Layer5 Meshery Development, 6 Apr 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Meshery Development Meeting (April 6th, 2022)

Description

Meshery Development Meeting - April 6th, 2022

Agenda -
- (Istio) E2e tests with a sample app. #406
- Meshery Test Plan Spreadsheet
- Patterns Card view. #5339
- [Blog] Patterns vs Designs
- E2E tests on Compatibility Matrix on Meshery Docs.
- Use Go SDK in EKS config function. https://github.com/meshery/meshkit/pull/168
- Add utility functions for leveraging the capabilities offered by kompose. #164
- Docker Desktop + Meshery
- SMP dashboard design spec [link]

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

Okay, uh abhinay: is this your first measuring development call if it is yes, sir.

A

Sorry, your audio cut out.

B

A

B

A

Hello, please take a moment to introduce yourself to the community.

B

uh Sir, my name is avinassen and I am a student from modern world management engineering college, and so I am working on the morning stack and android development. So I am starting contributing on the open source project.

A

Awesome sounds good great to have you in the community here. Okay, so I don't see any more newcomers.

A

Let's get started off with the agenda items. Then, okay, I think jose is yet to join us. So, let's start with raj raj. Could you just give us a brief about the e2? We do it as.

C

Foreign, no, not really. Oh.

D

E

It's a little bit better.

C

If you, if you can take the mic and sort of shove it down your throat.

D

C

I don't have a mic.

D

I have to basically uh put my face into my laptop.

C

It's about the same as it was a second ago yeah, but but you're you're, audible,.

D

Okay, so let me just quickly uh quickly explain.

D

Yeah, so uh I've been working on the end-to-end tests uh for uh istio with the sample app. So this is the pr that I've made and uh yeah just couple changes with the a couple extra workflows uh and yeah so they're all passing right now, which is a good thing um and yeah. So basically, I I used the minimal installation for sto here, because apparently, nine pods is too much for uh github actions and some of the part or one part, exactly was not getting enough resources to spin up.

D

So the difference between a minimal installation and a full demo installation for istio is that the two of the parts called sd, ingress and eggrest do not spawn in minimal install the minimalist only spawns one part for easter, which is the sdod um yeah. That's about it, but also uh this is another concern that uh lee had brought up in the comments that we should probably be using meshuctl to deploy the adapters instead of using the kubernetes, manifests in the github actions that we're doing right now so yeah.

D

That is another thing to do, which will actually require us to change the workflow across all the adapters and yeah, and also change the main script for the testing in the machine repository.

D

Yeah, that's it if you could hear.

C

Yeah yeah yeah yeah. That's that's good stuff! um Secrunder. I wonder if I wonder if you have any thoughts to offer.

C

Or mario, to the extent that either the two of you might have looked at that.

F

I have discussed with withdraw threads, how to we were still figuring out how to use the how to replace the manifest with with measuring sit here in the workflows, so we are currently still working on that, but otherwise I think the rest of the stuff is good in in that request. Apart from that, thank you.

C

Nice yeah uh a couple of educate me if you would.

C

Yeah man, I want to say that second use so bad root, rash, root, rash moon, rash.

C

Sure I can just imagine in your mind, you're like lee. Please stop just stop man, it hurts my ears.

C

Sorry, uh the would you kindly repeat a portion of what you were saying with respect to the different deployment styles that we might support for istio.

C

uh It sounded like with respect to the use use of a profile that would have an impact on whether or not the test should assert um multiple control, plane components is steel control, plane, components or just like one or two.

C

um I thought I saw the word default in there. It was that for the default profile.

D

C

D

Share it again.

D

So this couple of profiles- uh I didn't quite see all of them, but the two that I even concerned with that like minimal and there's another one called demo yeah.

G

That just spawns up all.

D

The yeah so that.

G

D

C

D

This is the name.

C

If the environment supports it, um I highly encourage us to do the demo uh profile because, because we're not gonna run this thing in production or much longer than 20 minutes or, however long it runs. But the point is to just verify functional completeness or to verify that we're touching as many things as possible that measures capable of deploying as many things as possible, and so, if we do the demo version, we might get prometheus and grafana, and you know other things potentially.

D

So that's what I was that's what I tried in the beginning. I tried it with the demo but uh yeah but github actions. Just so I looked at the status of the part and it just said that it doesn't have enough cpu to spawn the spin of one of the points in all this space.

C

I can just imagine some of the other service mesh projects getting a big chuckle out of that, because, because they like to differentiate between how lightweight they are versus how many resources istio might take. um Well, that's a little bit of a conundrum. So for now the default profile works with an istio d.

C

Okay, that still is, you know, touches end to end like can mescheri provision istio, and it can answer that question. um Did you happen to do any research by chance about uh potential things that we might do differently like? If we're the type of kubernetes that we're running? Maybe we could run a different one or lighter weight, or if there were configuration options afforded to people afforded to workflows when they run the github runner like the host environment?

C

I don't know that they really let you you know, request a larger or more resources or, if maybe I I don't know, if we want to go this deep, but it could be that we would spin up a portion of istio verify that it was there turn that part off turn on the other. The next part and kind of do that step wise, so that we're staying underneath the limit, so to speak, um that that might get really hairy. If it does, then you know, but just sort of a cursory analysis of potential solutions.

C

Another one another potential possibility is if, if the measuring environment that you're the measuring deployment that you're using if it is anything more than measury server and the istio adapter in this case, we could not spin up any of the rest.

C

D

Yeah, maybe so, can we just not spin up, let's say missing all those spots potentially.

C

Potentially there is a like that's a little that's yeah, it's slightly dangerous. We might be able to like all of the rest of the adapters for sure you could leave off all the other service mesh adapters.

C

um Potentially you could leave off um measuring ui. So I don't know that that takes much resources. um You could potentially leave out the operator and mesh sink and the broker.

C

The reason that I kind of wince at that thought is that.

C

C

Actually, I think in the assertions that you're doing right now or that we're doing right now. I I don't know that they're accounting for um the full round trip, that's provided through mesh sync and that's a little bit unfortunate.

C

Because we would really want for that, because if, if mescheri deploys istio but but we're not verifying that mesh, sync actually discovers it and returns that information back like that's kind of that full round trip, if you will of provisioning, it then discovering it and sending back the info, then we're kind of already only sort of only halfway there. In our test case design and our integration test.

D

So yes, already like, we still don't have that assertion right on in the test money.

C

There might be another workaround uh might be that there are multiple workflows that initiate and maybe one is able to call to resources in the other. One.

C

Which would require that there's some networking that's opened up somewhere yeah.

D

Yeah, I'm not sure how that works.

C

Yeah, it need to be dynamic, like there's system yeah um anyway, if you don't, if you don't mind if you capture each of those thoughts, just as simple little bullet points so that we can follow up on each one and just figure out if any of them are easily possible.

D

C

We can yeah, we can just do a cursory check.

C

That's great uh sounds like we should go forward with merging that and the thing that you mentioned about use of measuring ctl, so we certainly like measuring, certainly capable of deploying service meshes and that's what you're verifying the issue with how we're verify how it's being done is like we're not using ourself to deploy but measuring, isn't using itself to deploy istio.

C

C

It's it's only using the istio adapter to do it, whereas we could be initiating the request from mastery ctl to the to measure server to the adapter, and that would be testing two more of our components, which would be fantastic, yeah, yeah. So there's something that I'm still hoping someone will really wrangle, and it's in the mesherie test plan. There's the spreadsheet the measuring test plan, one of the sheets inside that spreadsheet, which I'll try to find is well. It talks about all the the different tests that are being run and where their.

C

Where they are where what coverage we have so there's a master test strategy, but that's there's a measuring test plan.

C

Here: here's the sheet that we look at in the building release meeting, which I think we might have this week and then there's an automated test sheet down here. I I realize this is a little bit small.

C

C

Talks about what workflows get run where at what they, when they get triggered kind of what they pull in what system is under test. So what area is under test and your? What whoops.

C

This is a good start to what would be a comprehensive um picture of what's being tested and where we're going to find, we can deduplicate some things, we're going to find that we can reuse a lot of things. The notion that we would use measuring ctl to provision measury like we're doing that in other places. Already it's not necessarily something you have to write um and so good. Where is that?

C

Well, this is a great place to organize that I'll put this into the beating minutes.

C

That's uh that's all for me.

D

C

That's also like.

D

D

Yeah, so just a small little heads up uh about the overall uh sample app test planners uh second under right now is working on also the patent file for linkerd, and once that's done, I think it's straightforward to have that on uh linker d also, and then we can do the same thing. Copy paste for other adapters.

C

Totally that's great yeah, or even even if we cannot copy paste, but we can just if we can just include so to speak. That's even better, that's great!

C

By the way, I would say secreta if to whomever is, if you're working on the pattern file format um talk to piyush, who also isn't overly familiar with the pattern file format.

C

The reason that you would talk to him is he can tell you about four other ways that you can ignore. Writing out the pattern file format. The point is not to learn that. Actually, um that's not what you want to learn. So push will hopefully post um about that about how to write designs without knowing the pattern file, format.

A

Okay, thank you for sharing moving on, uh we have nitish nitish. Would you like to talk about the compose capability that we've been working on.

H

A

There's some background noise.

H

A

uh You're up next with the patents cardview.

I

uh Yeah there's uh like there's not much of an update prior to the last thing uh we were checking like if there was an error which our posted that time like was it uh replicating in other systems as well, because I was not able to find that bug. So I did like. Did you check like if that's happening.

A

I haven't, but if you have, uh if you have your measuring up, could you just give us like a brief view of how it looks.

I

uh Yeah, just an adjustment.

A

I think you're on mute.

I

uh Yeah, this is how it looks uh all the functions and all the buttons that were needed for this. uh Individual cards are intact and it's working we can update file, we can we can update uh through the ammo, we can delete it. The same only thing that that is left is the uh maximize button and the changing of the page.

I

Yeah then, regarding the changing of the page, I wanted to ask one question: like do. We fetch uh the patterns in a similar way that we fetch performance profile like we fetch for the profiles? We fetch profiles as long along with the page number do we face? Do we do the same in the pattern? Also, I think we do the same, but it's not working similar to performance profile.

I

So I have to look a bit deep in.

A

That, I actually don't know.

C

Yeah, I don't know that that yeah, to repeat your question, I I think it was um do we have uh do. You have a count of the number of the account of the total number of patterns.

I

Yeah yeah, okay, we it's like uh we are following the same fetching pattern like uh for the performance profile as well as the measuring patterns, and I try to like copy the paging like component from the performance profile itself, but it's not working similarly for measuring patterns. So I think there must be something that that performance profile does different than mystery pattern. So that's I have to look into it.

E

I

Yeah so like, if anyone knows like, if what is different by pack of that now, then it will be helpful. That's that was what I was asking.

C

Yeah, let me yeah there is. There is a difference. The api endpoint for the performance test for the performance profiles for the performance test is there's one api that will just give you the total result, which is not a well-formulated api, um the other one, this other one about patterns.

C

All of them need to support kind of similar methods about, like total count, is a fairly universal need across them all. If we go look at the reference for either the rest, endpoint or the graphql endpoint.

C

So we have applications filters um here under ohms where we're keeping the patterns or I'm sorry, I guess not actually, where we're keeping the components and then here's the patterns.

C

So page size returns all mystery patterns.

C

This is confusing to me, total. Okay, that looks.

C

That looks like what you might be looking for.

C

Yeah, I don't know something.

I

Yeah I'll look into it more okay,.

C

All right, um other dude, do you mind sharing your screen one more time to see if we're maybe ready to do you consider that ui is ready to incorporate.

I

Yeah, the only uh in the ui pad, I think only the maximize button is like. I have just added the length 100, which is not required. I think we need that same functionality, as we have in table view like this thing. Yeah, I'm incorporating this kind of thing in the card viewers.

C

And then um it'll be even more confidence inspiring when you- and maybe you do, but when you have it connected to real data. Is this connected to real data yeah.

I

It's connected like the pattern, is what we have created itself, so, let's edit the pattern also.

C

Nice so wait a second okay and then do we have the ability to pop into yeah. We do that design button on the front. Okay, we'll take you here all right, oh fancy. Okay,.

C

Who, who else has feedback here? What what? uh What don't you like about this? Yes, yes, you on the call. What don't what don't you like about this there's white space? It's not getting used. Maybe there should be three columns, not two. If that's what we're gonna do, uh maybe the modified on should be either, should a hug and be a little more friendly with the title or should disavow itself and like move all the way to the lower left-hand corner.

E

Just like they created.

I

C

Yeah but it's sort of otherwise kind of giving the title some personal space, maybe a little too much it's sort of like not. You know not sure if it wants to be friends.

I

I think uh we have like some other, like other components in mind like which will be which we will place in the future. So I think that's why we are leaving some space.

C

It's like you're saying when someone uses your own words against you, it's just there's no room. There's no wiggle room, that's right! So betty I don't you know. Maybe we would do well to drop a link for everyone to look at the figma designs for the second iteration of what this would be, and that way even sounds like. I should go. Look at that too, because clearly I forgot.

C

Yeah, so there's a seamless transition between grid view and table view is what it looked like it.

C

You know, for my part, understanding that there's a second iteration coming forth, one that will require a couple of back end changes. I suppose I you know. For my part, I don't see any reason not to go forth with.

C

You know put this with what you have.

C

But what does anyone else have an opinion.

C

So I'll share this just as a refresher, which is oh boy patterns,.

C

Which is here so alvin schinendu has helped do some designs.

C

If, for all of you, that can access figma, which you all should be able to, um there are two drafts of what this grid view would look like. So the first iteration is like this.

I

C

E

I

The background color for the back side of the card.

C

The card should flip on click of the card, and it should flip back on clip click of the card. There's a little bit of bolding. That's here.

C

um I think you have it if you've got an italic site, I think yeah anyway, a little bit of review just between here probably be good. So that's the first iteration the second iteration.

C

Begins to identify the particular service meshes that the design is compatible with, or maybe the more appropriate thing to say, the way of characterizing. It is to identify what types of components are included in the pattern so because there could be kubernetes components or linker d components or argo, cd components or other.

C

So if you start to think about this in a little more of a generic way or expanded beyond a service mesh, so so we can identify the types of components you can go to details. Do the deployment you have a label of like what, if there's a if it falls under a particular category.

C

C

C

I think this this draft here probably needs some work.

C

Any other feedback.

C

Ultimately, there'll be the need for like identifying the user that created it, maybe if the user has shared it with their, if the other teammates team members have access or if they want to yeah one of the numbers that would be nice to show outside of what suvetti had said about like the total count of designs, the total count of patterns is.

C

How many components outside of what type of components are included in a design? How many total components are included in a specific design like one design might be quite massive, might describe all kinds of things, whereas another design might be pretty small.

C

Another thing that would be nice to track over time is how many times a particular design has been used. How many times it has been applied, potentially if it's an active use as well, so there'd be a nice linkage between the fact that you have a descriptor, this static configuration and whether or not there's any actively running infrastructure. That's using that configuration.

J

So so also which clusters or contexts it's being used at the moment right.

C

Yes, yeah good great point.

J

And what is a component? Would it be a specific kubernetes objects or anything.

C

Yeah, but if yeah yes, it's it's actually yes to both. um If you consider that kubernetes itself has core components that are part of kubernetes proper, if you will like a deployment or a pod, and then there are any number of custom resource definitions or custom components, custom objects.

C

Like an istio virtual service, it says crd and there are any number of applications or pieces of infrastructure like in, like a service mesh whose components can be represented and so yep both.

C

Because those yeah, because those custom components that come from a service mesh get represented, they would be represented in the same way that core istio components would be represented. So.

J

C

Yeah, so there's a couple of things like one. For my part, I'm gonna, we really have to go clarify the like uh designs and patterns start to er.

C

We need to start to imbue particular properties of what it means to have a pattern versus a design, and let me say it really quickly for everyone to kind of grasp it, and that is that, hey, if you're, using measuring to configure your infrastructure, wonderful, you can do that for kubernetes components, uh kubernetes, proper or core components or custom components like custom resource definitions,.

C

And you create designs, configuration designs, they're called designs, so fine, they're designs, you can use a visual topology mesh map. You can use the designer configurator, um you can hand code them, which is what I was saying too was intimating to secreta. That, like like, no you shouldn't do like don't hand code them. That's why the design configurator is there use the design configurator. It will write out that yaml for you, that's one answer or use mesh map or whatever.

C

Yeah, but a pattern is a pattern, is different than we should go: write a blog post on it because I've said it. I don't know how many times- and at this point it's shame on me for not just being able to point people to the faq. That explains the difference.

C

So so yeah uh devil priya. If you happen to be taking a couple of notes or two feel free to give me a task on that.

C

Nice well good, so suvetti like dude before I imagine that by by this time next week, I imagine by this time tomorrow your pr will be merged.

K

Which is good, yeah, probably yeah,.

K

No promises from him.

A

Okay, thanks uh subway, uh okay, let's circle back to nitish before we move on to the next items,.

H

Yeah, I hope my audio is clear. Now yep yeah, that's great okay, so.

H

Okay, so my topic about yeah, thanks for watching so actually like uh um yeah, there is a utility called compose and uh which, which actually converts a docker compose file into kubernetes, manifest so essentially like uh there's been a. I guess like this is publication that is like docker is going to release something called docker extension and uh yeah.

H

uh um We as a platform, ask mostly like we're actually going to uh embed, create an extension for docker, so that, like so that, like people who are using docker desktop, can actually use measuring along with it and then yeah they'll have they can have a negative experience, so that sort of thing.

H

So, as we were working on that, like we realized, we realized that if, if there was a feature in machinery which could actually allow all of the users to convert a docker compose file into humanities, manifest and then like you know, and then leverage the existing functionalities that are presented, it will be if.

H

uh This library called compose, which actually converts a docker course file and the into internet is manifest so which would allow us to do things like displayed on mesh maps so that the users can visually design their uh not exactly docker compose file, because, like obviously like right now, the utility does not allow the users to convert a cumulative manifest into a document. But then like that's something that we can actually uh write the logic for and make it possible, but then like as of now.

H

It doesn't actually allow the users to convert this manufacturer, but then, with with the amount of features that it supports right now with this other pr. All it does is that it adds uh a couple of iterative functions in meshkit which allows the users allows us to uh convert the docker compose file into humanities, manifest so yeah. If you're interested, you can actually take a look into this pr. uh I guess probably other link of that in the chat box, and there are some.

H

There are a couple of caveats to note such as uh the composed that particular library is using a docker slip compose, which is uh just not actively being maintained right now, so they're good, and there is actually uh some dependency problems with the title and we have. We are actually kind of tackling it somehow, but uh as it's not that big of a deal right now we're not actually that was concerned about it, but in the future we might actually move away from it or look for some alternatives, but right now yeah.

H

This is what it is and yeah you guys can take a look at it and if you have any comments or anything just you can put it in stack or right now.

A

All right any comments.

H

And I guess we will be talking about the docker desktop extension, so it might actually make more sense, but yeah.

C

C

If no one has any comments, we'll get to secretary's topic and then we'll give kind of a demo of nitish's work, a fake demo.

A

Hamza, would you like to share your screen.

F

Sure just give me a minute.

F

I hope my screen is visible now right, yep yeah, so there is a link in the meeting minutes about this function that I was working on the one. That's that we're using the in the measure is ctrl command, but it seems like match kit, so I've made it like.

F

It it will give the the user warning here on line 988 like we are about no line line. 93 we are going to overwrite your cube, config file. Then it will. It will have archived that cube config file somewhere in the tmp directory so and um the other thing is the data that is returned by this by this function. Here I don't know much about kubernetes, but it's to me it doesn't seem sufficient.

F

So I'm calling on for anyone experiencing kubernetes to help me figure out the minimal like what is needed, because what is this is what it's going to write. It doesn't seem enough, like the utility section here is empty.

F

Under yeah under user, there is very many more data, so I I don't know the the thing I thought of was getting a kubernetes client set, which is available inside match kits.

F

There is the kubernetes client set inside mesquite, but then how do we access that in measuring ctr to get the rest of the data that is needed for the config file, so I'm still kind of puzzled there? Although cool sign, has not been feeling well this last week, so she hasn't been so available. That's why I kind of slowed down, but she promised she was getting okay and this coming week we are, we are going through someone, but so far this is where this work is.

F

So anyone with any comments, advice, the store, will come.

C

Does anyone have access to any ks.

G

F

We were using the the there is a lab inside a cloud guru that we were using, so we were using that lab for the eks clusters as a sandbox, but a live eks cluster we haven't got any.

C

Of the sandbox cluster that you were using did were you able to successfully connect after.

C

um After you invoked this.

F

Yeah connecting with it, you will invoke this function and it will give you if you have configured your your local setup properly and the I identity stuff. It will give you back data this function, but, like I showed you last time, the data it brings back is not okay, it it could be enough, but to access the eks cluster, but to write a good config. It seems this is all that is returned the way you see it here from around line 476.

F

This is under cluster. This is what it returns: certificate authority data at the server, then under the context it returns. Just this part here where you see it, and it's not really named. It is amazon resource names that are kind of cryptic to the user. You may not really understand even under users. This is all that is returned api version as well, so.

C

Yeah yeah yeah it it seems like it might be enough. I mean I get what you're saying I I I agree with you, um although I think of what was shown there, that might be enough to connect one thing that concerns me a little bit is whether or not that yaml is will work, because you know you've got multiple colons in the same line.

C

I don't know if you're going to need to encapsulate that value as a string.

F

Oh, I I I I just copied this from from the c from the terminal to here just to demonstrate with color and all that, but in in the config file. Of course, we shall have to put some quotes quotations and all of that yeah by the way very good point. Yeah too many columns.

C

If, in that, kubernetes sandbox environment, if you're able to take that cube config as a file, if you have an interactive terminal, you could use cubectl with that config and see if you can just run a simple get command, and that would help confirm that that you have yeah.

G

The other it actually works after you have written it right right, right, right, right.

C

One thing just on line 88: it might be the case we I think we should have some convention for where it is that measures ctl is writing temp files. There should be. If you do a quick search within the code base, there should be an existing spot, and- and so you ideally want to reuse the same spot in part because, like that full, while you might have write permission to like like the folder structure, would need, maybe the folder structure is being created as part of right file.

C

I don't know, but there's just some common hang-ups that you might bump into um right.

G

F

Right right file creates this directory tree up to the file. Okay,.

C

Okay, do we are we still using to log that printf is that just a temporary.

F

This is just them temporary. This is just the product. Okay, I may have to import the logo yeah, like the official, the one being used in the project right right.

C

Right right, oh okay, is this totally standalone what you have right now.

F

um Yeah and the other thing is what it takes in here. Oh I see: okay, it's texting the cluster name, the region and the config path which inside mercury city are there or the the the conflict. Path comes from the utils.config path and the cluster name are prompted from the user, the cluster name and the region. Okay,.

C

Oh, I see okay, so you're handwriting out the config file. I got you okay, but you're and you're, using the aws sdk to in order to help retrieve the certificate data, but then the rest of the file you're handwriting out so yeah, so so yeah, you probably do have an issue with yaml then and so yeah. I would highly recommend like like. I would scour the net for utilities that help you write, write the config, the cube config, because it shouldn't be something we have to kind of hand code.

F

Yeah, so even if it's another, it's another library, you know in the in the in the kubernetes, it's like library, the goku is client. They have a. They have a function that can convert this one into into bytes that you can into yaml bytes that you can write to the config file.

F

But then we had kind of already settled for this right to file with resigner and she wasn't feeling well. So I just kind of tossed it to the side, but but it's still on table just in case it can. It can also come in handy with writing the cube configurable.

F

Yeah yeah, I think.

C

It might feel easier up front just to hand write it out, but it is in the long long run way way worse of a path to go. You'll end up with bugs you have to sustain them.

C

So yeah yeah, whatever whatever utility the uh gates client, has it's good. uh Do other people have feedback here. Questions comments. This is the first of two or three sdks that we want to be using to rewrite refactor the cubeconfig so for gke for aks.

C

I'm saying you have your work cut out for you.

C

Yeah, I think you're gonna get this too. I mean it looks it's looking good.

A

Okay, so uh we have lee, should we talk about the smp dashboard, dashboard design, spec first and maybe the docker desktop plus machinery? We can go on waiting or do you want to yeah.

C

Let's jump to the docker desktop one we can just because I think uh I think subeddie's gotten he's gonna send out uh requests for feedback today.

C

So good so so, hey um dish was talking earlier about docker compose and uh we've been working on a bit of an integration of meshery and docker com and docker desktop so um docker. Let the cat out of the bag recently. So it's okay to talk about this now at dockercon, which is in about a month or so daca will be announcing um a couple of things. One of them is.

C

Docker extensions, so if you've been using docker cli, which you all have at some point, you may or may not have most likely not have noticed that there's a docker plug-in sub command. um This has been here for quite some time. um I've had teams in the past write plugins for the docker cli. um What is it? Does kubernetes actually have that yeah, so they both have.

C

They both you know, borrow from each other pretty heavily and in this case, cube ctl borrowing from docker, but so cube, ctl, plug-in docker plug-in is a way of bringing forth additional capabilities that aren't core to the projects, but want to give you a docker-like or a kubernetes-like experience, and so building off that concept um and kind of building off the concept of the of docker hub um that there would be instead of docker plug-ins.

C

There would be docker extensions, so extensions being within docker desktop itself, so not the cli, but in inside of docker itself you can you'll be able to go to a marketplace of extensions of plugins, of extensions and install them and use them like uh this one that we've been working on so to have mercury here um locally and.

C

Run mystery um at the end. um There's a guy, william charles that just remembered he just asked the question in slack and it's he should have joined this meeting so he could have gotten his question answered it'd, be good to tell them to do that so anyway, uh on the extensions.

C

This is how um the mystery extension is working right now right now it gives you full access to. The extension gives you full access to mesh re running inside of docker desktop, which in a lot of respects, is like super convenient because um well, because you don't have to bother with uh mastery ctl system start you just install measuring now, and then you get access to all the same functionality that meshery otherwise has, uh and you get it connected to your.

C

Local kubernetes instance, or to other kubernetes instances, nice kind of convenient right inside of here, so why mesh? We can do this really easily for a couple reasons. One is because, when you deploy mastery.

C

You get a choice of where to deploy it to.

C

By default will like sort of suggest that you're using kubernetes things are just a little bit easier inside of if you're doing everything inside of kubernetes. That's fine, but but um initially we measure is supported and still does support um docker as the as a platform, and so when you choose to deploy measurey to your docker host meshri deploys as a docker compose app, which is a collection of containers, meshre server, the adapters, so the extension aims to hopefully continue to give people full access to all of mashari, but also a couple of convenient operations.

C

One of those um is so there's two, and one of them is being you know, faithfully delivered in terms of the compose this the ability to import compose apps and to take your compose app like this. This is a compose app.

C

I'm sorry! Actually that's! This is a kubernetes manifest. This was the meshery compose app converted now to kubernetes yaml, so so what you would do, as a user is come over import. Your compose app, your your compose yaml, so something like in measuring. If you go to the root directory of meshri, there's a docker compose file.

C

You take it and import that meshrie then treats that docker compose application as a measure. Application measuring applications can be sourced from uh kubernetes manifests today, tomorrow, docker compose apps in the future, helm, charts and so measures management. Software that, like you know, compatibility, is a big part of what it does amalgamating and bringing things together, making them easier to manage under one roof.

C

So that's you know, applications and kind of part of the vision. There is no different in that regard.

C

And so yeah bring so so then you can, you know, work on those designs here and deploy your compose apps. When you deploy the compose apps, it's a migration. It's a conversion from docker compose to kubernetes, so meshri converts those compose apps to kubernetes apps.

C

The second piece of functionality- that's provided directly in the extension is the ability to deploy a service mesh.

C

And as soon as we get the sound effect in, there will probably be done, but you can deploy your service mesh here with the check of the button which you know sometimes is like absolutely not what you want. You really want to configure this quite deeply um or um a lot of times. It actually is exactly what you want if you think about it. One of the nice capabilities of capability that I use of docker, desktop day in and day out, is its ability to run.

C

Kubernetes kubernetes is highly configurable and yet, to a large degree that doesn't matter for some of my use cases. I just need a checkbox so we're giving the same checkbox or a similar checkbox.

C

Yep and so that's what the dish was talking about earlier, so this is yet to be announced.

C

Go sign up if you want to try it out early, though we'll see if we can um get it to you,.

C

So good questions comments. Do people see other use cases that could be done? Would there be something more useful to show here like without having to go into the full-blown measuring environment?

C

Would it be nice to like run your most favorite performance test.

C

Wouldn't you know anybody have any thoughts.

I

The alignment of the browse button is a bit lower than the open measure button.

E

C

Yeah this whole, this whole section here is a little bit taller than these two, the two button, the height of the two buttons, aren't quite the same. The positioning isn't yeah yeah, there's a there's a long list, but in terms of use cases do you have any other ideas things that would be nice to expose right here, results from the last performance test.

J

I say: keep it simple, but.

C

Yeah yeah yeah.

C

It might be something like if you're part of the the flow of building your app and wanting to deploy it on a mesh like deploying the mesh good. um What about deploying your app on the mesh?

C

This is a start. Let you import it so.

C

Think about that, if you would um you're dropping your your comments, slack.

A

Okay, uh so we are at the top of the hour. uh Does anyone have any comments about the docker desktop extension.

A

I guess not, so thanks all for joining I'll see you guys next week. At the same time, bye guys bye-bye.

F

D