Cloud Native Computing Foundation App Delivery Special Interest Group, 6 Jan 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: SIG App Delivery Operator Working Group 2021-01-06

Description

SIG App Delivery Operator Working Group 2021-01-06

A

Hello, this is gautam from.

A

B

B

um Hello, you can hear. Oh me, can hear you, okay,.

B

Can you hear me not sure.

A

Yeah yeah, I can hear you you can. Can you hear me.

A

B

B

No, I can't hear you.

A

B

I think now I might hear you hello. Can you hear me now yeah now I can hear you, okay cool.

B

So um let me introduce this this working group meeting because I I really don't think someone else will join and we just call it. Okay,.

A

B

So I'm olmer, I'm one of the active people in this working group.

B

Other is thomas, which generally take this part, um so the working group is about operators. We are a part of sig up delivery.

A

B

A

B

Goal to define what an operator is under the cncf uh concept, so why is an operator in general and then go a little bit deeper in how an operator in kubernetes is built.

B

So I'll share my screen: let's give it just a second.

B

So can you see the white pepper, yeah, okay, great, so basically, what we are doing or aiming to do right now is completing the operator white paper, and the way we need is that we take sections and each time we open a pull request with a couple of sections, give it a week or two and then merge it, and I don't have the github right here, but it's somewhere in the notes.

B

So let me give you a brief introduction about the whitepaper. You can at any time just go in and comment and add or delete stuff and generally what we're doing in this uh meetings is going over and just approving or discussing changes. So just look at the executive.

B

Summary and the introduction, which will both be written um more in the end, so it's kind of wording here and the goal is like I described earlier, and the target audience foundation, which is basically a more generic talk about kubernetes operators, the operator design.

A

A little bit uh to the is I've been uh like actually developing some of the operators using uh uh some of the uh available uh operating generation frameworks. Like cube builder operator sdk, previously generated code using code, generators and client go along. So all of these things. So I one of my questions is more about.

A

So when you talked about like more generalistic approach of what an operator should do and uh uh what in the world of cuban it is how we can define an operator like where can I find more information about those architectural discussions that happen for generic operator versus kubernetes operators?.

B

So this is basically the operator design pattern.

B

Okay, it's a more general approach that where we think uh where we think about a software that have the operational knowledge, what and current is called a controller and it watch some declarative state report, some declarative status and manage.

A

B

Resources, yes, so it's just taking the concept of a controller into a more generic design pattern in the sense that we don't define it in the wording of controller and we're not requiring it to run and extend the kubernetes api.

B

Okay and after we described this part, uh this also describing the characteristic of it. Sorry, for that um there is the kubernetes operator components which is taking the design pattern, but giving it the right names. So here you can see the contour, it watches uh custom resource it reconciles the desired state, just the all the things that queue builder and everyone else gives out of the box.

B

So once again, we talked about the leaderbit controller and custom resource in codeloop, and the second part of talking about uh operators in general are the operator capabilities. So we want to.

A

B

Few capabilities that an operator can say I'm implementing this capability, and this, if you know this capability, you know that I'm doing everything right as it written. So, for example, an operator can have the install or ownership capabilities, and then the operator will know how to provision everything, to check everything and or to take ownership on external resources that were already provisioned and.

A

B

There can be the backup, the scaling, the auto scaling uninstalling.

B

So all this is our kind of capability model that we try to build, and uh the third part of more general operator is the security and risks which is once again we don't want to get inside of how you write the control that controller itself, so the language and all that, of course, that can affect the security and risks, but that's a little bit deeper than what we want to describe here.

B

But, yes, we want to describe how you build trust, how you.

B

Discuss the api informations, you need! How do you monitor the audit log, so the more general thinking and less so that can be good if you're running inside the cluster outside the cluster or maybe even if you run some terraform in a way that it can continuously watch a gita, for example,.

A

Okay, uh another quick question, so this this sig uh also care about the authorization and authentication with controller uh being like a native support for any of such kind of uh authorization. Authentication within the controller itself like is that a whole different chapter.

B

um So in general, we're not approving or designing any operator. It is we're just trying to laying down the foundation of what an operator is and what are the capabilities, but we're not. We will not take part on uh proving or for.

B

Adding or vetting for operators.

A

Okay: okay, okay,.

B

Yeah, that was answering the question.

A

uh One more question is, so: are there any kind of guidelines from the perspective of this fig uh when to opt for an operator, uh for example, an operator can be a stateful set or not uh can operator be scalable in a sense that there could be multiple versions of a controller of a given operator and load balance among them. uh Does this lay out any kind of guidelines or or is that all sort of scope for this.

B

B

B

We will probably need to discuss somewhere around even in the general design pattern. So if this software is should be only one of it or should be able to scale it should it keep state, we should probably go into it. I'm not sure if the documentation goes into it right now, but this is a great part to dive in and add, maybe somewhere in this section.

B

Or in the controller section and yeah, I'm not sure if this is written right now, I'm not sure if there'll be a very hard guideline around it, maybe more in the discussion of okay, you need to understand. You might need to know how to scale it so uh don't keep state. For example, okay,.

B

Sorry very good, so, just on the warning, we are not a sig uh we're a working group under seek up delivery, so we are okay, very specific task and once we might close down and say goodbye.

A

Okay, okay, okay, make make sense, totally make sense, yeah uh yeah I had it. No. I thought this was the entire sig meeting. uh uh So I was asking all kinds of questions. I didn't realize that this is a very uh working. uh A working group uh confined to one particular area of discussion.

B

A

uh Yeah, I think those are the only questions I have regarding this. I actually have gone through this document, uh not very much in detail, but at a high level um yeah. I think this document actually helped me a lot in understanding uh where the actual design comes uh comes for this, the kubernetes operators. I think uh I found the answer now yeah.

A

Thank you thanks for that explanation.

B

A

What usually you do uh like uh once you convene and meet here like what is the outcome uh of every uh working group meeting.

B

uh Yeah, let me just a.

B

B

Okay, so, and generally we go over the document and solve uh comments and addition text and as we try to get a good consensus inside the working group and then go out in a public pr and get consensus in a bigger white. The finish the end result will be a document that we will present in the seek of delivery and get the seek approval for it and from there, I'm not sure. But from there that's the sig responsibility and okay.

A

B

Is an ongoing process, uh not short one and uh with all their uh situation this year, so uh yeah going slowly. uh You are welcome to add comments. As I said before, on every stage you want, you can add multiple people right uh parts, so you can just take part, write it up or even delete it and write it again. uh We also have a slack channel and the cncs slack called sig up delivery operator working group and so also feel free to jump. There ask questions, sorry and walk around.

B

I I think that all the holidays and the situation so people didn't join today. Probably next meeting we'll have more people.

A

Yeah yeah totally make sense yeah. So I should be able to comment. Add any comments uh in this document right.

B

Yeah, I think that comment is open for everyone suggestions. Okay, uh I.

A

B

Just write in the uh channel and I will figure out okay.

A

Sure yeah make sense all right yeah. I think uh that's it from me uh I'll try to put my comments or questions into the channel. I think that is more active than putting comments here in the document.

B

Great thanks, it's great to have you here, we'll have to do more people joining and contribute.

A

All right, thank you. Thank you very much have a nice day.

B

Thanks goodbye cheer.

B