Kong Kuma Community, 11 Nov 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Kuma Community Call - November 11, 2020

Description

Kuma hosts official monthly community calls where users and contributors can discuss about any topic and demonstrate use-cases. Interested? You can register for the next Community Call: https://bit.ly/3A46EdD

A

So for next, uh uh so so this is more rest for next week uh we uh actually the the the my next thing that I wanted to to actually share that coupon eu for next year was announced. It's going to be virtual.

A

They kind of how do you see? How do you call this anticipated or.

A

They've foreseen that uh it's going to be it's not going to be okay like by may next year, so just decided to drop the whole idea from what I hear from cncf. Probably 2022 will be the year of the events if uh yeah yeah- let's see, I hope they do, they do one in australia so that you can have the chance to travel there.

A

uh Yeah uh those uh coffer papers is open. That was the the main part to december 13. So if anyone is interested into submitting any joint or even even not joined, I mean uh talks, uh is it called for papers or call for presentations? Should we go presentations? I assume cfp, um but yeah so mid december. They.

A

I remember that I think last year, in san diego, they it was like almost the the the initial deadline was like uh the same day when, when the conference ends- and it was the deadline for the kubecon europe for the armstrong, so this year they at least gave us some time after the conference, because sometimes you know it's.

A

I don't know with virtual event, but with live events. Sometimes you get back from the conference. You have some. You know some thoughts, some remarks, new contacts, new people, you've met and this having some time to think about what you want to do next is interesting.

A

My previous uh adventure, celia might remember this. We had this. You know conference driven development where you actually you develop features for the presentation on the next conference. So you have things to show.

A

A

So I don't know.

B

A

Hey hello, so uh in general I don't know if you are interested in these things is to detox hello, but I mean, if you think, there's some interesting topic around.

A

Meshes and something along like this, that you want to do, um I don't know, maybe maybe we can, we can join join forces for something or I don't know. I mean I'm, not I'm not telling cost in this, because we probably will have separate shots about these things, but uh yeah.

B

Oh, so you mean uh for uh submitting uh uh proposals, uh you are asking.

A

Yeah, no I'm I'm I'm just saying that the next virtual conference for kubecon you will be in roughly six months from now and then the deadline is still mid-december. So we have kind of amount to the deadline, and you know, since kuma is kind of um new project in the cncf landscape and yeah.

A

We would like to see some some some some more talks and proposals at least submit it, uh because I think that this this for this new, like this coupon, that is next week, uh we we didn't, we didn't, we haven't had anyone accepted, and I was on the on the program committee. But yet I was not able to get one so uh probably yeah. I mean, because you know it's kong kuma, then, when kong submits about kumar, then it's not really.

B

Yeah, okay, I'll discuss with you in offline after that.

A

B

A

Course, of course, of course, of course, I'm just talking.

A

So deep, this is uh this is, of course, an open.

A

Like an open topic for for everyone in the in the community, so we would really like that on on next kubecon being it virtual. Yet I would like to have more more presence.

A

Good, so someone.

B

A

Don't see the name here, but.

A

Please add yourselves to the attendees here so that we have a record.

B

I'm sure you can.

A

Okay thanks, so apart from the still virtual conferences that are going to be happening, uh we I have only only only two two other things. uh These are the releases. So last week we released rc2 of 1 0 0, which we believe is kind of.

A

Close to what we want to do- and we are planning like what we want to release as a one- zero, zero j. um So um next week is the plan. So we want to go to kubecon. We are going to to have a boat virtual boat, of course, about kuma at the conference, so we'll take shifts.

A

And uh if someone is interested you can draw by and we can chat, uh I guess the virtual experience of chatting won't be as as nice as being able to see each other face to face, but it is what it is. um So we want to get uh to go to the conference with the with the 100 released. So probably.

A

Sometime monday,.

A

Or maybe late monday, u.s time like pacific, which would be you know europe early on like morning, europe we will be releasing one zero zero. Next week a plugging official release will finally update the dogs. There are a lot of dogs uh pending their.

A

um We just had like high ambitions for this one, and uh it took some more time which put us put us in a not really great state where we have features released in rc, then the docks are a bit strange, but we're planning to to improve this um and uh yeah actually to to not get in this uh situation again.

A

So uh for the changelog, I guess that people interested already checked this out. um We significantly changed the multi-zone deployment flow. Like changed, I mean improved.

A

Let's say um I know that some people get issues with this new workflow, because it's simple, but yet the since the documentation is not really public on the site, it's still in the pr. So, as I said, I hope that this is this. Is I mean like we do our best to be to not get into similar situations again.

A

The performance optimizations, as we discussed last time with tilia they finally are in so we we we're pretty pretty confident with what is shipped now um in this in this series and what will be in one zero: zero, the gj release in terms of performance of the control plane. That's uh at the point. um The communication security has really three great improvements. There, uh jacob was the one working main was the main one working on this, and so um just as you can see, the number of pr's related is uh is like a good amount.

A

A lot of code changed migrated to the new uh workflow, but I believe that we are in the in a situation where.

A

Where, where will we want it to be um locality where load balancing? That was one thing that was uh that was asked a lot in the community charting kind of interactions, uh people trying out kuma. um It's essentially leveraging uh invoice functionality, built-in functionality.

A

We kind of I think that we enabled intina in a really simple way, like uh it's, based on on the already automatically set zone tags, meaning that people have very little to do to actually use it.

A

External services, of course, uh something that has been also asked. uh We still have some final touches to do here before the release.

A

uh Terun thanks for contributing this I'm actually uh today I was doing some tests and I'm taking advantage of this one. So the fact that you can push a lot of uh resources with a single kumakata. That's uh that's nice!.

A

Kafka support for some metrics in parameters.

A

And some other improvements here then no need to go through the whole list. um I like the list of the new things when rc1 and rc2 actually will will be the final. Let's say: change lock that we will have in in one zero zero um very little things like some final touches that we're doing here there for people that are following the prs they probably have seen.

A

A

So I'm doing some something about you know: pagination, we're trying some configuration last minute changes, and mostly mostly things like that, we are trying to make sure that, even if small, but if, uh if it's a breaking change in the ui like the interface environment, variables, whatever we do this before the release.

A

um So with this um my my list, pretty short, of course, this is seeing. um I come out of out of topic, so it's like.

A

We are open to two questions, uh suggestions, remarks.

C

So uh I was, uh I didn't like the changes, which has happened, I mean it has made the configuration really easy, though I had to like struggle to how make it work with the latest and greatest, uh which is their own master, so uh documentation, as I said, uh if it can be made a little better, it will really help.

C

So uh there are like couple of I actually uh had one issue also which I for which I've raised.

B

C

Issue get up issue, so when you do a kuma dp run, it says cp address, but actually it is not cpr. It is actually tp address. Now that you have a new dp server, so might want to change that because it's very confusing for anybody and the there is one suggestion I wanted to uh know from you so when, when we do prometheus when when we are basically scraping from all the dps and putting it to into prometheus, so let's say if we have like two prometheus and uh we want to scale it up.

C

So how do we shard it I mean, will? Will there be a prometheus s, t for both the prometheus or or how? How will the deployment be? uh I think that the issue I mean the issue is, uh there are multiple dp and we want.

B

C

Sure we have redundancy for prometheus resiliency of prometheus, so we want. We divide all the data planes into two and then make sure that one premise is scraping: half of the data planes and the other one is scraping. Half of it is that possible. Now. Is that a valid ask? That is also one more question.

A

Yeah yeah yeah, I'm thinking uh it's also me like a valid task. I mean like you, should be able to scale up, uh of course, uh every component. Now, I'm not sure.

A

um How is this supposed to work? I don't know, does anyone have any any any ideas here.

B

Oh, so the question is possible to scale prometheus yeah all right.

A

And do we have two like if we have two parameters, do we have two two sds per meters, sd container that we built a binary there? Oh.

B

But promises promises d is a side curve, for you know right for me, yeah, I believe. If you escape, it should be twice of it. Now, I'm not really. uh No here.

C

Always what I'll do is I'll.

B

C

Meet your ticket and uh if you can help me on that, so there are like two things that I'm trying to do here. One is make sure that we have resiliency with respect to prometheus and the other thing is we are able to shard all the dp agents or.

B

C

Dpr onward, with respect to number of uh prometheus as we scale, because we don't want the same dp to be scaled, to be straight by uh both the primitives right. So so.

A

uh You're talking universe, all right.

C

Yes, yes, yes, yeah yeah, all right so I'll raise a ticket. Maybe we can discuss more on that.

A

C

Yeah yeah yeah, uh when we uh scrape the mattresses from prometheus, is it possible for us to add extra tags to that some custom tags so that uh we can use the tag later on to build dashboards or maybe segregation or or anything.

A

Actually to inject in the metrics custom tag based on whatever I am not aware of such capability, but that that would be an interesting honesty capability.

B

Are these tags that you don't want to include on the actual um uh kuma object, because I think all all the dp tags should be included in the in the um prometheus labels? Is that not.

A

Okay, does that work for you.

C

Yeah I mean if all the dp tags are automatically taken care. It will be great, so we can add any custom tag right. Any key value.

B

Yeah, I think so.

A

Yeah I mean, uh except for the reserved ones which are you know the well-known. You know: dot io, slash service, you know protocol and this but yeah I mean you invade your own prefix and you should be, should be fine.

C

Okay, I'll I'll test that out. Thank you.

A

So this reminds me uh austin. Is there any any anything that we need to discuss about this? I haven't opened actually since last week, but.

B

Yeah, so I haven't, there's no uh talk, um I I haven't been able to uh get to the xds research, but that's what's next.

A

What else there has been a question uh regarding, but it's what's in an old wish, I don't autist, I don't know if I will find it was.

B

A

Or something maybe terraform was it yes, so the essentially the question boils down to. Is there any recommended way at the moment to manage the different resources like the common policies using github's workflow? If anyone can share their experience here that that probably would be appreciated, I am not aware of something that we are doing on our site. It's essentially, you know. We provide the api and the resources and then.

A

It's up to the users to to decide what's what they want to do with this yeah, okay, hosting I'm just open this for for later.

A

So yeah, if anyone has any any any experience that they can share here, and that would be, that would be great.

C

uh I have a quick question that I wanted to ask so uh when we create a mesh and- um and we enable um mtls in that so for all the meshes which are created on a cp- uh they all share the same root ca or for each measures. We have separate separate roots here.

A

I think that that you, you you're, setting the the provider like when you're going to set up the mesh yeah.

C

So um yeah we are planning to use the provider so just wanted to know, as in uh do we mean do we need to give like multiple, uh cs3 or other roots here, or only one is good enough, like in case of easter.

A

C

To give one and and all the measures are created out of that- is that something similar here.

B

Okay, yeah, if you're using built-in type of provider or provided which one are you using.

C

um We want to use provider or.

B

Provide I mean, if you use provided, then you can uh put a single secret for all of your mashes. I mean it will be single root. You can configure it yourself, but if you're using built-in, for example- and we generated pair mesh, so every mesh will have own yeah nikolai, you can also click on built-in just to make it clear. In this case we will generate a route for each mesh.

C

All right so for built-in we create roots here, for each mesh and for provider. It's just one.

B

For provided it's up to you, you just uh point it to some secret or what ways do we give to configure? Probably.

A

You can put.

B

C

Okay, all right.

A

But, but for built-in, actually you are not doing doing anything, I mean it's automatically generated for you and it will be different. That's what I was saying. So we are almost at the end of the of the meeting, so I will need to switch to another one in a couple of minutes. So if uh if there are no other questions, we can wrap it up.

A

So final call. Okay, then hope to see you next week get our at the virtual boots, we'll make sure to post some links and whatever the on the community channel so yeah.

A

Thank you thank.

B