Cloud Native Computing Foundation KubeCon + CloudNativeCon China 2018 (Shanghai), 21 Apr 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Deep Dive: SIG Apps - Janet Kuo, Google

Description

Join us for Kubernetes Forums Seoul, Sydney, Bengaluru and Delhi - learn more at kubecon.io

Don't miss KubeCon + CloudNativeCon 2020 events in Amsterdam March 30 - April 2, Shanghai July 28-30 and Boston November 17-20! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects

Deep Dive: SIG Apps - Janet Kuo, Google

SIG Apps is dedicated to all facets of designing, developing, and running applications on Kubernetes. This is a working session and the content will be driven by the interests of the attendees. Topics could involve the workload APIs, tool development, tool interoperability, or any other topic relevant to the space. This is your opportunity to let your ideas help shape the future direction of the community.

To learn more click here: https://sched.co/FuLe

A

Welcome to deep session.

B

A

Start, oh and ask: have you attended stick-up ain't roll yesterday.

A

Supposing you already know what else it up and where it is used? If no, would you please raise your hand.

B

A

Introduce what is sick at.

B

A

Is responsible for workload, a PID measured deployments, they fall sat in kubernetes how to deploy in communities and run your application. All of the relevant api's are taken charge by cigars. What.

C

A

Circus SI g special interest group.

A

Way of organization or rosacea mayo not like there is about in intercession. I.

B

A

That we hold meetings every week also have discussions online or offline. You can write as miles to talk with us to how to prepare any slides for the session. I would like to open the floor for further discussion. Any issues relating to the gaps, local bonitas contribution.

B

A

Development use any questions you're welcome to talk about in this session.

B

A

Use your microphone when raising questions or I, repeat yours.

A

Anyone would like to ask questions.

B

A

Their applications, an important topic, is data recovery in a multi database center and flow distribution.

B

A

B

B

A

Application past layers responsibility to do this. Networking an infrastructure layer are also included, but I haven't seen SiC.

B

B

A

Such as logic cluster, because an application is distributed in multiple data center, that there's only one endpoint exposed. So do you have any plan to extend the scope for second.

B

A

Haven't found a plan to extend the gap scope to multi cluster.

B

A

We have another zip code, not a cluster.

B

A

The discussions are in that application: Sega is dedicated to developers or operators, user experience or the tool that they can directly use all the API. They can't directly use I, don't know whether I answered your question. If you are interested in a cluster.

B

A

You have any questions.

B

A

Can go to another clusters.

B

A

Management itself is a very large scope.

A

The current scope is around one cluster and even only one, there are multiple release management issues. Ok, thank you for your answer.

B

A

Other business.

A

Or we can have a discussion on slick apps next.

B

A

There are some.

B

A

B

A

Founding members of sick apps erase those issue were contributors who are interested in sick up discussion if they have.

B

A

Tools available, they can demo in a meeting and afterwards they can also discuss whether to donate the project to say app or not. The advantage of doing this is that your project will have a higher visibility of visibility and also you can enjoy other people's contribution to your projects.

A

The currency caps, so project is called application. Crd.

B

A

You heard of it application CRD.

B

A

You know what is the Rd.

B

B

A

B

A

Resource definition.

B

A

Kubernetes there were a lot of API, such as demon set deployment. You can also create your own API, for example,.

B

A

An API code machine so by using Yama file or Yama formats, you can set it to API server, saying that I wrote, C or D resource definition of people, my API machine, they PR group. What is the API group after sending a to API server? It will be recognized by API server and Q control command. My tool is a good way to access this API application, C or D is based on this concept.

B

A

It does is that you can define what is they have.

B

A

It's a self-defined API to manage your application. They married.

B

A

B

Does it mean by.

A

Managing the application, if you pull WordPress.

B

A

Will be a web server and database behind.

B

A

You also need the outsiders to be connected to the press, so you need this kind of service.

B

A

The application, C or D- how do you view these three concepts? Conjugate servers still set.

A

If you want to install the application.

B

A

Can control the leads each resource? If you do a web, UI or dashboard? How are going to show the interaction or the relation between the three acts in your UI? You will write which resources does the app.

B

B

A

Replication CRT.

B

A

B

A

When you get this resources, the application icon can be put in the application CRD, and you can show the icon in your UI, where the users can see the admin URL I and they will know who is the maintainer of the app? What is the description of it and if you they would like to be supported, who they resort to? So it's more like a.

B

A

A

Anyone can upload the information relating to this application. The users can also have access to this information. Now the application, C or E.

B

A

The place to put information.

A

There is not a controller.

B

B

A

We're doing some automation, so if you want to do application controller.

B

A

Perform application description or the concept I saw, who are the developers and icons? You can also view its status.

A

You can find the resources concerning all these applications. I'll stay for sec fold, deployment, healthiness and return these status to application status.

B

A

You can show this data to stay in the cube control, o web UI.

B

A

That's what make apps? What do you.

B

B

A

B

A

A package manager of cabinets, perhaps.

B

A

To version two.

A

He's working on the version 3 of help, you also change the home charts templating for that, originally it used to go template later. It also supports other parameters, templating approaches also, it will remove tiller.

B

A

Grade or template.

B

A

Is an RPC server.

B

A

It but it's not a goober Navy's native server.

B

A

Order to upgrade or downgrade.

B

A

A security perspective: it's like a rote server, because it has all the authorities.

B

A

Install all the resources, so all the settings.

B

A

Quite risky in terms of security, so the next version of home it will be really removed in the next version.

B

A

B

A

Crd support could be put in home, 3.

B

A

Abdication CRD, in-home related updates, so much for the updates for them.

B

A

Using web load, APR I do have any concerns or feedbacks. Oh when you use how or other our company these apps.

A

Do you have any questions concerning feature requests.

C

C

C

Time when we realize all that, because all the functions you've mentioned currently are quite simple, not appropriate for application.

B

C

These functions.

B

C

Far as I know, there are now in the better version if you want to make them more stable. Maybe we need to wait for a while until next year, because now increase their of their actual application, for example in Google Cloud, because I have Google cloud.

C

In our cloud there is a UI if you click on the UI, and you want to look at your kubernetes cluster and it will ensure all the apps installed your kubernetes cluster. It is very much like a.

B

C

B

C

Look at and manage your apps in a UI on the kubernetes, but it is not that complicated application so.

A

Not that complicated.

C

Application, why do you position as such, because you know at kubernetes, you have all of those base and for help. There are like those package management tools that now do you have any more complicated concepts to manage these applications. Why do you they make note that complicated application I mean.

C

It is mainly to.

C

Filling the blanks for existing apps. Do you know how the garbage collection of kubernetes run.

B

C

B

C

What for garbage collection, it means.

B

C

Each resource you could appoint their interrelation, for example, deployment will create a replica set and deployment can also assign itself to be the owner of all the replica sets, and this is.

B

B

C

The garbage collector of the communities to delete.

C

Current as well as the kit, at the same time, it sounds kind of strange, but it is like you know, you delete the route and then their children will be deleted.

B

As well and when.

C

You install an app of the resource.

C

All the resources are loosely dispersed, nothing to integrate them, so application CRD is created to solve this problem. Okay, then, for applications, the Rd.

B

C

B

C

Or crate describe everything when creation of an app needs it, but how about managing these applications? Can we control them from the bottom? No.

B

C

Can imagine it to package all the resources bundle all the resources together. It is easy for you to look at them and delete them. But if you want to update the resources, you need to update the bottom resources, not through applications the Rd not to serve up the updating applications Authority because for applications, the Rd. It has no controller.

B

C

B

C

It will know there will be no reaction, okay, so even for the future. If in the application it deploys like five applications, for example, for application service, not storage, but you cannot update it to three or four through applications. The already you need to change the deployment you're right, but if you think this sounds not reasonable or you think you have a better solution, you could.

B

C

And then set your feature request and we can improve it based on your suggestion. Okay, how many people are doing there see gaps now.

B

C

Not sure me about the exact number.

B

C

To the actual number of people, but in C gap the number of people really developing.

C

Api would be like a dozen I.

B

C

Several questions: well, we talked about application, C or D.

B

C

I delete the CRT and the relevant resources will also be cleaned yes, but if I want to create them, I need to create them separately or just a CRT. You need to create them separately, so you mean apply apps in a yellow package, including all the resources. Yes,.

B

C

It would be like an application package with some metadata. Yes, but I noticed there.

B

C

Face and not completed, but will it be updated? Will there be a wall to rhythm or for that face.

B

C

Know need to write a controller yourself and then update the face. For example, now I want to write a tool to install and.

B

C

Of them could be.

B

C

Use of applications they already writing data into. Therefore your UI later but applications, the CRT itself does not include the controller. You need to write a controller by ourselves. So, for example, the install is finished and you need to update all the applications. They are defects, but in the future we may add a application, CRT controller into it and the controller will help you manage some stuff.

B

C

Gatorade she went wrong.

B

C

You mean, if you want to update application and for different applications, they might have different sequences for updating or what I mean any updating strategy between different layers. You mean for the same application for each resource. They need to be updated in different orders. Yes,.

B

C

To make you some concept of the operator to know about operator.

B

C

B

C

Using CRT and controller concepts, actually it is a controller.

B

C

It is different from the deployment controller of their forceps controller from deployment and stateful SATs that they are quite kubernetes oriented. There are.

B

Genetic affinities.

C

Genetic, but for this controller you could define by ourselves what you wanted to do. For example,.

B

C

Cluster and every time we use scale up what you need to do, for example, I needed to.

C

Know and how well I fell over and all the logic would be waiting into the controller.

C

B

C

B

C

Customer resource similar to deployment, but it is self defined by you, for example, it is a CD cluster resource and you can write in it. Your replica sets to just define all the information you want all the parameters that could be considered by.

B

C

Writing, like I, see the resource data is generated and what I should do so in your case.

B

C

App is named, for example,.

B

C

Just let's assume it's a WordPress and now I have this word price and in its back what is written and then maybe in the snacks you could clarify the order. 1 2, number 1, number 2 and your controller could be upgraded according to this order or you could create different resources.

B

C

In our group, we have quite loads of discussion about our own application management internally and whether we should use home for templates or use operator. But while you mentioned home, is going to be transferred to application seniority with its own controller, so it seems like based on template, complement template. Its implementation will be more and more similar to an operator.

C

Did I understand me right. They.

B

C

Kind of different, because for an operator it is more like.

B

C

All the business logic only related to this project to the controller like that is with the controller concept plus.

B

C

Which is highly relevant to this application, but it doesn't mean all things used using a controller is named to come operator well for him.

B

C

B

C

Or specify any parameter which could be configured by.

B

C

Know the parameterize: it is a parameterised way.

C

You may need to do it through hell, but if you use operator, maybe where you want to upgrade, you are just changing the originally sent customer resource just like the way you are changing the deployments back, so there are totally different in terms of the approach, except for the parameterize. The way you could also be overlay, for example, today I have a home.

C

I'm required to change part of the parameters. Some specific parameters should be changed, then I just changed those specific ones, but, for example, now we have t2 teams for my team. We want to change, parameter a and the other team want to change, parameter, B, so parameter a and B should both be all apparent for change.

C

After all, this may be for this helm. The changeable parameters of would be. A lot would be very large and another way to do it is through overlay.

B

C

Is an open source tool: customized, okay, okay,.

B

C

Is the word customized where the first letter C changed to K, so this overlay for this overlay? You don't need to specify in advance which parameters should be changed. It's like you, have a base parameter and in the production scenario you could use the overlay with llamó to assign which field should be changed.

A

A

Use operator to manager.

B

A

The last five minutes.

B

A

Other questions.

A

Focused on different resources, which is managed.

B

By the users themselves, as.

A

I listened to you, the lecture.

A

On all the resources.

B

A

May be some hyper structure, for example, of a service broker well,.

B

Is used to toast one some.

A

Certain party resource, so do we have similar directions? You will consider the development.

B

As a whole, application.

A

Cic can be integrated with open service broker.

A

Very good question we haven't considered it.

B

A

We might as well if it's needed.

B

A

Not aesthetic but anytime, there.

A

There's no strictly related how to do the communication, so they bind to the service to the other application just to transmit some.

B

A

Application CRD is more static. The he was really completely know. What is the status of their application.

B

The application.

A

Said that to a controller so that vacation can be.

A

It's quite useful in one time from applications aspect people would like to see these does, especially in terms of emergency switch, the endpoint, it's quite useful.

B

A

Resource endpoint can be put in the application CID and used it in a convenient way. I think it's a very good idea. Thank you.

B

A

We have time for the last question.

B

A

They'll think of very much for your attention.