Kyma Core Special Interest Group Community, 25 Sep 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Kyma Core SIG community meeting 20190925

Description

1:12 Answering questions from the community
12:10 Community update - Kyma@Conferences
14:40 Function controller - Why Knative instead of Kubeless. How the Function abstraction will work. What we already have and what are the next steps
29:46 GitHub and Slack connectors. How they work, what can you achieve with them? Demo showing the automated text analytics of comments posted in the project
38:17 AsyncAPI 2.0 in Kyma with 1.6 release. Why we even need the spec. What we have thanks to it. A new feature of validation and conversion. You can skip 42:40-44:30 as Lukasz had some emergency to overcome during presentation :)
52:50 Overview of the proposal on how we will handle the migration to Kubernetes 1.16
59:08 Q&A. Follow up on how Headless CMS gets configuration for AsyncAPI validation and conversion

Meeting notes: https://docs.google.com/document/d/1vWleTon7sJIk0teee4SoVezS4mR3K8TlkvXkgLJwRD8/edit#heading=h.cr48g5rz2sss

A

Screen, okay, so with Mila. Let me welcome you in our next korra meeting and I see a lot of new people here. So just kindly reminder put your name under the attendees list and be aware that our meeting will be recorded and then upload it to key my youtube channel on today's agenda. We have a lot of nice thing, so at the beginning we will start with answering questions from our community. Then there will be a short term media date to say on which conferences?

A

Will you find the keema committee crew, then for Binion will show us the overview about function, controller and next I can already spoiler that we will have a really nice demo about new github and slack connectors in keema. Then we got Ronnie's key, we'll talk a little bit about a syncope supporting keema and last but not least, you know who they will share with us details about strategy for upgrading API version, kubernetes resources. So let's start with the first one.

A

So questions from committee I know that we have P be here so probably hear you answer those questions and let me read the first one. So what is a kima runtime.

B

The answer is simple, so kima runtime is another name for a class kubernetes cluster that is running kima in it. Yes, oh, this is just the name for a kubernetes cluster and we use that mainly in compass component, which is supposed to be somehow the component that is managing many kima runtimes. So you can imagine that you have more than one installation of kima or more than one cluster.

B

You can have one cluster per customer, probably maybe different scenarios for for different environments and who you want to manage application connectivity in one place and cluster is something that is not maybe something permanent. You can have new one okay and delete it. You can create it as as you need, and then we came with the name of kima runtimes, to have somehow notion of this item. We we can manage in in compass, so compass actually is using the schema runtimes to assign some applications, some application connectivity to some random.

B

So you can imagine that you have someone on premise, services, cloud services and you want to connect them to different kheema instances, and this is the name I'd hoped do. We have jung-in I.

A

Think that we have.

B

C

A

B

Okay, here sorry, if it is clearer than here, I will stop there and.

A

As far as I know, we have some documentation, but it's explaining that right.

B

Yep, probably we have yes, so with this inside of compost, probably in there some architecture, the thing but I used to be honest, I haven't prepared it yet so yeah. Let me yeah put maybe the answers in the Sakura later. Okay: let's do it.

A

So maybe you can start with the next one. Is there a compost demo on the horizont.

B

It's not a question to me, maybe more to pk, who is not in the call, but I think that we will have soon something that is really usable and we will present it also on this meeting in the succour. But I'm not sure if it will be next week or and in two weeks or in four weeks. So.

A

Sooner or later, yeah yeah good to know so.

B

Yeah, if you can actually start playing with it, but yet this is still working progress, so yeah. Definitely it's not that easy think to start it. This teal incubator, it's still not finished, so you can expect some problems running it, but it's more or less working yeah. So we did some some demos with it and if you are interested you can start and you can ask in the in the slack channel for help. If you face something some problems anyway, we will face them again when we are live with with it.

B

So this will never the better.

A

Good, the last one any advice on how to run general were closed in a chemical, aster I.

B

Need, yes, you enter maybe elaborate a little bit yeah.

C

I have some Ivan like a shopfront right and I'm, basically building a little application proxy that allows me to like that occasion, part that's gonna talk to coma and I'm like where do I host that do I put that on there like a cloud function or do I put that I'm, a VP, C or somewhere else, or can I just run that as a container within the Karma cluster? That's already in front of me and I'm already managing you know, yeah.

B

So definitely you can run it inside the kheema cluster. This is where it is supposed to to be. So if you have an workload for that is extension for your application and the quemic roster is, is the the right place? Of course you can use some external hosting, but this is not the first scenario we thought about yeah.

B

So when we thought about kima indexed and the extensibility scenario is that you want to have something like kubernetes, some cloud native environment, you can benefit from auto scaling the the you can benefit from all this observability monitoring, etc, and you want to run your workload in such environment, not necessarily in the environment that is brought to you. With with the application you extend yeah. So that's our main motivation to bring it to the kubernetes. We do not hide kubernetes and kubernetes workload is the way to go right now.

B

You have two options, so you can write lambda function, which is currently based on cublas, but this will be soon replaced with key native and serving, and the second option is regular deployment. So you can deploy your container whatever you have and- and you can also bind it to the application you can trigger it with with the events.

B

So that's the two options you have right now and the third option that to to host workload outside of Keima is also possible, but it's not something that we actively encourage or support yet, but on the roadmap we will have something that that will allow you to put maybe some workload in the functions provided by the cloud hyper scalars. But this is not in the stage that we can talk too much about it. I hope you ever concert.

C

A

Any other questions from you from Yerevan.

A

Hey I guess not. The next question is from Holger she's, not on our call, but he asked that question in context of Hamas jihad, provisioning kubernetes cluster, using Timothy lime, and the question was what about hidden hitting in the direction of class of API, which the different providers for any many clouds. Basically I think that he is asking why we choose terraform instead of cluster API, so cluster API for those who don't know and the project under kubernetes six organization, and basically it aims to provision also cluster using own API, but on the other.

A

Another type is clear: site. I discussed that and ask people behind our commercial I project, and us answer was quite simple. The cluster API is not robust enough for now. Basically, you can see even in your in their documentation that they are saying in the prototype stage, and that was the main reason why, which is more production version like terraform. But what is important is that tera for implantation inside crema CLI has been done as a plugin, so we can implement more operators.

A

So in future we can add support for faster API, and this stuff is not closed right. So we have some extensibility here. I see also a new question from John.

A

If you want to have development staging production, does that mean different namespace in one cluster or multiple cluster I.

B

Can answer as you wish? Yes, oh and as you can afford for here, so if you can afford for multiple clusters and have separate cluster per each environment, of course it's battery, so it's better separation, but it's more expensive. Yes, so usually people do they're staging tests in one cluster using separation for namespace and most people are rather avoids having production cluster shared with with other environments. Yes, so usually you have separate cluster for production and for other environments, you can have one shared cluster with different namespaces. That's what most.

A

People do yeah and basically just you are running kima on community cluster, so we applies the same rule like you use for communities. It's alright, so I think that better is to use the dedicated cluster for production, because you know staging and development you can cure, for example, API server and things, and because of that, you will have some downtime, so I think better and ways to have different clusters at.

B

Least, I wouldn't put developers and production clusters to experiment.

A

B

Some mastico TIA yeah.

A

Yeah, that's true because you, you assure the control, plane, okay, so any other questions.

A

Maki so let's go to the next item committee update, oh gosh, your turn.

D

Yeah yeah, okay, so we just wanted to share with you that we will be present at those two conferences mentioned in the omitting notes. So in case you would like to meet with us or even have a dedicated session showcase of kima. Then you can let us know so in case of API specifications conference, that's a conference in Canada happening in October and I'm, the one that is going there to present there.

D

So if you'd like to have me even at your company to do some lunch talk about kima, just let me know up front so I can plan and prepare, um and if it comes to keep calm, we already confirmed that we will have few people joining cube con. It's gonna be me and without my daughters for sure it's gonna be Mattel's. That is moderating this meeting and he will also share some responsibilities related to his presence.

D

As a chair in the six service, catalog he's gonna be ahmed running the native initiative from our side and also doing the notes at the moment. The reversal pavel that was also recently presenting on conference cognitive conference in Warsaw and there's gonna, be also caption scheme, who is a product model behind compass, so we are also open to do any additional presentations there, regardless the con. We already actually had one discussion with one meet a local meetup about cloud native. If we got accepted, we will have three or four sessions as far as I.

D

Remember so one and a half hour just about kima, so yeah a long story short. We were there in North America. If you want to have us involved in your projects. Let us all thank you.

A

Very much any questions: okay, good! So next topic overview about function, controller, carbon! You can proceed.

E

So the function controller.

E

Why did we change something first question if this was the first question, so the idea to change from the current implementation based on cubelets to a now based one on K native, is that we get scale to zero out of the box with the new one or as we wanted to leverage the fastest out of times. That's could be achieved using K native serving and obviously yeah. We want to base as much a schema on K native, and so we want to base it just on native and use it's modular design.

E

That's the reasoning behind the switch now a quick architecture overview on how it will look, so we will have 2 lambda UI, as we have currently then the lambda. U I basically creates somehow a K native serving okay native service. The code is somewhere stored, either on a blob, storage or disk. It part.

E

Get repository this is not and not implemented yet, but today we decided that we at least want to try to releases in the next release. Then, as soon as the code is stored somewhere, then we trigger a build. Currently, the first implementation was implemented using K native built, but as Kennedy filled is now deprecated we switch to Tecton.

E

This builds then a fully full-blown docker image with all the code and all the the runtime features that you require to run the function, purchases to a docker registry and then K native serving takes over and runs the image with all the benefits of kinetics, of like scale to zero. I said.

E

So what we did so far is we have something working, so we build images using Tecton. We have our initial CID specification and we run our functions using connective serving and I think then it would make sense to just show you something.

E

E

Here I have my little cluster in this cluster. I don't have anything except for.

E

I installed just default kinetic release. Currently it's based on a native ordered 9. Our function, controller kheema comes now with a native ODOT, 8 I think um so. On top of kinetic I installed also tacked on as we required for the build and that's basically it I did not put it in in a full-blown kima installation for this demonstration, because we have a little problem going on with the scale to 0. It's it's down pretty good, but never scales up again, but this will be fixed in the future so and obviously I prepared a function.

E

So, as you can see, you specify, as you did, for the old Q plus based and implementation, you specific specify your function code and this function field. You could add dependencies, we will support sizes, that's something that we will support latter on.

E

We will also support different content types and you can specify currently different runtimes. We configure out-of-the-box noches 8 and no js' 6 and that's it. So then what you need in your namespace.

E

Currently we we have not decided yet if it will support only one namespace to build everything or if every namespace will be built enabled and how we do the Tocco credential management, but for an hour you have we configure one namespace and there you can build your functions. To do that, you will need to set up a service account.

E

This service account is referenced in our function, controller configuration and it basically just contains your talker credentials because currently in kima you don't have a docker registry in kima. So you need to authenticate against something external.

E

So once this is there and then you also need some docker files.

E

Because, as I already mentioned, all the images will be built using, they will be built using taka files on Chronicle in the cluster. So, as you can see here, there is no magic going on just a normal. Talk of this is one for no j6 one for no trace 8, and it's just differentiates between the base images.

E

You can specify as many taka files as you want. If you have special needs so then let's apply our function.

E

So this is our function, that is, that was created if we check what it did in the background is it creates a task run for a Tecton.

E

This is currently running.

E

Once this is it's done, it will create the image, push it to my private docker hub account and start a connective serving.

E

So, as you can see, there's already a lot of stuff created, so we have now the successful build. We have our successfully running function and let's try to curl it.

E

And it responds as expected, so if I don't hit it novice trafficked and it scares down to zero using the standard, K native serving configuration and later on, it will scale up again but I. Don't think that's worth showing you.

E

Here and then there was questions about our plans for the future. So currently, yet, as you have seen, we are not fully integrated into FEMA. Basically, we are not integrated into FEMA at the moment, so we don't have a you. I can't support service bindings yet and you cannot expose functions running on the function. Controller using API objects. So that's the first thing that we need to do and the next plan would then be to support kit integration so that you don't have to modify the function CRD.

E

Every time you do a code update, but you can just reference a it's a github repository and your function would be rebuilt once you push there and I think that's it. So if you have any questions done, go ahead.

A

Okay, so maybe one question for my side: I saw that you have built back for node.js and I replying to add, build back for go by default or not that's a question.

E

That I would really like to hand over to Ahmed um I think it would be. It would be easy to implement it, but I don't know if we want them limited.

F

That's a good feature: request of I, don't think.

A

I, don't see a reason why not that's right now: I'm liking that feature in a current team, all right, Thank, You bliss, it's probably doable, but we don't have it I. Just.

E

To be honest, with the proper taka file, you can already do it. It just.

A

E

Correct base image and some magic in the dockerfile and the networks hmm not.

A

Basically, where I come up, platform use heavily go and also we are working cloud native word. So yeah I think that support for goal should be something to consider. Yep yeah.

E

B

Have another question so you mentioned that you haven't decided yet if it will be one namespace for building and or amended or many and I think that we should always have the principle that you can use many namespaces with which are which which can be used for building these functions. Yeah yeah.

E

As I mentioned currently with the current implementation and with what we specify them read me, you can do it in in each and every namespace, but it's only for ones in namespace. You have to do the steps for each and every name says that you want to build an but later on. We might just do this automatically for you.

B

Okay, you just wanted to ensure that you don't.

E

D

B

Give up this this.

E

C

D

B

D

So I have another question. So during your demo, you, basically the docker image was pushed to docker up right. Yes,.

B

D

What's your view on it for like for production, so I mean do you plan, or do you see it in the way that at the end, it's pushed to the registry local in the in the cluster? Instead of going to some my.

E

Personal opinion is that I hope that we never have a private talker repository in kima, rather than have a secure, publicly reachable docker repository.

F

By the, where discussions about it, yes for sure, because for the pros and cons for both, they have an opinion and we didn't decide. Finally, on that, so this publicly. This is probably one of the discussion that is very soon.

F

So when we we have a tick we, there is even an issue on keema for for doing this doc registry investigation, and there was an initial one that was made, but we still need more discussions or on it, even with the community and stuff, because one of the one of the point to view there for the pro having a docker registry is that if you backup your cluster and restore it, then you restore everything, including your docker images.

F

But one of the also valid points on the other side is that the kubernetes cluster is a set of configurations, plus storage, docker images that are available there. So the docker images are on the same cluster with the configuration that might be not a desired behavior. So maybe you need to you want to restore. We want to store your docker images separately from your cluster and then you with the state from somewhere else to have separate backups lifecycle, I, would say.

E

And just to add to this currently the current implementation. You are not limited to docker up only so you can basically use whatever talker registry. You want to use that's publicly available.

F

So there is an issue: I will put it in the readme and in the meeting minutes about the stalker history and feel free to discuss over it. Also.

F

Because we didn't reach a conclusion from somewhere, yet it's a bit need more consensus or more.

D

Discussions I think that in this particular case, it's it would be important to in both maybe like the users that we already have like. Maybe I would talk with net economy, guys yeah.

F

D

Someone because at the end I mean I, would imagine that some customer can say that he doesn't have a external registry and he just wants to take something from us. Yeah and I'll, just like I, said, pass the link and then I'll make sure that whoever I know in the community already is using him address country if they share their feedback.

D

A

Any other questions.

D

A

That was really nice demo. Sure thank you turn me on and now it's time for the next demo about github and slack connectors. So really a stage is yours.

G

So I will just share the screen and then tell you something. Give me a second okay.

G

Do you see yeah.

A

G

From here, yep.

A

G

So I want to show you what team flying skills a group of six interns have done during last three months. We have prepared two connectors for commands, one for slack and one for github connector for github. You can install them through ham chart or for an atom in Tomas. They both register in application registry. They have API specification in open, API standard and even specification in a single PI standard of both slack and github. We provide the specification for all events during installation.

G

You provide token and secrets for those services, so you don't have to care about authorization later. The connectors also also expose end points on which slack or github can send events. What's more github creates a web hook. Github connector creates a web hook on github automatically, so you don't have to do it manually manually on every repository you wish to handle and you can handle, of course, many repositories at once.

G

You can create one connector per token. You don't have to create many instances of it, but the installation of connector. It's just a service activation. It's it adds a service class in command. So you later, you need to create an instance to be able to handle events and you use application gateway. We have also created a scenario which uses these connectors.

G

It labels an issue on github that is offensive and sends a notification to slack so community manager can manage this issue tell something to the creator.

G

So on this chart you can see the flow.

G

So when a new issue occurs on github, the event is sent to github connector, which wraps the event to a format which is acceptable by comma. Then it's forwarded to even the bus and lambda which we created subscribe to this event and sends the issue by the end style, to a text analytics which analyzes it and sends back the sentiment of it. If it's too low, then notification is sent to slag and a label is put on the issue on the github.

G

You can install the whole scenario with an atom which we prepared and it's just one click you don't have to copy and paste the code and many other things. It's really easy. So, let's see it in come on.

G

You can see that we have add-ons for slack connector for github connector for scenario.

G

They have a documentation explaining how to install it, explaining the fields you need to feel you can add it here. There is one term you can provide many repositories or organizations. I have installed them before because it takes a while. So to install the scenario, you have to add the slack connector github connector as your service broker and the rest of magic will be done by this atom. So it creates a service instance of the connectors. It creates a lambda which already subscribes the issue event.

G

It has the whole code inside dependencies service binding the channel name because you can specify during the installation or do it on which slack channel you would like to send it notifications and thanks to those service binding, you don't have to worry about the authorization in the lambdas code. As you can see, it's really hard. It has 110 lines of code. So it's not that much and let's see it in action.

G

When you create an issue on github that is offensive.

G

You should get a notification, that's on slack, but oh yeah it has here, can you see or I? Can you see slack.

B

G

So maybe I first show you the label. When we freshly paid the label cautions offensive, appears and I have to change the screen share. So give me a second okay.

G

You can see that message. There is a author, the title, the body and link of course.

G

So that's it. That's not for me. That's how it works. Maybe you have some questions.

C

Yeah, can you just show me where you're hosting that OCU you've got the connector? It's.

G

Hosted inside c'mon it's it's, the connectors are services written in go so they handle an end point and.

G

It's it just put in in key McCluster.

B

This is more interesting: where do you have this repository with add-ins somewhere published that you can share the link with yeah.

E

B

To enable that scenario in the.

G

B

Area so I think that would be interesting.

G

The atom is on camera, incubator, hack, showcase in folder Adams, and there is an index for connectors and index for scenario. So you can paste it in comma and you try it at home.

B

G

There is all documentation of the connectors, so you can read something about it: how to install them.

A

Okay, any other questions. Okay,.

A

So, thank you very much really nice demo and pretty well explained. So what we can do we can just go further and once again, it's time for our own okada-kun is key to say something about I. Think happy supporting Chima, hello.

D

Again, I can show my screen this time. You should not expect a good demo. Sorry, just not raise expectation, hmm so ask me API to zero before I go into demo and actually show the stuff that we did. I have to give you some context, so you can actually understand whatever I'm showing in the CLI, but first thing for most important. Is that whatever I'm showing you now is it's built on a cluster? The cluster is built on a on the latest master, which means whatever I'm showing is going into the release next week.

D

So here you can see a lambda UI in c'mon, and one of the functionalities that we were proud of in Kim is that it's not only easy to write. Such a function doesn't matter if it's running on cube list now or later on, key native, but it's also the usability how easy it is actually to bind to some service and also to subscribe it to some event and in the UI you do it just like by selecting Lee I trigger even trigger, and then you get a model with list of events, the other cool stuff.

D

About that you can see in the UI is also that somehow magically. We know what events, not only what evens you have, but also you can get a select a particular event and then we generate for you an example payload. So you can test your lambda through from the UI, without really going into your real application to to test your code and those these magic. Things like this drop-down is listing and everything it's all.

D

Thanks to the fact that we adopted in in commodious an API spec, nothing API stag is a specification to document your events that you are sent by your application. So now this this information in the UI it comes from the actually. You have to excuse me for a second.

D

Excuse me um I had to send my kids away from home now I'm back, so you can. This whole functionality comes from massing, API and a long term. The the issue that you can see here is that you can do those things you can have this discovery of events, and example payloads only it's only possible with the UI, because it's handled by the UI back-end we want to change it.

D

But first we wanted to adopt the latest specification because we didn't want to invest in the old specs that will not be adopted by community and to make sure that it all works and people that upgrade c'mon that already use it and they already integrated with kheema, with the old specification I provided in the version, one one zero one, one one two. We wanted to make sure that the introduction of the asking API to is as similar as possible.

D

So that's what I'm going to show you that we provided a special service that is doing conversion of the specification that you registering the in turn into keema. So for you, it's super trans Barham. You just can register whatever spec you want at the you end.

D

We all make sure that it's converted into the latest version, and we also added additional feature that was requested by community, but we also validate this pack, so you can actually from kheema, learn and up front that something is wrong with your specification when you register now, the demo is based on the two mocks that I've registered, so the application excuse me for a second.

D

D

Okay, um because that's an emergency, can we pause for a second. For me, how would continue and I'm back with you in like few minutes I would appreciate it.

D

What a whoosh mijo yeah I'm here yeah. Could you hear me order yeah.

A

D

I have heard I had to drop off for a few minutes and Sony Vaio. Let let me open you if you spoke a no.

A

Problem: okay, yeah: we can switch and then back to you.

A

Okay, the question is how to stop sharing hookers green and they always try to share my screen. Mvc yeah, all.

C

Right, maybe just.

A

A second kicking.

C

Something if you can't view options, you can request remote control.

A

Stripper, hmm that's a tricky thing: I, don't see any okay, but he's like oh gosh. Do you me man, yeah, yeah? Okay, because you you had screen-sharing, so I can already continue. You.

D

A

That's good so.

D

Let's go okay thanks a lot for patients and excuse me so I have those two applications and what I did I registered them, which means I, also provided specs for two of them, but one and one the one for C for C is provided with me as an API 1-0 spec and the one EC mock it's provided with to 0, which is also broken.

D

So just now, let's go to the CLI, which is less nice.

D

D

D

Can you see a terminal?

D

I can see okay awesome. So now, what I'm doing is that I'm I'm checking the details of a cluster Docs topic that was created by the application connector when I was registering the the the application mark.

D

And just you need to remember that the headless CMS components that we invested in this release to make this far because this functionality, at the end, underneath it, creates an cluster asset resource which is part of the asset store. Another component of ours that is taking this back and thanks to a proper configuration knows if he has to do some additional operations with your file before it's stored in the amino in s3, like storage.

D

So you can see that, in case of this EC spec I could see that the cluster Docs topic status was yeah. It was waiting and in the asset cluster asset resource I can see that the specifications that I a specification that I tried to register the status is failed and thanks to the thanks to the service that is integrated in an internally asset store asks um the service to validate and convert. We know that I had an error in my spec and I.

D

Exactly know that there's a fills that is required and I'm missing it and I also know that there's a additional field that should not be there and reason is simply because I broke the spec by changing channels into channel. So we can see that and as a result in the UI.

D

You, when I go into the service catalog and the events service class I, don't see the specs generated and rendered so it I see that there's something wrong. Something is not showing up in the UI and then with the CLI. I can check the details in the resource that what's wrong, with thee with the specification. So that's the this bad scenario but yeah. What's it if it's successful, so an application that integrated with kima and they did it with the specification 1-0 specification, you can see that when I.

D

Stirred it that's the location of the original file and just to prove, let me open it. You can see that it is as an API one zero. You can still see it right, yes, yeah, okay, so you can see that what I did I registered one zero.

D

You and now let me check the tales of the was registered.

D

D

D

D

And the fact that at the end is registered and will be used inside FEMA for various different functionalities, that's asking API to zero and when I go to the UI and I go to the details, this class cloud for customer events I can see that the the list of the events- it's it's visible. All the channels are displayed and all of this this is handled with the new component that we contributed to asking API the react component that is rendering those those channels in V in the UI.

D

So just to summarize, we we created the service to do the conversion and validation, and we also made a number of contributions to the assassin API to make sure that they have a a converter provided with a written and go, and we also have the proper components for the UI.

D

um So that would be all from my side. Do you have any questions to the park? You.

C

Yeah I have a question mm is the is the validator executed as part via in the assess, or is it or was it executed external like? Is there a dependency on the asset store to get the validation? I guess is what I'm asking no.

D

So the the validation, the actual validation is, is done by a specific service. That is a exposing an API where you can ask the service to value the validation or the conversion. What asking a pia what asset store is doing? It's the architecture of the icing store allows us to say: okay in this particular asset before you upload it to the s3 storage, you should call a specific webhook service.

D

Send this back to the service, ask for validation or conversion and then go back into the controller of the icing, the asset store and then send the valid converted file or whatever to the s3. So in to answer a question.

D

The the asset store is a massive component, because the implementation that we did in the headless CMS is using asset store, but the the service itself it can be used without kima, even and it's and it's part of the of the headless CMS component and it's a standalone service that will actually use also in the in the compass to provide the same kind of functionality.

D

So yeah I, don't know more question so just to give you a heads up so, of course, to have nice full-blown user experience at the end. I should already you can imagine that the best would be if I would go here into easy and before binding the application to my namespace. Already in this, you should know that my spec is is wrong. It's not working, but that's that's something we know about it.

D

We want to fix it, but that's again, yeah compass thingy that will address it with a really nice functionality that you can even edit the expect in the UI. But that's going to be part of the compass time on the future.

A

Nice thanks man, so last thing on our I'd end up proposal for upgrading, API versions of community resources currently is a hot topic. So we have Sheree was your proposal proposal.

D

Okay sure my screen.

D

Okay, okay, some time ago, I created some proposal about operating a peer purchase of copernicus because, as you know, or maybe don't know yet with capital as 1.16. The some RP versions that were at the placated since I know, kubernetes 1.9 will be no longer set by API server and will be removed from kubernetes and unfortunately, we use that deprecated versions in karma and we need to somehow upgrade to new IP versions.

D

Unfortunately, the IP version field is immutable, so we cannot just edit it and update resource the resource needs to be recreated and I made. Some investigation I also asked on cabinet as slack how to handle it, but nobody asked for it. I know that Lukasz Kaminski asked on helm, slack and people just say we don't care about.

D

We didn't just recreate the resources, remove the old one, create anyone, but we were thinking about something like the Qunari deployment, but it is ok when you work with deployment stateful set on others, and you can just control the law to yourself ease, but sometimes it is not possible, for example, in case of ingress, because every ingress will have its own IP address and you need to point your domain address later, so that will be complicated and so I decided that the easiest version.

D

This is why, for us, we'll be just like others in the prea brief who removed all the sources and then just in the help, charts create the new resources, and so that will not have a problem with replicated. This also will not in. We not have to tell like a time to investigate which resources can be duplicated. For example, alt control is created with leaders who are not ready to be multiple instances.

D

Only since the computer to zero, it is possible to have multiple instance of controllers, so for that controls it it must have to remove all these stances, as here is the proposal and deadline is probably next Monday, for it, I think and, as I said, I propose to directly create instead of Canary deployment. Any.

A

Questions basically on the pre upgrades hook. You want to always remove that objects or decide.

D

To be removed, if the version created that is already in the cluster is different than a suggested person for that cabernet. This version, for example, you can see the short helpers for deployments and if the version is smaller than 1.9, then we will use up for all dot-111 1 beta 2. But if this new version then we will use, does the new IP operation, apps, apps version 2 and then in private hook. We just need to verify which version we use. If we use old version, then we will remove it.

A

Okay, so it's removed the resources only in case if there is a chance to abate that resource manual, you are right and really not. Okay change it back. Even if you have a new version already then do.

B

We support kubernetes version before 1.9.

B

So it's not possible to install key mother just commandeer, so we do not need this conditional figure. I. Think.

D

But the new version is available since 1.14, okay,.

B

Then that is a case. Yeah I just wanted to point that one mine is yeah.

C

But really an ancient one.

D

But the important to notice here is like emphasized clearly that it means that the upgrade process for the customer will cost down them. So for us first, we're gonna I really have to make sure that we make all the changes in all the components and yet kima as many of them. Let me make sure that we do it in one release and don't cause the downtime for next to three releases. So that's a challenge, but next release we want to because 116 was released recently.

D

So the plan is for next month to prepare a pipeline for checking in to check kima on against 1/16 using client. So once we have an automated pipeline that this it's going to be easier to track, the changes come.

A

Any other questions.

A

Okay, that's very much for saying that, and you said that is the lines for next week right.

D

A

Money, okay, it might be at the.

D

Moment this is in buoy quest the days.

B

A

Like so, take it mind that you need to provide feedback before next Monday. Thank you.

A

So we already a little bit of out of time, but the last thing on our agenda is Q&A. So do you have any questions regarding that represented or done any question about FEMA itself.

D

Then I can jump in quickly. I quickly show my screen quickly. Yeah quickly, I'm super quick, um so just to follow up on Ian's question to show exactly where they were the magic happens. So when we, when application, connects or creates a resource and I told you that at the end we create cluster assets and outclassed acid is calling the service. So he knows it. We we know about it.

D

Thanks to the approach that we have a config map for our had a CMS controller, at least where we can specify that a for a specific type of the source that it's processing, it must ask Li the asset store that when he is processing the for the file and saving to s3, he has to perform some validations and mutations, and we specified to which service from which names based which end point it has to call and ask for for a particular action. So yeah just say for you.

A

Okay, so thank you for attending our Seacole meeting and CI in next two weeks. Thank you. Thank you. Thanks, bye-bye.

C

Thank you, Mike.