's mind when they're actually reading through this output, are they trying to verify something, however lightly or meaningfully one thing.

C

That I don't know if we can guarantee at least right now is the order that these are going to get copied. So I'm guessing like right now we can't guarantee it so that would be added functionality. If we wanted to do something like that,.

E

Yeah I mean maybe we're describing like a future story that improves the ux or something yeah.

D

The imported five images happens that after everything's been uploaded, so we could print out whatever we like. I think, you're right. I think it's like we.

D

We don't have a lot of control over the importing like the line items, although after the performance story, we do have a lot more control um in how we do that um because we changed it used to be like you know, each line item used to be shown up-to-date progress of what's been uploaded because we were handling all the uploads itself now everything's been uploaded by the library at once, and then we print out the importing messages. So we do have a lot more control.

C

Now, okay, maybe this is worth capturing as a as john mentioned. A follow-up story. That's like focused on ux then sounds like a.

D

E

D

E

My head goes back to, but but the point here was: is that whatever's in a tarball, whether it's a bundle or an image that just happens to be referenced by a bundle? It's still five images, and this is not expected to add any complexity whatsoever right now in the scope of the story. Okay,.

C

Cool we'll move on to the third scenario, which is a failure case of uh not using the experimental flag on a recursive bundle.

C

um So if you were to try and push a recursive bundle.

C

Let's see so down here, sorry, when you're, when you're copying the bundle.

C

Then we return this error.

C

And I think our code block just kind of shows.

C

I don't I don't know if this is an intention of providing what the right command should be or if this was just a markdown templating thing.

D

I mean if we leave it as.

C

We wanted yeah if we want.

D

To supply the command, it does add a bit more complexity in terms of implementation, so it sounds like it's intentional. Okay,.

C

Yeah, it does look like it is intentional.

C

Yeah, I don't, I don't know like happy to hear feedback on you know one way or the other or again I'm kind of representing joao at the moment, if, if the pair, that picks us up wants to have a deeper conversation on this piece in particular, I think you'd always follow up with him.

E

I feel like it's really friendly to do so to say, I think what you meant to say was everything um from like opportunity. Cost perspective like it does add, just just a bit more to the story and wondering if, like perhaps we can get sufficient mileage, especially something where there's no parameters for the keyword, arg, it's just it's an additional flag and we'd say include that I up arrow, add in that additional flag and hit enter.

E

Maybe that's not too terribly bad, but maybe in other cases, in our tools it might be helpful to like oh yeah. This might take a little bit to calculate. Let me just show you what what it probably would be uh is even greater value for for the user, um so I'm proposing perhaps that we could consider this part just mention the flag so make sure you include this. If this is what you want to do with your previous command.

C

That makes sense I'll uh I'll. Add that as a note to this story as well.

G

Question I have is, um I know the story that we talked about last week, which was pushing and pulling recursive bundles. um We use the flag dash dash recursive, I wonder like. Is there a reason why we're not using the same play here as the previous story.

C

That is a good question.

C

D

Think it's the same, but I.

C

D

Should they should these scenario? Examples also have a dash dash recursive flag or are they I feel like they're just missing him correct, like you know, scenario one. When I run image package push, I don't see a dash r flag or the dash dash recursive, or is that the experimental one.

C

Yeah, I wonder if we just have an inconsistency in the name of the flag.

E

No, no, I I think these were meant to be two different things. um The dash r genuinely means do this recursively.

D

E

D

So this command is missing the recursive flag. Then it sounds. I.

E

Don't I I actually, I can't speak for joao, I don't know if he was intending it for actually to mean that, because I know at the end, we want to have the ability to say I don't want a recursive operation for some reason, if it's like for bandwidth or whatnot.

E

So unless we've decided to change what the default behavior is, I suspect it's just a minor omission of like we needed to include the the dash r or dash dash recursive to make sure that it happens recursively and that this other one, this experimental flag, is a feature flag. Is whether or not we even turn this feature on at all.

G

I'm not sure how you could use the recursive flag without this experimental, recursive bundle flag seems.

E

Like they'd have to go.

G

E

Yeah like at first, you have to keep them together, and it's like this really long command line you're using a an alpha feature, I think, is the signal there. That's how I was thinking about it.

D

Yeah so back to carrie's question then like should the copy command also have that dash dash recursive flag? I think I mean with if you don't have it, what does it mean to? I guess you? Don't you don't know you don't recurse down through the bundles and you just miss some of the layers. That seems like not what the user intended most of the time. So maybe it's emitted intentionally. You always want to recurse bundles when you copied to a tar.

G

Yeah imagine a scenario where you would want to use the recursive flag without experimental recursive bundle. So to me it seems like maybe they're meant to be combined or like. Maybe this is just a an accident that we name this differently.

G

Maybe it's just something we need to check in with with. Would you wow.

C

Yeah, we'll clarify those wow. I think he's back in tomorrow.

C

But we'll certainly add that, as a note to the story as well.

E

Maybe for today, in terms of having clarity on the story, uh what if we stipulate the experimental flag was on purpose and it's meant to enable or disable anything really having to do with recursive bundles period, in that um uh there might be hints elsewhere, whether or not there needs to be a flag. Let's say if you're copying to atar or not, but that the overall functionality here that, like creating a tar ball from recursive bundle, needs to be able to happen, quote unquote somehow so like.

E

uh Hopefully, it's a knit is what I'm trying to get at as to whether or not that flag exists, that, like a dash r flag, exists.

E

And that we're not trying to um do the the truth table of okay did you include the dash r and the experimental and the dash r without the experimental and then reverse like I, I don't think that's the intent here.

E

C

Sounds like a good assumption for today. Okay folks feel ready to point the story.

D

Just wanted to make sure I understood the last sentence in this thing, which was something about a note making sure something delays get deduped if you scroll down uh yeah. That note, um is that just coming for free or is it something we need to do for that? Is it like why? Why do we? What do you think we added that night here for.

C

um Yeah, I think.

C

So when we do, was it the first story or the first scenario.

C

C

I guess ensure that like when, when.

C

I guess I yeah I'm trying to remember my conversation for a while, and I think this should have already been done as part of the first push story, um but I think the note is just to ensure like let's, let's just again, double double check. So I think that's what that was about.

D

Okay, basically to know what make sure like, I don't think it's even possible. I mean to have duplicate layers, because the file names you know are content.

C

Addressable, so you know I.

D

Guess you would just overwrite worst case the layers on top of each other, but they're going to be de-duped in the top or at the end of the day, and that's also fine, okay, cool.

C

Cool good question.

C

All right should we throw points and see.

C

Cool three two one shoot.

C

Twos consensus, twos.

C

Cool, uh let's see we'll keep going, then um I think with the three-pointer in flight and then a two after that we're probably going to be okay for the week. Does that sound right with folks yeah cool?

C

um Do we have a schema story to discuss, looks like it.

E

Oh yeah, we started talking about this last time. We didn't get all the way through right.

E

Okay, so if we remember, we have schemas that are defined in our what we call our root library or base library like that base directory, where your ytt files are at, and you can include a schema in that as well. Now. What we want to do is turn our eyes toward a more advanced usage where you're able to collect some templates and data values, and you stick them into what's called a private library which sort of sits there.

E

Latent doesn't do anything until in your main library, you actually do a library get, and then you can do operations with those like evaluate them into your output as a main library.

E

It also means that those things sort of operate as as they're called libraries they can get vendored in um and then pull and then incorporated in multiple uh ytt invocations.

E

So we want to make sure that a that this is going to be working for that usage and b, that there are proper, uh either interactions or or lack of interactions between the root library and the private library.

E

So so here we have the first scenario, and this is where we we do have a private library and it does have a schema defined in that private library. So here our private library is named foui and it has a schema file here. It's describing the values that are expected by its template and then when we invoke uh sorry so here here that here's that actual the contents of those files, so there's the fui schema and the food template you can see, foo is defined and it's going to be an integer.

E

We use that there and then in our root library. We happen to have a values file here, let's see sorry so in our in our root library. All we're doing is we are um loading up that foui library set library get fue, we're evaluating it. So now we've got a doc set, that's been fully rendered now it's just yaml and we're going to say: hey, take this document start and replace it with the contents of that evaluated library.

E

That's what this line is doing so if, in effect, we're just invoking the private library, then the other file, that's sitting here in the root. Is this tricky little data values file and it's tricky because it's a data values, but it's targeted declaratively to the private library. So here this is an attempt to change the value of foo from the default 42.

E

Sorry to seven, and in this scenario we run this whole mess. Then what we should get is just that net result, which is library, foo evaluated with the value, seven and replaced at the root, and it gives us a single document that has foo. That is seven.

E

Any questions about this first scenario.

F

D

Knowledge of this everything in.

F

The first scenario would work the same if we did not include that enable experiment schema flag and there wasn't a schema file for it.

E

um It doesn't because this schema file is currently not scanned as a piece of schema for the outer library, and so it actually gets interpreted, as I think just like a template. I don't know it's it's it's awkward, it's not right, it's not right!

E

So no, there is work that needs to be done to um ensure that when files are being fetched in the same way, note that there's prior art here about like how are data data value files being handled they're, probably assuring some when we think about that same code path for schema, there's, probably some short circuiting somewhere where we well.

E

There is where we say: okay, if you're loading up a private library, it's schema is currently any schema and we're not scanning for and executing that step of loading up, schema preparing it and then applying it to the data values inside that library.

E

Cool great good question good point anything else on this first one, it's good to know this. The next one's gonna build off of it.

E

Okay, great thanks, aaron, so good news. Everything that you just saw is exactly the same with one change.

E

So, instead of that values file that was at the root of our library, instead of it having um foo, that is an integer, it's actually a string which is the wrong type, but note we're still targeting phui with that value, and so when we go to run that pile of ytt stuff, we should actually get the corresponding schema.

E

E

Any questions about this one, I think.

D

This was mentioned last time, but the error message: it says: values yaml, we're going to include the path to the private library to differentiate between the private library values. Actually not given the path, I should say to values yaml. If you have multiple value, multiple uh or maybe I'm talking about schema.yaml yeah, maybe.

C

D

Schema yaml yeah. We should give the path to the private module library for schema.yaml yeah. We.

E

So the net of all that is, that is a really good. uh That would be a great improvement. It would be out of scope for this story, so we want to do that. Yes, but not as a part of this work.

E

Because we do have that mirror problem with I mean it, it can be awkward when you have the same name files. Yes, in your data values.

E

Okay, so far, so good, okay, so that whole thing, those first two scenarios were declaratively setting a data value all right. So this next one scenario: three it's very similar, except now the difference is in our route. You'll notice, there's no values.yaml file.

E

Well, instead, what we've done is we've actually used a programmatic approach to do the same thing, so first we defined a function, it's called dv's data values from root, and then we define this variable and it's foo from root library, and uh I guess to keep things interesting, I'm going to jump down to the schema here. I guess to keep things interesting. We decided to change the type to a string, so it's no longer think foo is now a string as defined by the schema okay.

E

So, coming back to that root, yaml, it's the same business library.getfui and here's where the interesting part is where above we just immediately evaluated now we're calling that, through this fluent interface, we're calling this with data values, and so what that does is it allows you to set the data values programmatically. This is effectively the exact same thing as what happened above from the yaml that you've uh set here.

E

So it's a it's a map of values, so we're actually passing in this foo and the string from root library directly in and then evaluating and then rendering out replacing the contents of the root document.

E

So here's your schema. It's apparently a string this time and then the template, that's going to be evaluated. Just displays foo, real simple, and when you run it you should get your nice output interesting. One way, I guess that's legit yaml! I don't know if that's exactly right, but just imagine it is.

E

There we go so programmatically doing it.

E

These are two different mechanisms by which these data values are landing in the in the private library. That's why we have these scenarios.

E

Okay, I'm seeing nodding heads feel free to interrupt. Let's keep going all right, don't worry this. Is it so scenario? Four is same thing as you guessed it it's identical to what's above, except uh we put in lucky number 13, instead of a string, and so we should be getting the opposite complaint that we expected a string as defined by the schema, but we actually got an integer instead.

D

This question might not make any sense. I'm sorry, but like the what the error message when it says values.yaml, was there a values.yaml file? Oh.

E

No dude, that was, it was copypasta, so it would that's a good question. I can't remember how we render that out. We will have. This is called in it in the code. We call it an associated name fancy, um so we will have some value there for that. I can't remember what it's called when you're getting it from here. It might actually end up being be from the the root.yaml that might be where that comes from.

D

Okay, let me just update that when we do find out what it should say, yeah.

E

E

I mean the the so. The spirit here is that you get the corresponding scheme error.

E

It could catch.

C

See any other questions from.

C

C

Get ready to point.

C

Cool three two one shoot got some twos and a five.

C

uh Maybe start with the five.

E

Let's see, let me think about the things that I think are so there's these two different code paths, there's the declarative versus the programmatic and within those there's, some part of like scanning for schema um that I mean we'll, hopefully be able to just pull that in maybe maybe there's a way in which that folds in um so that was one source.

E

um uh Down down the two different paths, so for me, that's sort of like almost three right there in in my head um and then there's.

E

I want to say like like they're: uh it's import. The part of that enumeration is like doing doing the uh that, probably generally generalizing how we're like plucking and setting aside schema values separate from um other other documents, so I just think there might be like some additional refactoring work there. uh So there's just enough uncertainty with all with, like the meat of all those different things to in in my head that I was like all right. Well, it's if I had a four, I might throw four but we're doing fibonacci.

E

So I'm like all right. That's a small five big three! That's my thinking! Those are the things I saw complexity, wise.

C

Anyone from the two want to speak to their pointy.

F

I think I really saw two main sources of complexity, which was one plucking: the schema from the library and two figuring out how to get data values to that schema. Think I'm I'm a little more convinced upwards from the multiple ways of doing that.

F

But those are the the two places I.

F

C

All right: how about we throw points once more so pointing three two one shoot.

C

So we got three threes and a five: uh how about we go with majority of threes? You okay coming down john, but we have this recorded, so we can go back in time.

E

The important part here is the conversation, absolutely yeah rough consensus is great.

C

Cool, uh let's see if we can fit in one more story: a cap story, so hopefully you're not getting too much whiplash with all this context, but uh john, would you be able to speak to this one as well.

E

E

Actually, I think, I think, if we gave folks just a minute to read through this, the narrative actually probably does a great job of summarizing. So let's just kind of check that.

E

E

Okay, so boy how? How timely was that that garrett gave us our boxes and lines on cat, because now we can collectively talk about the dag about the diff, dag and one way of another way of like complimenting this description. Is that hey there isn't there's this resource? That's in the cluster and uh the most obvious one is a crd where it's not in the pile of manifests. It got created by something else in the middle of the deployment, usually a controller.

E

So I want to be able to tell cap that that will exist, but how how to do that. So that's what this explore is about and instead of just focusing on crd, let's talk about well any resource that we might wait on. That. We know is going to be generated, that the um the author is going to know is generated by some machinery inside the cluster.

E

So this gives us the ability to uh sort of point at a thing that we don't have the definition for that. Somebody else will generate for us nice and then wait on that thing at the right.

E

Moment, whereas today everything has to be inside those manifests and we build up our dag from what we. What cap cap builds up, its dag based off of all the resources that have been specified and given.

E

Any questions about sort of the context here, the wording of any of these questions.

D

So I'm just gonna assume that the the work to just like we, this is clearly uh like something that we're lacking in cap. Like you know, we have change groups, we have those other kind of ways of grouping resources and and having some sort of dependency using those potentially but like that. Just doesn't cut it for, for these cases at least, and we want to explore a solution and we have- and all the links to you know, to describe the problems like link.

D

52 will have a bit more like a customer's interpretation of what their problem was as well. Okay,.

E

Yeah you'll get that that felt sense description, which is that it fails. The cap deploy, fails.

D

And we want this solution to be a bit generalized as well, not just for this specific issue, but just how do we depend things on general and how do we deal with deleting things and cascading deletes and okay.

E

Yeah, so the knee was expressed in terms of crds, so we know for sure that that is a genuine need, whatever whatever feature we built will meet that need, and what we want to do is think about it. More generally, our solution, although it'll, solve a specific problem, there's real value in thinking about a generic mechanism, more general mechanism that will satisfy this, and so today, in order to affect a change group, I need to have a resource on which I am dropping that annotation, but if it's being created by something else, then what so.

E

This is a this is an attempt to explore. Well what would be event what so uh there's all kinds of ways to approach this but like in order to make it a little bit concrete like what, if we had some way of describing a resource like it's another file, but we've marked it in some way that it's not a resource uh that gets applied.

E

um There's challenges with that like, for example, we've done work in the past where we said well, we want to be able to take the exact same yaml. You could pipe in a cap and we'd want to be able to up use cube control to just allow it to go through without it having any hairy challenges.

E

So we've done we've like wrapped cap controller in a config map before uh to sort of allow for for that to happen, um so that sort the same spirit still exists. We want to be able to co-exist with other.

E

Ecosystem tooling, so you wouldn't have to necessarily say well. You can't run that at all with, say, cube controller, cue cuddle. However, you say that.

E

There's this described sort of like a suggestion of these outcomes so like working on a proposal uh getting conversation going so there's people here that that we could talk to who've reported issues talked to folks who have mentioned, uh go back and say: hey, you know, there's an idea. Is this gonna how's this gonna? How does this feel how's this looking kind of thing and then synthesize that.

E

D

So gatekeeper and decks, these are people who want this feature or I'm reading this as the products that have this feature.

E

Yeah, so these are products that create their own crds, okay, so the if a user was using one of these features, they've and they're trying to cap deploy with with these uh products um installed.

D

It breaks cool and so, like tying us back to the round of feedback, we reviews to present the proposal we want to help me tie that again, are we thinking of like asking gatekeeper or the decks like what they think about a proposal? Is this just making sure that our feet our thing works with gatekeeper index?

D

E

A part of it, the other, is that whoever reported issue 52, which was sort of the spark of this um and those who have responded in kind and on that issue, and I think we might have like some references to slack conversations. Any of those folks are great people to come back to to say: hey. uh We have this idea. We want to make sure that it's going to meet your needs before we commit implementation to it, got it great questions.

C

Folks comfortable pointing to see uh where people are landing right now.

E

Yeah, and so what we would be estimating is sort of like an equivalent to kind of a time box around the proposal is that fair.

E

Or to get started here yeah, how do you, how do you want to manage that.

C

It's a good question.

C

I don't know, is there a time box that seems reasonable.

E

Well, one thing that that can be helpful here with with the time box, is sort of giving us a sense of, or putting some back pressure on uh getting ensnared in lots of detail, uh but having enough space to to really think this through and and poke and explore. These things also note that, like many of us, are aspiring to be reviewers in this code base, so that's some work to to load up um the design of what's there as well. So I'd want to incorporate that in the time box.

E

I'm thinking like I don't know three three days and then kind of checking in and not necessarily having a proposal necessarily written by that time, but like have have a sense of direction. At the very least.

E

So I guess that would mean if we were trying to have a little bit more firmer time frame. It would be like four.

C

C

Yeah, that sounds good, so we can add a couple notes here at the after this then to just mention the time box and checking in.

E

And anything like this, uh we'll want to lean on folks who are uh like any approvers of the tool, or you know my job is to help support these things too. So it's not about going off in a corner and uh sliding pete's boxes under the door until an answer comes out, but like really having that space to have uh time time to think things through, but lean on lean on your teammates.

C

Absolutely does it make sense to throw an estimate on this? Oh, you can't throw a zero.

C

Oh interesting. I have to create one. Oh yeah are people okay with the zero.

C

And we'll just add the note about a time box.

C

C

All right or a couple minutes over, but we got through a few stories uh anything else from others before nancy closes out.

A

Okay, so thank you, everyone for joining the community meeting. Today uh we had some really great topics to to go over looks like we didn't, have any um discussion items that uh we didn't have time for, but uh we will have another community meeting on monday, which is uh march 8th and we'll also kick off our office hours next week, which will be on march 11th.

A

So if you are watching this from home and there's something that you would like to bring to the team to discuss regarding karmal they're happy to help you in any way, whether that's how to get started with it or if there's something that you're blocked with on implementing on your own, please please join those office hours join the community meetings. We would love to have you participate all right with that? We'll see you next time.

A