Layer5 Meshery Development, 14 Sep 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Meshery Dev Meeting (Sept. 14th, 2022)

Description

Meshery Development Meeting (September 14th, 2022)

Agenda:
[Cyrine] [Helm] Restructure Meshery and Meshery operator Helm Charts #6212
Two commands to install Meshery - not quite as nice as a single command, but…
[Ashish]Demo operator installation/uninstallation #214
[Kanika] [UX] Update page design to incorporate MeshMark
[Abhi] RJSF Form refinement for Object-field-templates.
[Abhi] eslint Fixtures in Meshery UI [issue]
[Abhi] Meshery Extension e2e Tests
[Alpha] Removal of k8s manifests in “system update” #6197
[Yash/Gaurav] Meshery v0.7.0 Planning
Refactoring Meshery UI: meshery-ui-restructuring | How to contribute
Volunteer Needed: v0.7.0 Roadmap feature champion assignment refresh
[Gaurav] Fixed E2E Mehsery UI tests PR
Call for volunteers for Cypress E2E in Meshery. [Docs]
Filters Tests are disabled temporarily. [Issue]
Volunteer for #6215 ( good-first-issue)
[Lee] Multi-cluster: Allow users to specify Meshery Broker endpoint manually.
[Lee] Operational Modal UX design review.
[Lee] Kubernetes Connection States design spec review.
In Meshery v0.6.0, connections that Meshery established to Kubernetes clusters, Prometheus, and Grafana are automatically discovered, and registered for use.
Looking ahead to v0.7.0, the following finite state machine is proposed to track different states of Kubernetes connections, specifically, and allow users more control over whether the discovered infrastructure is to be managed or not (registered for use or not).
[Subedy] Discussion on Integration test failing
[Uzair] Fix inconsistency between K8s context switcher and settings page #6205
[Pankaj] Unable to reproduce issue #5850
[Pranav] Update dictionary in Meshkit PR#213

Join the community at https://layer5.io/community

Find Meshery on:
GitHub: https://github.com/meshery
Twitter: https://twitter.com/mesheryio
LinkedIn: https://www.linkedin.com/showcase/meshery/
Docker Hub: https://hub.docker.com/u/layer5/

A

Hello uh yeah uh good morning or good evening, based on your time zone to everyone uh today, uh yeah welcome you all in the mystery development meeting on wednesday. So if you forget to uh put your name there, just like me uh put your name here so that we know that you are you. You have joined the column other than that uh we have a tradition like long tradition, to uh get to know some newcomers uh joining the call.

A

So I think alex uh has joined the call for the first time if I'm not wrong right, alex.

A

No, this is not my first time, uh okay, someone else who has joined this call for the first time like I I keep forgetting the name. So please, you know, keep forgiving me for.

A

So okay seems, like uh everyone is, you know old, timer here right.

A

So so, let's get started with our first agenda item uh in the list.

B

I wonder if um um asutas has had a chance to introduce on this call again. Mr panda, I can't remember, have you got a chance to say hi to everybody on this call.

A

Probably he has some issues with his mind properly.

A

Okay, I will get back to you once your mic is, you know completely set up. Meanwhile, you can write your introduction like who you are in the chat.

A

So let's get rolling uh with our agenda item and the first item is by syrian restructuring machinery and measuring operator, helm, charts yeah, it's very easy.

C

Hello, so I'll start I'll start by sharing my screen.

C

Okay, so last time uh we need to discuss the measuring.

C

We found that, uh thanks to trying to delete the the charts that are uh that exists in both places are replicated, will kind of, in all cases leads in a conflict because of the release names mastery mastery operator, so the only so the best way to deal with that was to completely separate mastery and bash.

C

I think I'm gonna stop the video for me so that doesn't freeze so easily okay, showing my screen again.

C

Okay, okay, okay, so this is the new pr that are made with ashish to restructure the helm, charts and make the changes that are necessary in measuring ctl and measuring server to in order for the operator and measuring charts to be completely separated.

C

So is I'm gonna? Do a small demo demo for the new set of charts?

C

What I did was I previously mastery charts would depend on the operator chart so took that off and and so now to deploy them. We have to deploy both separately. So the way to do that is by doing helm, install measuring separately and then mercury operator separately too.

C

So this is the release uh that you get when you only install measuring.

C

This namespace, so, as you can see, we have only measure installed and these are the dependencies that mercury has.

C

So so one thing that I have to do on my pc so that I can show you the machine interface is, I have to kind of configure the new ip that the message server has in engines.

C

This is something that only I have to do because of my setup.

C

Copy this here.

C

And then put it in engines.

C

And genuine, that's that might be interested in how that might be using the vm too and interested in how to do it in order to access uh mastery using mini cube.

C

You can kind of take a look at my setup.

C

Have you kind of have to do uh configuration for websocket and for mercury on its own separately?

C

So I I'll restart the endings. Now.

C

And there we go now we're gonna.

C

Okay, so just maybe wait for a few seconds to detect the cluster.

C

C

And here in the cluster settings, since I've only just deployed mercury, we don't really have the mystery operator yet.

C

Is it lacking a bit.

A

B

But just to add some questions or ask some questions. While this is going so serene the so we have two separate helm releases, helm two separate sets of helm charts. Today the proposal is to essentially start trying to de-duplicate them um today, you're able to install all of mastery, including the operator, using a single command um with this change, the user, if they're using helm to install they'll need to invoke to home commands or they'll need to do two helm deployments, not a huge deal, a slight step back.

C

Yeah yeah, because well after discussing it with ashish the release name when you deploy them both together, both operator and machine, would have the release name and the way it's currently works in server and measuring ctl.

C

It will key when uninstalling it will keep looking for mercury operator release name instead, and so that's that would create conflict so now to kind of uninstall. The operator just like he said, have to install it on its own.

C

And that work we get both release names.

C

So they have both that are separated like this.

C

Right now, I'm kind of showing everything as a helm demo, but later on ashish, he did a few more commits and can show you how how it works once it's well integrated with uh with the hv server.

C

So here we go. The operator is active.

C

To on solitude, you can have to uninstall it on its own.

B

So there is there: if you just install the measuring helm chart, you will not get the mystery operator.

C

Yes, they're both separated now completely.

A

The only reason to do that is because helm does not allow to selectively just delete the sub charts. If it does, then we can have that thing, because if it it doesn't, then so we can't uh okay. Matt has some question or maybe some position, but.

D

um Yeah yeah uh have have you looked. Have you guys looked at the prometheus operator before um uh it's an example of of something that the core os prometheus operator? Rather it brings in a bunch of other helm charts. So if it's the case that when you install the helm operator, then I'm sorry mystery operator that measures should always be there right. If it's basically a dependency, then you could have the measuring operator chart reference. The measuring chart.

A

Yeah, actually it looks like that way because of the naming, but it's it's not.

D

Like diamond had a chance to look at the pr, but um I was I was kind of is: would we expect the user to do both or you know like I could see like I installed the operator and now I'm done right. I make crds and everything is just handled for me. I never really actually install measuring and the operator as a user right it's either or where do I have that? Actually.

A

Yeah, so actually measuring is not like it's not a operator thing, so it's not completely kubernetes native it operator acts as an extension and operator has two resources under it, which is mesh sync and the nats broker that we use, so messy server can be kept completely outside of the cluster, not managed by the operator at all. So it it's not uh technically, they are not dependencies of each other. They can't completely exist.

D

But but you could run the server as a crd as well.

A

D

A

D

I see so that's the case where you're actually running the server outside cluster. um I see.

B

Or even another way of another way of characterizing, the interrelationships between the components and kubernetes is to say that um for any kubernetes cluster under measures management should have the operator deployed to it.

B

um Only one or two or some number less should have measures server and its adapters deployed to it.

B

So the statement is for any kubernetes cluster, always deploy mesh reoperator and then sometimes deploy mesh reserver.

B

So there there is a there. Is you know to matt's point like yeah this thing? I I can't tell you how many times I've had this conversation and I've never really read about helm, myself and so yeah. I mean to matt's point like about the example of the prometheus operator.

B

It just it just seems like it. Just feels like and helm may be limited in this way, and but it feels like you would be able to have a single release that has a universal component. The operator that's always installed every time and then um and then a values or that and then occasionally sometimes you want to install sub char charts, and sometimes you don't.

D

So that's how it works. Basically, the the configuration that you send the values dot. Indian values.yaml ends up being sort of a hierarchical aggregate. Composite pick your noun, um it's a blob of configuration. So, for example, you know that and then there are crds associated with instances of those sub problem charts. For example, the graphana chart alert manager. You know um you know prometheus itself.

D

um Various service monitors, alert rules like all of these things then are sort of like. If the crds are there, then they get instantiated. So if there's a server in this case, if there's a server uh component, then there'll be a server crd instance. But if it's not there, then it simply won't be made right. And then, if you want and.

B

C

D

It's using the measuring helm chart under the covers, potentially um at least to install the controllers and stuff.

B

D

There's other ways to package it up, but the helmet you write is kind of limited, so you kind of have to scotch tape and bubblegum it. I can grab a link to the prometheus operator if you want to see an example of how they do this, which may or may not be all but.

A

Yeah and to least point that uh yeah I know it feels like. If you have an operator you can, you should have measuring under the subchart, but the problem comes when you try to uninstall the operator and if operator has meshi server as a subchart, it will try to go ahead and uninstall mesherie as well, and if so, basically we run into problems, because the uh the flow is not that smooth the installation and installation it has. The permutation com.

A

Combinations are too much varying and helm is a little constrained as to how much I have read the docs of help, if someone so basically, if someone knows a way to delete a sub chart uh uh selectively, so in a single release, if we have some the dependency charts and if there is a way to just delete those dependency charts and not delete the parent chart, then we can have a single chart. Otherwise, we'll have to move with this.

B

I don't think that this particular requirement has been lifted or I think part of the um part of the bounds that ashish and serene and others that have looked at this, that they've operated within has been true to the statement that ashish just made, which is that that there's a desire to be able to like remove a mesh free server but leave the operator residual or potentially to remove the operator and leave measuring server residual one interchangeably.

B

It could be that you know we, you might be able to make this statement, which is, if there's a cluster that has mesh reservoir deployed and therefore also the meshi operator, if you're going to remove measuring server and uninstall uninstall it as a subchart that you are forced to also uninstall the operator from the cluster.

B

That doesn't you know when you say it like that, or at least when I hear it like that, it doesn't sound too bad. It's not as flexible as you might like, but gosh. If you're, if you're removing the server, then most of the time, wouldn't you want to remove the operator not all the time there are use cases where you you don't, and I think that that's what what we've been trying to achieve and then you know hitting the wall against um helm's sophistication.

B

So, but I don't know, I don't know if it's worth exploring whether or not it's really necessary to be able to allow the user that much flexibility to say if you're going to remove the server that you're forced to remove the operator as well.

A

Okay, so if you package it inside so let's say someone deployed machinery and message server deploy in that release. We deployed the operator and let's say that now someone wants to undeploy the operator. They can't use the meshline chart because that will undeploy mesh three and they can't use them actually operator helm chart because that's under a different uh because you, you are installed it via different release, name and now you're trying there is again. Will there again will be a conflict? I have.

A

I tried a bunch of permutation combinations like that, like the way you are going through. All these different combinations I tried in all in all of these are always some kind of release, name conflict. That comes so. Let me just demo real, quick, my changes so that.

A

Okay, so my mystery server has started and uh inside I don't have anything in my meshing, name space, so I'll go I'll, go with a bunch of system flows. So first is let's say, open my mesh leaf for the first time: okay, so uh okay yeah, so there are one cluster and a data coming. This is from my uh this is the uh operator is not deployed. Currently it's. uh It was deployed 10 seconds ago.

A

So, just like just when I started messaging, the first thing it did was check if the operator was deployed, if not it deployed the operator and uh this old. This is the old data that I'm getting from my database so anyway. So if I come over here and I look at operator and as you can see, the operator was successfully painted operator is there and uh here as well, you can see the button is turned on the operator. Is there? So if I turn it off now, and so it's turned off.

A

If I go back here, my mesh, the operator has terminated now, if I uh reload the page now, I have added one more uh basically one more.

A

If I go over my pr, I guess for somewhere around the controllers here, I've added one more state. So uh currently we had not deployed state, and uh now I have added undeployed state, so this undeployed so not deployed is when is the default. So when uh the first time you log into mesh, you have your operator is in not deployed state, but if user intentionally undeploys the operator, we don't want them to reload and then the whole operator comes back again.

A

So when someone manually undeploys the operator, it goes into the undeployed state and when the operator is in undeployed state, we don't uh automatically do it unless user prompts us to. So. If I come back here again and if I turn on my operator again- uh and if I go back to my so the operator comes back so now it is in uh deployed state again, so uh there I've added one more now, let's okay, so this was, um um I was doing everything from hd ui. Now let me demo it from.

A

Let me turn off my operator from here and demo. It from hdctl.

A

A

Actually operator is gone so if I do message eating system start, so there used to be conflict that uh earlier, if you had installed it from ui and then you try to install it from meshley ctl. The thing is that the chart name and the release name all are like scattered all over the place, and I am trying to centralize each of these. So there used to be conflict that if you have already installed it or the operator from ui, then you do try to do it from the mexican system start. Okay.

A

Actually I uninstalled it so before I demo, let me install just to see just to show that there is no uh conflict.

A

Okay yeah, so it's back up and if I do mystery, sequel system start I'll have to close my local server because both of them listen at 908.1, localhost yeah. So uh I had link ready adapter in my mesh config mystery server got started. My operator was already there, so no problems with that mesh thing is that broker is there and if I uninstall so yeah yeah here I can come. I cannot uh do it right now from this ui. The new.

A

This ui is basically the is container image running, which is the latest table and of course my changes are not only this table. So if I try to undeploy it here, old methods would be used. So I will not do it and if I come back here and if I do message satan system stop, there would be a clean uninstallation of everything so that, um if I do the the the only thing is that meshes currently doesn't allow users to only un-deploy the operator.

A

If you do machine serial system stop, there is no extra flag for you to just stop the operator. So uh if I go back here, there would be like everything has been terminated and um the whole message name space is cleared and deleted. So this is like a complete cleanup right now we don't have granularity and messy ctl to allow for them to selectively just deploy operator, although we can now, uh since we have moved to this new approach, we can completely do that so yeah the operator has been, um everything has been undeployed.

A

It takes time to delete the machine in space. So first part of this pr is to have this extra state um inside undeployed state inside of the. um So we have a centralized way of uh managing the deployment and undeployment. Basically, the life cycle of controllers that we have controllers are the imagery operator and the crds that come under, which the operator all are come under the umbrella term of controllers. So we have same so this generic interface that we use for their functionality.

A

We didn't have an undeploy state, so earlier what we were doing is we were deploying it using this new approach of having this centralized logic, uh but while undeploying we were using some legacy function, which the comment said it has to be deprecated, but we were still using that to underprice there was this conflict again so to centralize. I have added this undeploy function here as well, so that everything all the logic related to every life cycle of the controls come over here so yeah the message resources are stopped.

A

The cluster has been cleaned up and yeah so that that was uh all from my side. If I start it again, everything will come up without any conflict, so operator again comes up without any. You know, because of charts are separate so yeah. That was everything from my side.

B

Oh yeah, here in a bigger conversation or a broader discussion here, it's mesher's genesis was in context of docker deployments and so being able to just you know, deploy as a set of containers, whether that was inside of kubernetes or or outside of kubernetes.

B

The operator was um a way of helping ensure kubernetes. Well, it was primarily a way of helping ensure that meshsync was that that message server was always in.

B

Up to speed in sync doing you know, service discovery in some respects with kubernetes, and so mastery has one foot in as kubernetes native and one foot out, um there's some duplication and overlap that occurs by not necessarily natively being fully kubernetes native.

B

There's some benefit to not being that way, and I'd like to say in so much as like the example being like the mescheri docker extension, which has helped mastery be well, which has helped like new users sign up by 25 by a lot.

B

Yeah, I don't know still a point of confusion. I'm not! I I'd love to hear other people's thoughts on this. This is an area that I've like it's super important to the project in terms of how it deploys and works. I don't know that it's necessarily a step.

B

Like yeah there's a step backwards to move to two: oh yeah.

B

We already have two releases: two duplicative sets of helm, charts, they're, just structured differently. The operator is mandated in one and maybe not in the other. It's not really working for us. You guys have been refactoring to make it work for us. The two use case: there's three use cases for installing mesh, reusing, helm, helm, command line with helm, the master ctl command line, which gives us more control over which helps the user experience and gives us more control to.

B

She should point over whether or not you just want to install the operator or you want to install both. You know both home charts and then for and measure ui as well. It becomes really apparent inside of mesh, for ui, at least to me, is that you may end up loading up and connecting your mesh reservoir instance to any number of kubernetes clusters and the the flexibility to just deploy the operator as the only component necessary as the footprint in each of those clusters is, and then that's a critical capability.

B

um I don't do this often, I sort of feel like I'm in some respects. I'm feel like I'm at my. My wit's end in a positive way, because others are here: matt and serene and ashish that might arrive at a you know I'll, acknowledge that there's some.

B

I think this is true, for maybe all of us to varying degrees is that maybe none of us on the call would say that we're an operator, sdk master or necessarily a helm ninja so to speak and that there might be some ins and outs some of them that I think matt is probably pointing out um that once you get three layers, a few layers deep, you might figure out that, oh, what we have today is achievable or you can.

B

You could keep it as is, but here's the caveats and here's how you'd have to change some of the behavior and.

D

Yeah, I think, like uh one way to consider it or one way to think about it in terms of like.

D

What's the predicate you know, what's the goal, um is you know it's all fine to have measuring ctl and have more fine-tuned capabilities, but so long as you don't, for example, require a user to do something in measuring ctl after installing an operator or something like that like that would be, in my view, an indicator that um you're kind of doing an end run around some of the core benefits of leveraging kubernetes, as a platform right you want to be able to say like here is what should be here is my desired state.

D

I want to connect to these seven clusters in this kind of way, and I also want to do this thing over here. uh I want to send a duplicate set of log streams over there. Whatever you know, whatever that config is, and it can just be like you know, coop could apply once like, or a git ops or like argo or flux can just like grab it from kit and like that's it soup to nuts right and like so long as that property is retained.

D

Then command line, tools and extra capabilities are super cool and and running it locally in docker might be great for development. There's also ways to like patch an instant, send and use service meshes. As we all know.

D

You know for debug scenarios, but like that core property of just like the operators replacing what a human would have to type as long as we don't add things.

A

Yeah here make sure you, after, if you use meshes ctl, to deploy something users won't have to do anything extra. The uh the only here operator. The reason I'm saying that mystery operator and message server are orthogonal is because measuring operator uh generally, like you, have prometheus operators and all these other operators which manage the life cycle of the components beneath them and like the all of those components, are defined as crds and all mesh server is uh like comp.

A

It sits outside of in the architecture, diagram itself outside of uh the operator, so op so operator doesn't manage. Meshi's life cycle and machine doesn't uh but but yeah you can deploy and undeploy, but you can uh manage operators life cycle through measuring. So that's why operator can't be like the single so.

D

That's an inversion of what most people would expect from an operator what, but I think you could still retain what you have today. If you just like made a minor like how you consider it tweak by saying like, for example, if there's a crd for measuring server.

D

uh That's fine! That's fine! If there's no instance of it created, but the operator could have the capability to spin up a measuring server. If you want to do some debug capability or some legacy capability, where there's an external to kubernetes on bare metal, meshery server, then they're simply the absence of a crd defined for the server and then in the configuration for the other components they point at that server. I would recommend using one of those external name or external dns, external name, whatever service types that would refer it to your bare metal server.

D

But but you know you, don't you don't have to do to have the server that would manage the operator turns everything upside down and then now you have corner cases that that we that this discussion started with that are not really solvable.

A

Yeah, actually it's because I think measuring operators, the better name, would be mesh sync operator, because machinery you sit outside it can like sit outside the cluster. So uh if, if you think hierarchically, it can manage the life cycle of operator because it's not uh it can like measure being inside of a cluster inside of on on a bare metal on a docker. It's like it doesn't matter where it is and it like. uh We have multi.

A

The way we handle multi cluster is that for measuring all those clusters are just some external entities to reach out to and manage so on that hierarchical level. On a cluster level, it measurably server, sits above all the clusters, so it can be in a cluster while managing some cluster outside of like it can be in cluster one and then managing stuff in cluster two cluster, three cluster, four. So it's not uh yeah, so it's not inside. So it's not clustered dependent cluster scope.

A

It's it can be deployed inside the cluster while at the same time managing operators across different clusters.

D

Managing operators, or talking to the components that operators would do like like see. I I guess what I'm saying is: if you leave it, if the operator is the supreme controller of you know, the supreme overlord, like like everything you do, could be done by the operator. If you want to manually, do some things: fine right that that's fine, but in terms of a core capability like if you do it, the other way than to anyone who's deployed a kubernetes compliant environment, their their release.

D

Processes are like give me your yaml and that's all that's required right so like if you- and this is how we install yaml and then you know, take your pipeline. You know, but whoever and then there's a huge variety of pipelines, but they all are predicated on.

D

You know the operator controls all of the things on a cluster, and now I can ci clusters now that we have a cluster api right now I can like a ci job, spins up a cluster and deploys my stuff to it, um and I don't have to have a separate auditing threat, modeling compliance approval process. You know to manually install the home chart if it, you know what I mean or having to require this ordering of things.

D

So I think again, like I'm, not suggesting you do or don't do anything, but I am saying that I hear an either or argument when I think it's a both right like like. If you install the operator with no server crd defined, then you've effectively got what you have today.

D

But for someone who is a kubernetes everything environment and they categorically don't want to run vms, they retain the ability to then create a server crd and how to config point that way, like you, don't have to choose smarts from the development side and I don't want a rat hole, but um I'm happy to chat offline or.

B

Yeah you can yeah, you can have your. You can have your case.

D

Yeah, I'm a resource use me but um yeah I don't wanna. I don't wanna take more time this big meeting than than the top before.

A

Okay, uh there's probably uh you'll carry on the discussions after this fall with ashish. So anything else uh we need to see. You are saying something.

B

What a great conversation yeah you can have your cake and eat it too, in some respects, uh summarizes what uh matt is saying and yeah. It's like this is this: is my own embarrassment not to be shared by anyone else, but um yeah the duplication and overlap that we're like painfully and slowly taking ourselves through by not leveraging kubernetes native capability is like a point of embarrassment for me there's you know you can justify it, there's reasons for it, but um well I.

D

Wouldn't go that far because a lot of the capabilities we're talking about, for example, cappy the cluster api, which like is one of the things that makes any of this even possible like if you want to treat clusters like cattle, not pets, although that's an anti-animal thing, if you want to treat them as ephemeral or disposable, that capability didn't exist more than a year ago and it didn't exist within like six months ago in a non-beta state for many of the major cloud providers.

D

So, like you know what we're taught, this is iteration on a fast-moving ecosystem. I wouldn't I wouldn't go so far.

A

Okay, uh so I I think kanika is not there. uh So moving on to my item, uh which is uh arches, a form refinement, and um I mean whenever I create any design or either in the mesh map or in the uh from the measure, design catalog.

A

uh I have a you know very big uh nested forms that comes up uh on my face uh without giving context of what I'm doing and whatever uh and the you know automatically expanding all the fields also create some validation issues, specifically in the measurement, because we are showing it uh to the users, also so yeah, so to remove all those things and to give you know a final, uh you know site uh for the users.

A

uh It's in rj sub refinement that uh we have added uh some kind of expandable upstate on uh object fields also, which has uh a title enclosed in it, and kind of uh have something as these children. So this is very simple or just a form, but if you can go to you know uh something uh like tron job, which has which is like hold something uh bigger in its schema. So you see a job. Template has metadata and spec. So right now you can see all these things.

A

Like what job template provides and what you want to specifically, uh you know, alter, for example, metadata spec, based on what you want to change. You can, you know, make your amendments are, but uh you know uh configurations there and and that's how uh the look is right now. uh So, um with this very minor change, uh the form um looks, you know much cleaner. uh You can also, you know once you're done with this. uh You want to get out of this view.

A

You can you know, I mean collapse, it any change uh done here, while you're by while you're collapsing, you will retain you know it will retain as it is remain as it is. So that's the update from for the you know. Rj uh schema, I mean rj form refinements, and that is uh solving a bunch of issue. I just talked about so yeah. That's the first step update. I have on register forms.

A

Any comments there before moving to the second item, so this is the pr that specifically does it uh you can see through it uh I'll give it. I mean I'll I'll, add a link to the doc. Also, I just forgot to do it.

A

Okay, so if I take that, I don't want to you, I mean talk too much in the context of mesh map, but in the context of validation errors. If I take a component and uh if I have all the parent level fields, uh okay, then there would be no validation errors, just because that I haven't filled in three level: three levels: deep child field: no, no! You won't see because that schema is completely uh thrown away, so uh any other palliation are done on schema and not the form data.

A

So these, two being you know completely independent. uh Basically, in a short year, you won't.

A

uh Okay, I hope I I mean we don't have any other comments there. So moving on to the next item, uh eslint pictures uh in machine ui, uh so this is this- has been the long standing item like a very long. It's been on from three.

A

So so uh I need some help are completing the items. uh I think.

A

uh Some checks in the esri darcy, which was uh I mean, check.

A

They are suppressed by the esl rc jsonpi, which was causing a big problems in the ui, uh sometimes, and uh sometimes the ui just crashes, because something is undefined and they are being called inside the code. So all these uh problems were originating from one small uh es10 configurations which were configuration, which was uh which was no one depth. uh I think to be exact. It was no unused wars, and uh these two I mean they also create a kind of confusions, are rendering things on the our ui side.

A

There were many duplicate keys, as you see, uh 52 editions, 85 decisions. So all these things were cluttering uh the ui, and you know time to time they are showing something uh like very big to understand like where the things broke so yeah. These are the changes uh that is done uh to get rid of all these things, and the last was our machine extension e2e test. So uh we have a meshi extension, which is called mesh map and uh there is a depository specifically that provides the package.

A

uh You know uh that if you have the access to mesh map, then all these packages will be downloaded from these. This repository and before you know, uh and you know, that the mesh map on mystery extensions are development is being done on a separate repo which is private.

A

uh So all these, uh all those I mean uh those being uh exposed to the user needs to be. uh The code needs to be tested for the quality purpose uh to ensure, like everything works, fine, even after all, that will do custom uh a bunch of commits.

A

So uh that's what we are doing uh in this meshed extension packages, e2e test, so uh I mean uh we have added some uh workflows uh and uh for for unit testing uh testing the uh source code on each pull request and on each commit and then on post release. Also, so, if something happens you know uh post releasing like something is breaking the then we should be the first to know like, and that is breaking so yeah.

A

These were the three updates that I had today from my side. If you have any comments, then.

B

Hope the ux improvement on the forms makes a lot of sense, and it's a little more than I guess. I guess it still qualifies as ux, but it's um the step forward does alleviate a lot of pain. Actually I guess it's more than ux. It's like it's a it's a bug, because if you don't, if there's certain configuration required fields that you're not intending to even use and the form is forcing you to use them, then yeah that's more than you, so that that's great, please alpha is up next uh all right checking my.

B

A

Okay, so my pr is about for removal of the use of kubernetes manifest and the system update command and measuring ctl. So previously we have transition from using the kubernetes manifest we have charts, but still uh there is being found the certain usage of this car manifest in the system, big command, which which uh ultimately caused some failure in the integration test as well.

A

So I have worked on it and, like I found out that, let me one give you yeah all right, so uh I was able to remove the uh manifest code which is present in the system update for both the docker as velocity uh kubernetes platforms and ultimately, uh the end result is.

A

The uh integration test was started to function normally, but uh there has been some kind of a failure which is because of the uh old version present here present here in the test files of the measuring ctl, which is the latest version, has been not been updated yet, uh but that part uh anthony is working on. So I haven't touched that part yet so yeah that that's much of it actually, so I have uh able to remove the cage manifest and it was uh able to work uh work as mentioned so yep.

A

That's much of it.

B

Any comments in it.

A

uh I have I have a comment, but um that's go ahead closely related to the life cycle of uh mesley through messy ctl. So while you run this thing in just um inside of, I guess in messages system, stop I've seen this thing that uh we have a force, uh a force, an installation feature which manually picks up each uh uh humanity's resource and undeploys it.

A

It's it's it's a long and it's a long and weird code, and I have I had seen it a while back and I was hoping to uh talk to someone who was working around this area. uh So since you you're already close deep into this, um you know system update command, because system update internally uses the stop function and inside the stop function.

A

If it, if measure, fails to undeploy something it it take it, it tries to undeploy config map by config map lan port by pod deployment by deployment, which is quite weird because, like the idea of helm chart is that we package everything into one thing. So if we add something in our helm chart, then what we'll have to do is we'll have to manually go into our code and deploy it in the force uh feature, so it kind of seemed redundant.

A

um So, while you're on it just take a look there as well, so that uh we are not trying to you know do do all that. To the there is a force flag or a.

A

So any more comments.

C

A

C

A

uh Yeah then look like it looks like it's time for uh gaurav to discuss about machine 0.7 blending.

C

Yeah, let me show my screen.

A

Yeah, so the gender item is about uh machine meshi.7 planning. So currently we are at uh dot six for measuring and parallel to it. The whole machine ui is being refactored and most multiple people who are working on this particular refactoring. And if you are interested in machine ui refactoring, you can also jump in over there and provide help in this particular area.

A

So let me share the link of spreadsheet in chat so for this refactoring. Some principles are being followed, such as react, 17, material, ui5 and redux as a state management tool.

A

So if you want to know how you can contribute to this particular refactoring, you can have a look at this particular readme file, and you can add your name to any of the component that you would like to refactor or say hi and messi ui channel and the best benefit benefit you can get out of this refactoring is that you will be get to know more about uh mystery code base, so the particularly messy ui code base, so volunteers are needed for this dot. Seven road map feature champion against all the issues.

A

Yeah next item is also by me, so it was a fix of end-to-end and integration, machine vitex messy ui. So let me go to the pr so yeah this pr was uh recently two days ago. So it covers a graphical utility package for queries and mutations, uh end-to-end test fix for user preference, pick and integration for your preference pick, and this was uh merged, and now we can have. We have all the green checks in any of the pr for the ui we check, but at a bare minimum cost.

A

We disable the filters for on temporary basis, and for that I have already created an issue over there.

A

So this issue requires a change for the request that we are sending in cyprus test using graphql interception. So I have tagged mario on this particular issue. He can provide some help and if anyone else is also interested in writing more end-to-end test cases for mashery, they can also jump in over there to know more about cyprus. We are using cyprus as an end-to-end testing tool for measuring. You can go over here to learn more about cyprus and how you can set up your local environment for machine ui for writing more test cases.

B

Yeah on on my prior topic about measure, ui refactoring, there's been a fair bit of effort. It's been kind of um well. The community has gotten a jump start on that refactoring, knowing that that was something that we desired to do now. Now that we're here and working on the dot, seven release um and my open question, and so that's so it's fantastic, there's a whole spreadsheet of all the components that are being rewritten and uplifted and um and being made to be mui5 compatible and next js 11 something something compatible and react. Seven.

B

I thought it was gonna, be 17, but maybe in 19 compatible or you know something anyway, just a bunch of uplift upgrade on the frameworks being used.

B

It'd be really nice to be able to move to some of those newer frameworks without having to go whole hog on displacing the current mesh ui with the changes that are in the refactored branch um pankaj. I don't know if this is even a feasible thing or maybe one framework is feasible to move forward, but yeah.

A

Anyway, that's because, uh since we are moving to material ui5 and ui5 on the restructuring, branch and nufi only works with react, 17 and above so, since uh the structuring branch has already taken that step. uh Those changes need to be made for those components that we are working on and only then uh react. 17 will work along with mufi and it's pretty good to move from uh material uf4 to mu5, because there's significant improvement that we get in uh styling and the different components that are there.

A

So I couldn't figure out a way that let's say apart from mufi and react, which ones are the ones to pick on, because uh next year's again uh the newer version will depend upon the act 17 and further. So it's better if, since uh nuf5 uh upgrade has been uh started with some components that is finished up, so react, 17 works and then in turn next 11 and further also works.

A

So that will be a really a big dump or a big change that will be brought.

B

Yeah darn yeah. I was hoping that it wasn't uh yeah because there's just more risk associated with okay, well, um gara it might, and for those that are involved in that refactoring.

B

um This has been a consistent topic of the mastery dev meeting. So that's that's wonderful.

B

We might, if you think about it, we might start to do a few demos or we might start to I'm just trying to identify how we can bridge that divide, because I'm because there's no doubt that the current measure ui um at the point in time in which some of these components were refactored, the current master ui has advanced with some new capabilities, and so some of that refactoring is out of date and um yeah it'd be good to just get a walkthrough or a tour of the current state of that refactoring.

B

From a demo perspective.

A

A

Yeah so I'll uh make sure there will be a demo on this. A new refactoring or next call yep back to you.

A

Yes, so uh there has been a good issue standing there uh about. You know adding tooltips uh to the card component, so uh I think kamal is taking this item, so our key that will uh reach him soon.

A

Next is a multi cluster. Allow users to specify max measuring, blocker and point manually like this is the last minute topic, but you can move it to the future call also if it is a big item to discuss right now. It's up to you.

B

C

You say jason or me.

A

I said uh it's your turn and if you know out of time, then we can uh further discuss it in another call like uh in the next call. If it can't be, you know, uh uh kind of completed or picked up in one minute.

B

Yeah yeah one minute on the this next one. It's really relevant to the operator sdk discussion and some resource definitions, and that is that in one of the one of the two current custom resources that are used.

B

Where it's just this particular issue, the the one called allow users to specify measure broker endpoint manually. It's really helpful for some of the for many of us that are on this call. As you go to run potentially multiple clusters on the same local system. um You you some in certain situations. You find that your ports are overlapping on the same local host interface, uh the the broker ports, the messaging broker ports.

B

So it's really convenient to be able to specify and tell the broker to use an alternative port you to impart to what matt young was saying earlier on the call.

B

Well, there's a lot of value to be derived from use of the custom resources and and honoring and sort of adhering the settings that a user might define inside that custom resource. So today, there's a custom resource for measuring broker part of it.

B

It's not that long of a configuration part of the configuration says it identifies the external like the type of networking to be used, the external ip, the internal ip and maybe the port number, and I don't believe that you know if if a user goes in and they modify those settings today, the operator will measure operator will, I believe, ignore those and kind of uh and and overwrite them overwrite changes that the? U um an administrator, might define so that that's what that point is about.

B

So it's something to kind of, as we dig dig it more into operators that level of control over the deployment is advantageous.

B

So it's an open issue out there.

B

I'll try I'll grab the issue number yeah. That.

A

B

A

Yeah and and with that, uh we are out of time too. uh So, let's uh you know continue the rest of the items in the next meeting uh in the next development call on wednesday again so till then uh good day. Goodbye.

A

Thank you. Everyone for joining us.

B

Bye guys, thank you. Avi abby.