Kong Kuma Community, 11 Jan 2023

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Kuma Community Call - January 11, 2023

Description

Agenda:
- New policies: MeshTimeout, MeshRateLimit, MeshHealthCheck
- Preview of the new GUI
- Inter CP API
- Survey
-If you’re a Kuma user in production, please reach out!
-Survey is on how people are using Kuma

A

Hello, everyone welcome to the Kumo Community call. So please add your name to the attendee list. I hope everyone has access to the document right, um yeah and feel free to add any agenda topic you'd like to discuss today.

A

We will cover this in first half of the meeting and then we will leave some time for questions.

A

Our first thing is that we implemented three more policies um and we're working hard towards our shipping Kumo 2.1 in the near future, and we want to have all new.

A

uh All new policy will replace existing policy, but with new Target not replaced, but they will coexist for a while and new policies with the target ref watcher and we implemented mesh timeout measure estimate mesh health check. Unfortunately, there is no documentation, I think for measure it limit. I might be wrong and I saw PR for health check, but still um previews preview of the release is available, so you can download and try these new policies um any question about new policies.

A

If there are no questions, then um I think Philip wants to show us new GUI.

B

Right I'm gonna take over screen sharing. If you let me yep,.

B

Right this is uh in this case kumagui, but all of the things that are new here are also available in Kong Mash GUI, of course, um sort of just a relatively basic demo setup running locally, and we have since last time a lot of like subtle changes and a few more substantial ones.

B

Maybe the most notable since very recently is that the sidebar is quite significantly decluttered now, where we now only find one general policies item combining all the policy views for the individual policy types into one item in the sidebar, and they can now be switched around here using this uh drop down widget um also what we changed is, or rather added that under services for each individual service view. We also now show the um data plane proxies attached to a service.

B

So in this case we just see one because this is not a very big demo setup but yeah all the dpps that are attached possibly to a service would show up here now. So that's the same kind of listing that you would also see for dpps and now also separately for gateways which are kind of powered by the same underlying table.

B

um What we're currently working on what you might be seeing here already is that you'll be soon able to filter dpps by a couple of different things, so, for example, the name of a service of a DPP or their text. So you can do things like filtering for protocol, for example um right. What else did we change I? Think we did some smaller changes here. In this view, the overview of a mesh um any any requests for things that we were working on yeah, oh yeah, right.

B

We have four dpps if you go to individual detail view for DPP and the policies um well, not in this setup.

A

You don't have new photos, yeah.

B

I, don't have new policies here, which is uh inherent to the demo setup that I'm running, but now uh in the inspect API for dpps would lead to a new entry here. That would show new policies here as well. So if you have a mass traffic permission, for example, that will show up here as well, and all of these got a bit of a different look. That is something that we're still iterating on. So this might slightly change in the overall presentation of how data is showing up here.

B

What else did we do?

B

Should I see it? What did I miss? It's Tracy here, I think so right.

B

I think that about covers it, unless somebody has a request for what they want to see.

B

All right, then, I'm gonna stop sharing for now. If you have any questions, feel free Johnston,.

B

Then it's bacteria yeah thanks.

A

Let me share yeah, thank you for the preview. It's very nice update, because I think we tried this when you can go from service page directly to the data plane super convenient and it's very nice flow yeah. Okay, um next thing we wanted to share today, it's uh Intercity apis new thing in Kuma and uh I. Think Jacob can speak more about this.

A

C

Want to share it or I can show them. No, not really, not really I, don't have any demo for this. One.

A

C

Yeah yeah I guess you could, but there are no like cool diagrams there. So, okay, you can as well just open empty blank tab um yeah. So uh some time ago we implemented the functionality of executing um XDS config from the GUI right, and this works well and it worked well and works now in Standalone.

C

But in multi-zone we had a problem when, if you are running multiple instances of global CP, you might execute a request for the XDS config band on a global CP instance, um which is an instant that has no connection to the zone right, because only a leader from given Zone connects to one instance of global CP right, which means that you need to be on a proper instance of a global CP to type this XDS config request to some CP and then to DP right, and that was well a problem. So um to solve this.

C

Unfortunately, it wasn't that super simple. We had to introduce a new API which we called Intercity, which is an API that lets CP instance, communicate with other CP instance right, so only in a given cluster. Only in a given zone or in like multiple instances of global CPU, they can communicate with each other.

C

um The API here is uh basically Health checking to keep a catalog because well, the CP instance needs to know. Where are the other instances right? So um we have this health checking to keep this catalog, and then we have a proper API, which is just executing XDS, config stats and clusters right. So the communication is over grpc.

C

It's TLS well mpls, with a CA that we have to generate so it's secured by default and that helped us solve the problem that while we, if we, if we executing XDS config, request on a global CPE stance that has no connection to a given Zone, then we can just forward the request to a proper instance of global CP, so that this change is um transparent for the end user is just fixes the problem that we that we had and Via this new API, which is kind of internal API.

C

It's not meant to be used by your user and um we may reuse this API in the future to coordinate other stuff. Like you know, we run bunch of things only on leader in global CP cluster right. We may, for example, schedule some jobs on other followers um in and we might use this API for that. But at this moment it's only used for the Envoy I mean operation for RB.

C

And that's it any questions about.

D

C

Okay, no question: thanks.

A

Yeah, thank you, Jacob um yeah, that's about Intercity API and you think, and finally, we'll be able to run uh inspect API on many CP instances and um yeah the next Point you'd like to speak. Today's survey uh I think shortly can share more.

E

Yeah um we, uh we are so one of our goal for 2020 phrase to actually um move from sandbox to incubation.

E

um One of the things we do need for moving to incubation is um to actually get um a list of users that are using kumain production.

E

um This is something we have little visibility about um so first, if you're a Kuma, a user of Kuma in production, please like reach out to me in kumaslak, or something like that, um just to make sure um we're aware of it um we're thinking about an adopters.md file but like it would potentially be better.

E

If, if we start um uh by not having like an empty one straight away um and then and then what we're also going to do is a survey to try to understand a little bit more how people are using Kumar um in what type?

E

What are they looking for, uh what they're finding difficult, what they're finding easy and all that um this survey is going to be advertised in the different um um social media, slack Etc, uh it's going to be as a banner on the website as well, and uh we hoping um uh maybe as well to be able to get it when you do um Kuma dot, IO installer.sh,.

E

um That's it it's just letting people know that this is coming. uh I think this is going to be quite interesting um to to hear a little bit about the different user stories and stuff like that cool and that's it.

A

Thank you any question about survey.

A

I guess not, and we are just at half of this meeting, so we have 15 minutes to questions.

A

um If you have any question about Kumar, oh, please feel free to ask.

A

Hey Avinash, uh how are you doing I see new faces of this uh call so.

B

It's been found out, yeah.

A

So how drink are using Kumon.

F

No not yet I've been now exploring or learning about it, but not yet using it in production, not really having use case yet, but exploring okay.

A

That's cool um any any questions, maybe so far about maybe even basic questions.

F

uh I'll just use the basic things uh just installed it on cluster and run some policies and very fundamental uh things. I don't really have any questions.

A

Yeah cool thanks.

F

There's probably out there, but um I was also interested in contributing to start.

A

Oh yeah I think if you want to contribute- or we are very happy to help with this- uh you can start by uh selecting good first issue yeah. We have issues marked as a good first issue, so you can or comment in any issues for help. If you don't know how to start.

F

Awesome thanks yeah.

E

There's a contributing the MD file as well, yes, directory.

A

Yeah but yeah, it's pretty simple I mean just open here. Nothing complicated.

F

A

Okay, um if we don't have uh anything else to discuss today, we can wrap this up. So um thank you. Everyone have a nice day and see you next. Communico have a good day, bye.

B