Kubernetes SIG K8s Infra, 7 Dec 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Kubernetes SIG K8s Infra - 20221207

Description

https://bit.ly/sig-k8s-infra-notes

A

Hi um welcome so this is the bi-weekly cancer from meeting and we are December 7th.

A

So as always, this meeting is under the code of conduct and it's going to be published to YouTube after the end of this meeting, so just a friendly reminder to be excellent to each other. In this meeting, um let me share my screen quickly.

A

B

So I'm gonna put the link of the dogs in the chat.

A

So I see new face in this call, so it's like is there like anyone want to introduce himself before we start like first welcome to anyone new in this meeting, and if you want to introduce yourself feel free to ask hi custom am I pronouncing right, I, don't know yeah.

C

A

C

I'm Carson Wade, um my pronouns: are he him and of a principal Community architect at red hat in the open source program office? If.

B

D

Josh burkus is a colleague of.

C

Mine, if you're looking for a comparable person, thanks I'm working with the awkward first project, um it looks like Mark I'm, maybe speaking.

E

From Marcel Hill's.

C

Point later because I think it got to be late for him in Germany. Oh.

F

There he is I'm, sorry, yeah I, just also want to say hello, I'm, a teammate of Carson I'm, the technical counterpart of this initiative and.

D

I put ourselves.

F

On the agenda, so we can go into more detail once we're in on the spot.

A

Okay, welcome to both of you happy to see you in the meeting.

A

E

You can go next.

A

My name is Igor.

E

uh I'm, just my first time in this sig um I work for Shopify infrastructure security, team I'm in Montreal um I met. The significant team at cubecom got very interested in what they do also because, like we're doing for security inference and what's related so I thought it would be interesting to help um so yeah. What else I was a member of bug Treehouse at some point, so I'm already in the kubernetes pretty.

D

G

But yeah but then maybe me uh hi I'm Mario uh from Germany I work as a architect for Cuba, Matic and uh yeah. We are currently talking to the cncf to support um with the with the whole moving to different uh cloud provider strategy and uh happy to listen in and Hope to support you, foreign.

H

Manager here, first time taking part in this meeting, 44 LF and here to listen and learn things.

A

Okay, uh welcome to everyone you're in this call happy to see the industry ISO during kubecon again I hope. We have interesting conversation in the future, but basically any one involvement.

A

Do we have someone else 's interest to do an introduction.

A

Okay, so I think we for people new in this call we doing billing preview of the overall infrastructure. So we have like the report you need to be part of mailing is to get access to that and each meeting we try to basically go over. This report see what's up what went during the last week.

B

A

For this billing review, I'm just gonna go over.

B

Is that uh no that's not accurate, so.

A

Yeah I'm just gonna go over November, basically try to explain what new what's happening currently in infrastructure, so.

A

For new people, this is basically a page expressing different the different gcp projects we use and and the different service we consume and the cost related service. This is I'm just sending a shout out to Justin, because it was the one set up all of this. So thank you again. I'm thankless for this report, so I just want to go over the course for November, so we spend I think 200k monthly, and this is something where current each month for this year, because we spend a lot of money in egress distribution.

A

So we basically blow up the budget like we spend annually more than one million enough just to ensure we can serve the community in terms of Contour images. So this is like this is an overview of basically what looks like in terms of distribution also CI and everything related to the equivalency of infrastructure in general, so, but compared to last time.

A

The one change that occurs like we now use um artifact registry, which is basically the new customer registry proposed by Google, to serve those to save the content images for people new to this car. We we did. We did some we did. We gave two call to talk during kubecon about this. I will try to find the link and put any docs that basically explain so, since we went here with registry dot case.io. We now spend money on artifact registry, because this is an end plan um using h53 to store and serve the container images.

A

So I think that's like the new elements in this building report.

A

I want to act like because, basically, if I go most of the time, if you don't track this billing report weekly, it's difficult to understand, what's really happening here. I don't want to go over debating report because we have other subject. We want to talk about, so we can. If anyone has like a question, we can try to figure out this later in the next meeting and answer to the question really to build uh I, don't think, there's like something specific right now.

B

A

And for on page six for people in interest, we also have the capability to compare spend of the current month basis.

A

Last month, foreign.

A

If there's like increase in costs at some point or basically something went wrong during the infrastructure, we have to investigate. Don't trust the president is here because it's kind of misleading, because we, the C infrastructure, is always evolving people's ad jobs remove jobs. So it's really difficult to use basically see this.

A

Those this data as a problem like oh I, went over, but over budget for a specific project. Is there a problem or not? It's it's like.

A

Something it's difficult to figure out, because we have big CI and it's really difficult to try to understand, what's happening over time. That's why most of the time I trust the monthly report, because that give you an over a general overview to understand, what's happening any specific question about this.

A

I, don't see all the face.

B

So I don't think, there's a question.

A

Let me see chat, okay, I think we can move on to the next subject.

A

I

Hey everyone um yeah, so I have just a PR that I'd like to have reviewed and merged it's just to do with an existing uh pipeline uh Go app that I I wrote a little like, maybe five months ago, or something and the changes basically uh enable the filtering of uh AWS cost Explorer export, which goes to a gcp bucket, um sorry, GCS bucket and then eventually into a bigquery data set uh I.

I

This is uh related or tangential to the the last thing that we talked about with Muhammad to do with how to get out all of the billing data. The I forget what it's called off the top of my head, the uh proposed method of doing that which I was uh preparing for with Muhammad it.

I

uh It doesn't appear to allow for filtering for accounts, which is problematic so uh I have this existing piece of code that works pretty well and will provide the filtered sufficient information to be able to get a billing out of the um I. Think it's four different accounts it's hard coded because it doesn't need to be configured separately, but um this is just the the first step into getting all that data out and making it usable.

A

Where do we export that.

I

uh So currently this is configured to go to uh a bucket in the I. Think it's the Kate's in for II sandbox account. Of course, that's going to be different when we eventually put this into production. If this is the right way to go but uh yep, that's that's where that will it'll go to a GCS bucket.

A

A

I minus one on the ID to basically export AWS data to gcp, because I think we can put that data in S3 and run analysis Beyond using S3.

I

This uh actually.

J

I

Foreign tegrated into the existing way that we do billing reports, which is why it's convenient to be over in gcp. It doesn't have to be it's actually, it's quite trivial to change that, um but uh yeah those are just the current state of it uh and I think it will be quite useful either way up.

A

Okay, I'm I'm gonna refresh this differently, so the the thing is for next year. I would like basically with externalize stuff, not put things inside gcp instead put things out of GSB, because we're going to run out of budget next year. If we don't, if we keep integrate things inside JCP I, think that's my my problem here like now, I want to put everything outside of gcp to give us room to absorb cost.

A

That's the one thing I want to do like that's what I'm saying I I would basically I feel like this. I would prefer to not integrate that with current billing report. To make sure we don't we keep money to absorb the cost, because next year we're gonna have again. The problem relate to cause exertion.

I

A

um We try to redirect so Google tried to do the redirection. So sorry I'll finish like Google. We try with Google to do a redirection to ensure we can minimize traffic for costs for Contour images. It didn't work and they decide to basically revert for technical reason which both us in this situation of next year. We need to make sure the credit we get inside gcp will be used just for that. So next year, Mighty is like trying to move things outside of gcp.

I

Yeah, that's great, uh and let's make that happen, but uh my suggestion changes are separate to that, so that my changes are in this. We're filtering this pipeline to only include those accounts and um changing the way that they're, exported and stored is a different thing which I'm happy to address.

A

Okay, okay I will I will get I will try to review that I'll. Put that in my backlog to before review, because I've Just Seen also in line for review.

I

D

Oh no, it looks like we skipped over the discussion of topics such as redirection, reverted.

A

Oh yeah, uh that's right! So I kind of oh yeah, this one. So so during the last three to six months, there was like a an attempt to withdraw traffic from case.gc.io to to community infrastructure.

A

It happened for maybe I think four weeks and we encoder I think mostly that trigger technical issue inside Google infrastructure, so it was decided to be Reverb like I'm I'm, not sure I I, don't think. I have like all the detail, because there was like I, see some security issue to communicate that information. I don't have that information. We might need to talk to Google us about this, but it was decided to.

A

It was basically decide to revert, so we just keep the traffic will go through Google for a long time until we have the the user and the community using the new registry endpoint.

D

So that back Port PR, those are all dead now right.

A

uh We we try something different.

D

A

D

Consider it dead now right.

A

We try something different, it's done, but we scope the the basically the backbot was to specific um line of codes. Oh.

D

That's the one that uh so the big one I made with 2000 files I can probably close that one right.

A

D

A

So for people inside the context we're trying to basically introduce the new endpoint to older version of kubernetes, and we basically did some git traffic to make that happen and Brian was working on that for maybe three weeks and that got merged. Today, two two or three hours ago.

D

Oh I didn't know that.

A

Yeah yeah I completely forgot to bring you about this.

A

Any any question related to the river something.

D

A

J

I know, there's no redirection but taking place to registered against.io or only the older Reapers.

A

Can you repeat, I think I miss I miss one yeah, sorry.

J

Sorry I just just understand, is no, no reverse um redirection taking place to register okay, another IO at the moment, yeah.

A

No no Resurrection at the moment until maybe gcp trying to do that do that again in the future. I have like no dates or promises about this like because it's like I'm, not a Google, so I don't have information right now about this, and the official communication is like something something got triggered and they had to refer.

A

um I need to maybe find a slight message from Ben or John about this.

E

K

E

Understanding is.

K

My understanding is that uh anyone with a lockdown network configuration which included uh like Google customers, but we also expect to include other people, uh was broken by this. Okay.

K

We detected it first for, like you, know, internal stuff, but yes, it was. It was that previously, uh it was basically like required more network access than we previously had which broke a lot of stuff.

A

Hopefully, hopefully we might basically have a new attempt in the future. I don't have the basically it did for this, assuming Google found a workaround to make it happen, but I we are flying for the moment we just we just we had to reveal.

K

I mean we might I think like how does a user say that they're good with it the easiest way for them to say that they're good with it is to switch to like registry.c.io right so yeah.

A

So if you know a customer accompanying using the end, the duplicate and press tell them to upgrade or switch.

A

Okay, so I'm next.

A

uh I think, like I, think it was last week during sick with this meeting I switch as we stopped doing image promotion2ks.gci.io for the future, because this is now they forget if I want. If you want to say that- and we just try this to make sure mainly, this is a project when I say it's a project, things like cluster API uh jobs, caps and other sub-project, using the convention structure to own the to all stake container image.

A

To do the promotion only to registry.gates.io and there's like a conversation happening with secret is we we try to basically we're trying to find how we're going to make it happen and what's the technical implementation of this?

A

So if you have any opinion about this feel free to go to the conversation, I'll put the link in Docs.

A

I, don't think, there's a specific specific conversation about this. This is just like trying to give an update about the subject.

L

Do you think we need to open an issue or not about this uh through the discussion.

A

um Well, the I think the discussion is already enough to capture basically different opinion about this, like I will put my plus one in the description and say: oh I, think we'll be fine with that, but what I think we need to discard the implication? What are the application of doing this for the community Downstream?

A

That's where I think the conversation need to happen.

A

Is it like? Is it like fine to stop doing promotion for all the release support by the community for, like, let's say, 2023, it's refined to do that, because this is like a direct impact for the for anyone deploying dining and kubernetes.

A

The empathy is most infrastructure, because we, when we went basically when we went when we went in production with new registry, we had people coming and say. Oh, we have security policy in place blocking case.io.

L

So I'm wondering if the gap about distribution should be updated with the deprecation plan for the older that registering.

L

So that that cap has been like massaged over the years to signify many things and I think now that finally, we have the new registry, perhaps one final mutation, capturing the deprecation plan should be because.

A

L

It needs to be captured somewhere so.

A

Yeah yeah yeah you're right about this I think I think we can. We can update the to get. uh Also, if you have the link of the cap, we can put that in the meaning notes: I, don't I, don't remember, I, never, remember the name. The.

L

I'll, add it to the notes. Give me a.

A

Second, yeah I think we we can try to introduce some kind of a deprecation policy for the for kids.gci.io and say bye, I would say end of maybe 129 or something like that. We we say we cannot guarantee.

A

We will push new container image to that registry.

A

Because there's going to be some neurons about the application, deprecation means the registry still exists, but we don't push any new image to it.

L

Yeah I I well, some of the conversations have been involved with around this topic suggests much earlier stuff of new text, so yeah as early as you know, January or so.

A

Yeah, because there's the the impact is on two aspects, the official artifact I would say. Basically the five uh kubernetes component will be impact. How we handle that this is the suppression, I mean we can say for kubernetes when I say kubernetes the five component. We don't deprecate right now, but for the sub project. We we basically do an announcement to all the six and say oh yeah. If you basically want to do promotion and cut new version of your project, also project of tooling, you have currently stop push to kids.gci.io I.

A

Think that's my main target, so project for kubernetes itself. I think it's fine to have a different application policy, but for the sub-project we should do something early good things. Ricardo is in the call, because Ingress is like one target as one of the most popular project use well, I mean those kind of projects are like the one thing I'm targeting right now.

A

I'm up to. We can delete the repo as.

K

Well we'll say: yeah: there is no replay anymore I'm.

L

B

L

Love your fan base. Don't you I do.

A

A

This yeah so for people new in the car. This is really to read this process in general. So if you have any question, go to secrety's Channel and if you have, if you are curious about how the release process happening, we I think we have trusting in the call as a former release team member. So.

K

A

Don't think I was.

K

Ever on the race team, actually so.

A

Yeah also both yeah sorry, we have just seen some Barbara and Justin Garrison, so I completely forget.

K

Sorry wrong: Justin apologies, hi, Justin,.

A

uh Any question about this: okay.

A

Okay, I'm gonna I'm, also next uh I'm gonna just give a cute quick update about this. So we we had a conversation with fastly last week as a follow-up of a request to basically we apply to the open source program I think like six months ago, and we finally got an approval pending contracts, signature between fastly and CFCF about I would say: support from vastly to distribute some artifact through the network.

A

So that's like my general update about this in general, so we request 60 terabyte for egress monthly and a few things like. So we will get 60 terabyte of figure tracking monthly for a year because basically they have to review the contract every every year with any project they deal with, but I don't expect we migrate away from them for a long time, because they've been serving uh offenses version for a very long time now, so we will get General support.

A

Basically, I put this in quote because it you have to check the superb page to do that. I need to put that somewhere. We get also traffic encryption, so mostly five TLA set figure. We can apply and some assistance from the developer, the devil team.

A

They are not here to talk about this so, but we will see them in the future. We might because my first goal is to move dl.k Style to like shell, that way fastly, because currently it's like inside Google, it's more like Netflix, true inside Google. So the idea is to basically move that in Sacramento of a structure and shield the binaries with fasting.

A

So that's like my update and Justin. You have your end race.

K

Yeah, yes, I pretty much! That's great news uh for those of us that can't do the math I can't do the math. uh How much is our download currently in terabytes, as opposed to in dollars,.

A

We I don't know because that bucket is inside Google, so you I think you Ma, you might be the best person. Okay,.

K

Like do, we know ballpark I'm, just like I guess: I can guess it based on like if we assume I don't know like 10 cents, a gigabyte I'm just trying to do the math.

A

So the I think last year, I basically talked about this with a googler and what I got is like it's massive and we're just gonna just migrate, because it's gonna blow up a Bridget for in like three months.

K

I guess: do we think 60 terabytes is enough because I guess like what I would my question boils down to you.

A

um If that's something, if it's not enough, we can reach out fastly to increase bandwidth, so they gave us that flexibility like we talk about this, like as if so because this endpoint is kind of scary into a bandwidth. I want to start with fasting and see what's happening. If it's not enough, we can reach out to them and increase the threshold adding that like, and we will get access to this metric so I'm pretty sure that also will get access to Metro and see the traffic um I.

A

Don't really expect a problem compared to the other open source project. This report, we I, don't I, don't think we triggered that much in terms of bandwidth, so I'm not really worried about that.

A

So the migration need to be discussed. I will open an issue about this and basically trying to put where are the technical requirement to achieve that migration.

A

A

A

So I think we can go to the next point which is operate first, so anyone need to share this screen or we just give us okay, I.

F

So I am prepared anything um actually I just put it this morning. My morning, it's now my evening into this agenda. You can follow the link. It has a nice web page, but it basically boils down to it's in Upstream initiative to open up Cloud operations and it's basically a concept. uh What needs to be done to to operations to do something similar that open source, the two software? So it's it's like this mental mind, shift or a concept, but the comprehending to that.

F

We also built out a open, Community cloud which is hybrid across um bare metal and Cloud providers to give basically kubernetes and and other resources to community projects. So you can just sign up via a GitHub account and get your namespaces in a community operated.

F

Kubernetes environments, we offer sorts of services like are your CD to deploy your applications.

F

Monitoring services like rafana and Prometheus and locally and other leading air quotes Leading Edge stuff, and it's fully backed by a github's principle, and we also use brow to do our cicd stuff and I've tacked on pipelines. So that's already a not at the size of kubernetes, obviously but um a small size, multi-cloud setup. And since we heard from this initiative the second part to move away from being too reliant on Google Cloud infrastructure and share the share load and the funding across multiple Cloud providers.

F

I came up with the idea and that's just a crazy idea. I don't know if that really lands with you and whether it actually makes sense because I'm not too familiar with and how kubernetes and testing works. But how about?

F

um We have a brow instance that is deployed on multiple openshift, slash kubernetes clusters um on at least two Cloud providers and what I think what red hat could bring to the table is one of these managed offerings like if we use AWS Cloud credits and Azure Cloud credits and Google Cloud credits and use them to spin up resources and then deploy.

F

Let them managed openshift deployed on these resources and then basically get a well managed kubernetes environment, where the community, it's like infra, can deploy their brow worker nodes or the brow control plane. On top of this would relieve the community, the kubernetes community, from maintaining the underlying infrastructure, because that's done by the managed openshift offering um and we can focus on Distributing the prowl workloads.

F

And I'll stop here.

A

Any anyone has a question comments. Concern about the proposal. I prefer to leave people having questions before I before I, give comments.

G

G

Was the one question? Would this basically mean that the um so that another company or a company is basically providing as a diminished open shift? Is then the underlying part of um of the um of the yeah distribution for the workload, so that, basically, this is then managing the Clusters?

G

um What does not eventually bring us to the same problem that we currently have um when we then say it is so much cost that is created on this uh managed site um that we are again in the same problem that we are now with Google.

G

Because it's not self-hosted on all of the clouds on all of the cloud providers, it's rather it's consuming one of it's consuming an infrastructure service beneath it so and if the company then says so, if Redhead Says, ah now it's getting too expensive move away.

F

Well so so I mean the I'm, not really insane I mean I'm in the opposite of the city. Also I cannot um talk to how credits I spend or how money is flowing through the system, but as from a Layman's perspective, I understand it as if I'm, a customer I'm bringing my AWS cost, which I'm giving to AWS to spin up ec2 resources and then I'm, paying redheads to manage openshift to manage the openshift Clusters, plus the subscription fees.

F

um So in essence, you're spending for the CPU your dollars are going to AWS and for the management layer. It's going to wrap up from a customer perspective, I'm getting a kubernetes API, kubernetes platform plus or the openshifts layering on top, which is essentially a nice UI and some other stuff and sales would help me for that saying that. But it feels like kubernetes and you can deploy brow on top of it. So I think from a cost spending perspective it.

F

It makes it way easier to spread the cost across multiple clouds across all the um Cloud providers that are supported by this, um and we don't really depend on anything open shift specific. It's just let's deploy brow there and that's where, where I might be too naive, so I don't know how the um sick I'm not sick tests, but there's a test infra test um I mean how the CI jobs are handled, whether that's more involved than prow and spinning up some testosters.

A

By the way, I saw a comment in in the chat about verification of like this is. This is not exactly free resources, just to be clear about what what's been seeing you it's more like an offer for an open source project to boost trap resources, but underlying costs is covered by where that uh I think you basically manage from my make sense. But I'm not sure we are the right thing to talk about this because we have seek testing So. Currently proud is a managed service for us.

A

It's like it's a money service provided by Google for us and that it's gonna be like that for probably two to three years now so I think I think Opera first makes sense. If we basically, if at some point we say, Okay Google will not or will stop make that offer, and the community need to own that.

A

But it's not for it's not like a short-term plan, because One requirement to move from from Google infrastructure to the community infrastructure is to make sure all the CI is owned by the community, which is not the case, because we I think we we work on a migration plan for profile like now two years, so we identify what are the technical Gap to ensure the migration, and the first thing is like make sure we have a good distribution of the CI between the different cloud provider like onboard AWS, not not on Broad, because we already use Azure, AWS and gcp.

A

The three major cloud provider right now, but we need to be more I would say more transparent about this, because that operation management is not clear at the moment. So I think the the one thing that basically blocked us to say: let's use what you're proposing is like the CI itself. That's like one broker currently and now aren't we doing we do that, we can see. Oh can we do pro.

A

Can we see that's great because currently I would say the CI Services the project use it's a man is of it's like a free offer from Google. For us we don't. We don't pay for that. They're distributed infrastructure and the people operating that so I'm, not sure.

A

It's interesting to do a POC, but short term I I'm, not sure I can give you that answer. That's what I'm saying I will defer to suggesting about this yeah from basically I would say in the scope of the community infrastructure. I would say all sick gets, infra I would say yeah dude. Basically, we need to finish move make sure the CIA is fully owned by the community.

A

I think that's like the first thing. We need to do first because we talked to someone right at about I know internally, where I did the proud migration and we we got feedback from that and one for the first suggestion was to basically fully on the CI and we can move pro control plane when this is done.

A

So I'm I'm not sure if uh I have the right answer for you right now, uh I, don't I, don't I, don't think we can do a POC for next year, but that's what I'm saying I don't this is my personal take on this talk to six days thing about this.

F

Okay and I brought up um brow and CI because we had a an email threat with Benjamin Benjamin yeah.

A

Benjamin is currently C chair for sync testing, so like I.

H

A

For a person is part of schizophren, but he's also sector of this acting so talk to six days. Think about how we want to learn, and the second aspect is like basically manage CR resources. We we already do that with protocol Cube test. They are basically that project gives the possibility to bootstrap kubernetes environment, regardless of the cloud provider.

A

I think we support I, think three to four cloud provider at the moment and that give us that flexibility to literally operate anything we want I, know Azure work on their own and plugin, just in the one. Maintaining the AWS per game for Justin Bieber is the one maintained, AWS plugin for cube test too.

A

So we in terms of like integration between brown and the cloud provider. This is kind of cover. Now it's up to each stick to basically decide because we also have like Sig and I know. Sub project also use third-party CI services to basically test I, think cap spray use or something else like of also cluster API, for example, use GitHub action to delete Twitter. So we I think our world is to basically provide the infrastructure and, at some point, make sure there's an integration between a third party service and us regarding the security impact.

A

But it's kind of its sake is a little responsible for True like that's. Basically, we trust the thing to behave about CI and not inter and basically do security not introduce security issues. Why that's, how I would say in general, we hope we operate, but the idea is interesting. I mean for I, really see value for a pressure interest to have power and not always want to maintain it because price kind of difficult to operate at some point so I say as a value of providing marriage a manager for this.

A

The thing is the current context.

A

Don't allow us to move to what you're proposing easily- or you know, I would say like next month, but if I want to bootstrap like a new open source project and say oh I would need CI and I don't want to get action. I would go to operate first, like that's like there's values worth doing that.

F

Carson also has this handle.

A

C

Yeah thanks I wanted to I mean Marcel spoke to some agrees, but you know that it was an email exchange with Ben about um what, where, where might be, some places that we could be looking and looking to be able to help and listed two items, and one of them um seems like it's been registry work that's been going on and then the CI and so I guess he was looking at that with the with the Sig testing perspective, um but but I think what the context I just wanted to give was that that this was just oh okay, we heard about that, took a look and saw what we saw in Marcel.

C

You know kind of bring uh something to propose, but but this this proof of concept idea is kind of a blank. It could be a it's um it's a blank line. If there's like, we don't understand enough about the infrastructure to make guesses about how we could be helpful in this space and what we could be doing um and coming and doing so and coming into the meeting to try to get that I mean I'm I'm already getting like just try.

C

You know keeping up with the various ways that prowl is being used and the CI is going across I'm like okay. You know, there's a lot of different stuff to try to get connection with in here and so I'm not sure what would be the best way, but maybe if there's something, if there's some way in which we could um uh get together with one or two people who know the infrastructure.

C

You know these different pieces, or it's just maybe just waiting a week and having conversation with Ben about what specifically I mean something like that, so that we could try to tune something into a a a you know, a proof of concept that would actually show something useful to where you want to want to get to, um because it's I think it's I want to just even going back to Myers comment at the beginning, I think it's important.

C

It's like we want to be finding a way to be helpful and moving forward to um to removing the dependencies and that the community has upon single vendors or anything else like that or wherever it is that you want to be moving forward. So we need help understand the problem space better. Basically so I think that's really the context. I want to say so it's you know.

C

Don't it's not just this one thing that we came to propose and like okay thanks give us a call later when you're ready for it it's more like hey what else? Okay, if that's not the idea, then what else could be the idea- and we don't need to brainstorm on that in this meeting right now, but but Marcel and and a few other folks, myself would be happy I'm trying to get together and talk about those things. Sometimes. So that's that's. The open proposal.

A

That that, basically that's what I'm saying like it's way better to. Basically, if you talk to see texting folks because they're like they're, they are the one making the decision about life cycle of throughout and said I would say to CI.

A

That's what I'm saying if you basically want to have like a conversation about trying to understand how we can integrate what you're proposing in the CI the other right person, because we are doing the line layer, we basically provide call call I, would say I.T resources to the rest of the community, but because we are a somehow a commodity for the community. We talk to a lot of people and there's this confusion. We had the right person to answer on specific question.

A

That's what I'm saying this is like my take on this, but I will not tell you hey, you should not do any is I'm preferred differ to sick testing because they are the one ultimately make the decision about brow and everything relate to Brown because they're like older, also integration program. We need to think about and I'm not sure, I'm the right person to talk about this.

L

C

Yeah and then to find out, if there's any anything else that we could be doing out aside from power, it would be um yeah.

A

We can we can. We can have different conversations about this, our black, no problem, having a lot of conversation about this any question for our prefers. Oh.

F

Muscle yeah, so um as far as I could follow the previous recording notes and the MC infra yep Mark's markdown file on on GitHub and kubernetes web pages. um I think we're mostly concerned about the registry and outbound traffic and storage and distribution of assets.

F

Yeah um is there anything that was besides CI and CI I only picked because we're also using that- and that was the most obvious example and use case for me for having the same interface across multiple Cloud providers and therefore having a some cost management and cost distribution, basically for free out of the box, because you're only talking to the same management interface that and you can distribute that. So if you're saying it's like infra is providing core I.T services to the kubernetes community. That's how I understood it.

F

If any of these are also distributed across multiple kubernetes instances which provide these I.T services and you could and you're federating them behind another layer- and it would have multiple of these kubernetes clusters acting in concert behind that Federation layer, maybe for other binary distribution or um if somebody's downloading something or web page you're, also serving stuff which is not necessarily going through a um yeah pop man or a Docker or what whatever command, which is doing something behind the scenes.

F

B

A

Oh boy, okay, this is this is a good one, because I would say yes, but you need more input from other people because we have C contracts and six entreprints will say yes or no off. Basically, I think watch your so just maybe I'm just expressing what you're trying to say, basically you're trying to basically automate a way to provide to to provide IIT services to the community. Like one Unified, I would say: user experience to bootstrap specific services.

A

Oh I want a DNS I just do some click UPS over I would say whatever setup and you get a DNS and try. You want a new cluster. These kind of things I think it made sense. I know other projects. Don't do that foreign, but, like I said again, you have ordered we just we are plugged to other people about this. I just cannot just say, oh yeah, so this is sick contracts running that.

A

So I see value on doing that, and maybe we should try to do that like basically, if we don't have like Pro, we basically say: oh, we have something. We can set up record recurrent services like DNS, how we simplify the contributor experience to get a DNS entry under case.io.

A

How we do that, is it doable? How we can look into this? Oh, how we simplify getting AWS account. I think it makes sense, but you need some kind of input for also other Forks, so I'm happy to do the plugin between the folks, I, don't have a problem and we can try to think about POC to to for Aubrey. First I think we can try with something like DNS as a service and see what's happening.

F

And I wasn't wasn't asking to Black operate first, the concept there just from a sick, infra point of view, whether you are also looking into migrating any of the sick, infra Services out of Google or across multiple Cloud providers, um where this model of having a managed, kubernetes, slash openshift offering would make sense in that context, and.

B

F

We're already on the top of the hour, maybe we postpone this discussion.

A

uh The core service: don't I, don't want to see migraine more like what we want to migrate, slow, artifacts NCI. That's like the first thing. We need to do first and maybe later core services and say: oh, we can deploy something, but the main goal right now and I think for the next six months is to ensure we can migrate. Ci and artifacts I. Think that's the first goal. I want to achieve first and later think about how we can read how we see I.T Services inside the community.

A

That's why I'm saying like if we want, if we want to do a POC about how we propose a managed Services inside the community, we can try with something like DNS or service and see what's happening. Why.

C

A

Really it's a really interesting.

C

Topic too about just thinking about the you know: the it for communities. Are it's it yeah, um there's a share. There's a shared Journey going on as well as as that's part of what operate first is providing for its Community sense. So you know there's some shared learnings and things like that. I would guess too, along with just the technical bits.

A

Okay, uh we out of time by two minutes- I'm, really sorry about this. This was like an interesting conversation between I'm, really sorry for people in the call, so I I think we. If anyone has a question, feel free to drop it on slack. I will try to answer like so.

A

I would say, have a good rest of the day and see you maybe next meeting bye.

I

Thanks everyone bye thanks.