Layer5 Meshery Build & Release, 28 Oct 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Meshery Build & Release Meeting (Oct 28th, 2021)

Description

Meshery Build & Release Meeting - October 28th 2021

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

Find Layer5 on:
GitHub: https://github.com/meshery
Twitter: https://twitter.com/mesheryio
LinkedIn: https://www.linkedin.com/showcase/mes...
Docker Hub: https://hub.docker.com/u/layer5/

A

D, are we not okay, so welcome everyone to the uh mystery, build and release meeting today is the 28th of october, and uh our primary agenda for today is tv 0.6 release.

A

We have some outstanding tasks that we were. We are. We were tracking through these uh to the through the week, and the main goal for today is to get updates from each of you on the call and hopefully uh uh get a bit more context on where we are with the v0.6 release.

A

So since we are uh fresh off the fresh off uh reviewing the uh reviewing the design spec for uh seed content in measuring cloud, we can maybe start with that. So if someone who was on the call could give us an update on the seed content in measuring cloud, it would be really good.

B

So we just had a brief call moments ago, and we talked a bit about some of the enhancements in this spec that we could have to make it more uh long-lasting and things like uh mystery cloud. Would the docker image that mr cloud ships with it would have the seat content in it and it would get registered as it starts, I mean not exactly starts, but as the release channel works, it would get the new content in the tables and some other things that we still have to.

B

I mean draw like the design of exactly how everything would be executed.

B

We are halfway through the design, but we still have to come up with a design on board with respect to how the flow would work yeah. That's it.

A

Oh so, oh, we are not waiting for it to land for the v0.6 release right.

B

I don't know like we can't still go without it, because I think mystery cloud releases work, independent of measuring, so misery would work with the latest release of the cloud provider anyways.

B

So I guess it would be fine if we make a release without that. That's that's my opinion.

C

Yeah, if we go forth the feature write-up for you know what patterns are filter, filter management et cetera, they need to include um you know, simple examples of how to take sample content, seed content and just do it manually.

C

And who knows by the time that we're done writing up that those feature blogs? You know those feature that those docs- maybe you, the second round of review- will be done for the content, um it's about as much effort on getting a good design as it is on at least the initial, the the initial part of the implementation that would get that would provide anyone who signs into mastery with some some content.

C

So so it's possible. So it's! The short answer is let's proceed at um under the context that it will be a write-up.

C

People will have to manually upload there, but rerocks. We should strive for as quickly as sanely possible, as as we can sanely get the right design.

A

Yep, that was my thought process as well. So if we are not adding seed content in measuring provider, we should document where users can get seed content in measuring docs. So what exactly should we have? Should we create a guide in mystery docs which walks through like importing seed content.

C

And then um the blog posts that kind of go we go over it high level and or you know, if there's enough time should we write an individual blog post just to go through all that, but but either way it's something of a regurgitation from what you would find in the docs yeah. Initially it needs to land in the dogs.

A

Okay, uh I was just writing this down so that we don't forget it.

A

Alright, so I think um rudraksha do you.

C

You'll get a second review now you'll put a meeting on the calendar for some time in the comfortable future, um and that doesn't mean you. You know you can't pull it in earlier.

C

um If you, if you get there, you do want to send it out to you. I think I think all of those that were on that review and all of those that will see your diagram.

C

You know we'll be able to digest that asynchronously.

C

So this is something like right now. We could certainly put this as a topic of discussion in the end for next wednesday, because we should be done like our second design review should be done. It should be part of the way through implementation somewhere in there.

B

Okay, make sense.

C

A

All right, uh so the uh other one other issue uh with helm shots in missouri ctl was uh it was not respecting the pinned version in uh in mesh config. So this this issue has already been uh erased and it is not a super urgent fix. We can fix this uh in a patch after the release as well, uh but uh related to helm, charts uh one of the uh one of the one of the things that we need to do was using help charts to deploy measuring operator in the in measuring server.

A

So I guess rudraksha was looking into it and has some issues proceeding with that uh rudraksha. If you could mention that here.

B

um So I guess we're talking about the operator thing. If right, I was lost and I was lost in the notes that she was writing.

A

Yep yep the uh deploying operators through helm, charts.

B

Okay, so from from the conversation that I went through in various places, I guess we are making operators the main chart and the others are sub charts, and uh I also had a brief, like back-to-back conversation with asuka on the same pr, and he was he was telling that he's moving out, I mean fixing the helm charts, so I guess once that's done.

B

We will need to make some changes in my open pr to not install charts, individually and message. Server would then only like. We can then skip installing misery. Server and operator would be installed as it is.

B

So I guess that would be the thing which should lead us into mystery server. Installing yeah.

C

C

It do we know who's doing that who's um moving the operator into the base chart and who's moving the server into the.

A

Yeah yep, so uh I talked to asuko and uh he mentioned that uh he will make a pr by tomorrow and utkarsh was also looking into this, and he had mentioned that he will check in with asuko and see if he has any issues with proceeding and if he has uh utkar mentioned that he can take care of this today. So I don't think utkarsh is on the gold right now, but I was hoping to get an update from him.

A

Right now so maybe we can get back to him after the scholar if he joins in.

A

So that would that would be the only change needed here. I guess like once that, once we move uh the measuring operator to the parent chart, uh rudrash can proceed with spr and.

A

We can go ahead with the release. I guess nothing else is holding up.

D

And no word back from darren.

A

A

Were you able to follow up with asuka on uh moving the uh changing changing of the helm, charts.

E

uh Not yet actually, as I mentioned, I was actually I was going to wait till midnight um and to get the updates from asuka and so yeah. I haven't yet reached out to him. Oh yeah,.

C

Does he know that um have you reached out to him at? Did you reach out to him the other day or like? Does he know that he's supposed to be working on it.

E

Yeah, actually, he mentioned on slack thread that he he is going to pick it up and going to wind it off asap.

A

Yeah he had mentioned that he can make appear by tomorrow, so.

C

So this is confusing to me. This is already done right.

C

We already did this, but we didn't land it because it needs to be refactored.

B

Refactor yeah, okay,.

D

So the buck stops with cars.

C

uh If a seagull doesn't get to cool all right, sorry to whoops, I'm sorry to interrupt.

A

Yep, uh so these were the uh remaining items for the next release, uh so I don't think we missed anything here like uh this was the things that we, these are the things that we discussed uh this week, and uh so what we can do now is uh maybe uh put put down names to some of the things we have to do. So if anyone can pick up the pick up, the documenting uh documenting using uh the seed content in measuring provider or importing seed control in measuring provider.

A

We can do that now and try to get this get it over in a couple of days.

C

Yeah, it's actually a really good um thing to go through, because I suspect we might find a bug or two so there and there's two ways for each of these pieces of documentation.

C

There's there might be two names needed in part well, just to help make sure everyone's accountable and can have has a has a qa buddy has a docs buddy, but one of those ways is measuring ctl. The other way is mesh for ui right. They would be covered in the same dock, but.

C

And there would be three docs right like.

C

For each content type, so mr.

C

Singing singer, mr singharia, I wonder if.

C

Just look how snuggly your name fits right in here. It's like it was made just for the square back. It's that's! Thank you, pierce. That would be. That would be lovely. Do you would you mind? um Would you mind.

C

Like this isn't about measuring ctl reference guide, I think we have filters apps and patterns covered there, that's good! This is about um under functionality. It's either a write-up under functionality like pattern management. If we go there in here, it talks about like what what are patterns. But what it's missing is: oh wait. If you go back up a little bit, okay, to deploy a pattern, execute this command like good. Now I think our experience there might have changed a little bit. So there's two sections: the ui and the ctl.

C

And so push do you think you could take that for.

C

Any of the content.

F

Like what exactly needs to be done, uh can you explain a bit more yeah so, where it says, measures including the dogs is the is that yes, yeah.

C

Yep, it's just improving the dog so like on the pad for patterns. It's like making sure that what it says is accurate and still and then to maybe explain some other things like. Oh, if you wanted to delete it later, you could or it's it's a bit more of a guide you're, providing a guide to the person who's using it like explaining what to expect if you do apply versus versus whatever right.

C

If you would like, if you end up focusing on measuring ctl in that way, you could then also there we're missing a page one called wassum filter management, probably and another one called yeah.

C

So this is something that um done wrote uh um and did a good job. This just explains the gender tech. This doesn't explain what mescheri does. This is good. I mean it kind of touches up against measuring, but this is great we do need. um This is just the faq we need. We need the other part of the dock.

C

Okay patterns filters and then kubernetes applications. So if you go back for a moment, thank you. There's a sample, there's a guide, there's a section called guides. I think yeah and deploying sample applications. Okay. This is a little bit different. This isn't about application management. I mean like there's a different thing like in it's nuanced, but you write these a little bit differently. This really isn't some.

C

I guess this is a guide. Not really a guide is more like a stepped thing. You're saying: hey, let's, let's, let's agree that we're doing a goal: uh let's walk through how to do this and explain it as we go, whereas the functionality section is is much more cut and dry.

C

So, ideally, we would have both guides and functionality for for all of these um so root push. We do need.

C

An application management page under functionality that has a section on measuring ctl apps.

F

It'll be a collaborative yeah I'll. Do the mesquite secure part, and I expect someone else to do the machine, ui thing and yeah. I'm adding this to my tool list.

C

Does anybody else want to do the ui.

A

Thank you. I I can work with the ocean. Get this.

C

A

uh Leave by the release announcements, uh you mean the the release blog that we are okay, yeah.

C

C

Anybody feel nervous about the release, there's kind of an aspect that I think keeps on getting swept under the rug a little bit and it falls right into obby's lap on the pattern. The pattern configurator.

C

If we do go out and and tell the world hey we're doing you're doing this thing around patterns, you should come check it out.

C

We either need to have the mesh map beta there, so they can really like use. You know do something with a pattern.

C

We need the configurator there.

G

So uh I'm actually working on the configurator only uh giving my prior focus to it and uh it will be done by today or tomorrow, hopefully uh fixing some bugs with the slow response. Time in the react. I don't know and yeah.

G

Are getting the new ui done for the configurator, so uh I would need the better choice for icons. uh That's the only thing that is required right now.

C

Yeah yeah, I'm sorry.

G

So the icons would be from mui or we would be using custom icons because there are very limited number of uh configurator menus like what should I say, the accordions uh like authorization policy or uh so let me see yeah yeah, so.

C

Yeah, you know icons are needed, so you know. Hopefully everyone understands that, like the pattern configurator part of the design calls for having um each of the service meshes custom resources available as a an icon.

C

To kind of so maybe that's not a good design, but it is it's what's written right now and as such, what we need is a visual representation of things like an authorization pod. All of the custom resources for each of the service meshes.

C

Some of those custom resource sources are one in the same or they're close enough that we use the same icon. I don't know we might be hurting ourselves by trying to do that. We might.

C

We might just want to use a drop down list of names of the customers, something I've had.

C

I don't know I'm not not suggesting that that's what we use, but just calling it into question right, but right now, though, under the current design, there's kind of a blocker or dependency on icons.

C

Outside of that.

D

What other, how else, how else are we tracking.

G

uh So uh if you ask about the the track is like, uh I have uh made some github check boxes in the issues, uh so that was the king, the way of tracking for me. But let me open it once again.

G

Yeah, it has certain items that needed to be done and I'm just going to.

G

uh Some items have been fixed and uh need more yeah that the performance optimization is the only concern because uh there is actually more than 700 uh list of items that comes with istio adapters and only when there is only easter egg running so are there is chances that, when someone will be running more than one adapter like uh all the adapters in this case, then their ui gonna be freeze, or it would be like very rough experience for them.

G

uh So working for that only uh there is a lot of state management being done here and we may not require all these things, but yeah a need to handle. Oh wait. um That's the only way to handle this thing.

C

Can you can you share your screen, real, quick or is somebody sharing what that list is.

G

Okay, uh just wait.

D

Share my screen.

C

Yeah you get in yeah.

G

uh So not confident with screen share, it is, is it vs code.

C

Is it? uh Is it bs code? It's um github pattern: okay, okay, sorry.

B

D

G

Yeah, so uh if I show you the current conf okay, I I was just our you know doing the thing like developing it and it failed.

G

Okay, that is previous ui, so uh I couldn't show it right now, because there is a lot of issues right here in the import statements, uh so we'll be fixing it in like another five ten minutes, so we'll basically sharing my screen.

G

That's fine! Okay,.

C

Yep sure, thank you, cool yeah yeah. Take your time. Actually maybe you know take your time and um and when you have that final well, when you have a screen, if you do just a quick recording like uh that, would be really helpful and if you have a link to that issue, when it has accurate checkboxes, I mean maybe they're accurate right now, but whenever it does pop over the link as well. That'll just give some indication as to where we're at some of what you're working on is.

C

Well, it'll be iterated on all right. What I mean to say is it doesn't ne. It has to be functional, not necessarily perfect.

G

Okay, okay, I I will be marking it as ready for review and we'll be uh adding this link to the slack also after the work has been done and uh making a short video explaining that other things and how things are working nice thanks.

C

Yeah all right um hey, so I tossed it so we tossed on another question about service mesh patterns, github action that uses measuring ctr. So there's a new repo opened.

C

You know very recently a pr opened up, um maybe we'll just put a link to that repo in the meeting minutes, and I think it might be a little too early to talk about.

C

Progress, it's probably healthy to just regurgitate the goal is to allow people to pipeline or automate use of nashville ctl.

C

You know, then, all of its sub commands, you know, apply a pattern which then like there's a lot within there. That might mean that meshri is connecting to kubernetes system, deploying a service mesh running a performance test like potentially any number of things that might be applied.

C

Potentially importing a pattern, someone may offer a url that they want to import temporarily.

C

Someone may say: look, we know we have a pattern that we use all the time and we also have our own application that we use all the time. So I want to.

C

um They might want to deploy applications.

C

If the pattern itself incorporates an application, that's sort of one and the same, we really have to get um a diagram. An architectural diagram that kind of shows how different content types are related and the fact that any given pattern might like the way that we've been using the word pattern that it might incorporate any other types of constructs.

C

The interesting thing is in mesh map we're talking about mesh map, a given design incorporating all you know, patterns and applications and filters, and all of these things outside of mesh map, though, when you're just using mesh for ui.

C

That experience is still the same, there isn't the third, there isn't a.

C

There's only patterns, and so I think that term gets conflated and confusing, because we don't have the concept of a separate template or a pattern. That is a template on editable versus um an instance of a pattern that you've created. That is no longer called a pattern, because the word template and pattern are synonymous rather.

C

You've been presented with a list of templates a list of patterns, and you want to create an instance of those. It's like a docker image. It's a an instance of that image, a docker container running, and so, if we have two terms to help us clarify that, there's a pattern template and something else.

B

Think about that.

C

Good adina, I want to say yes just I'm not sure that that's.

C

Yeah yeah, we do want to have a catalog of patterns publicly available people can use mastery ctl to import them. They can then use mastery ctl to apply them and they can kind of do that in one in the same motion or they can do it in two different motions: um yeah and that's probably the core goal of the github action.

C

Fortunately, we have two other github actions that already wrap mesh for ctl they've already had to deal with like tokens and other measuring ctl-specific stuff. So it's helpful.

C

We do have some definitions of terms that that will help our on patterns also on mesh ops, we're trying to help define those terms. If anyone saw the latest message in the messaging adapters channel, there was a definition of mesh ops. If you want mesh offset v2.

C

So good, um you know, at least for me. I think I'm walking away from the meeting feeling pretty good about where we're at with our most critical item, which is refactoring of the helm charts, I feel pretty good um that actually adds a little pressure to the writing of the release. Blogs writing a bit of a bit of documentation around.

C

It gives bobby a little more time to kind of wrap up and then solicit feedback for the pattern. Interface.

G

Yeah so yeah, so I have the pattern interface ready for now. So can you share us this game just.

G

Okay, I hope it's firefox now so right, uh-huh yeah, so uh yeah also as discussed in the figma design. So uh there is this uh where we can select adapter, so it will only show the list of available adapters. You can choose from so for now, only histo is running. So yeah only stressed.

G

G

Sorry today, once again.

G

Yeah, it just doesn't work in the first go ah so giving it the context uh so yeah. This is how it will look so here it here. There will be a toolbar and there will be list of options here, so that will contain all the options. For example, here applications and kubernetes services. Then there will be two icons here, which is one for application, and one for community service and uh yeah.

G

So thinking upon uh the colors and then uh like, will the icons be colorful, as we have in the mesh map latest design, spec and yeah. So these are the things and do we need any changes in the in the way this uh this rhs form is rendered.

E

um So obviously, just one thing that I noticed- which I uh which I have to mention right now is um so the name of the service is on line. Three is uuid and that's fine, because we recently made some changes and that's feasible now, but uh we also have to ensure that there is a new field uh either below line four or below nine five. It has to be a sibling to these fields, type namespace and settings, and it should be the it should be.

E

The field called name which will be, uh which will be the actual name. For that particular thing, the reason that is required is yeah. Basically, that is sort of required, because it could happen that an authorization policy doesn't accept the name with hyphens in them and so yeah. uh The the the point is that earlier it was like like on line 19. We have something called application and that's completely fine. That would work just fine uh and that's a and if you would see your cluster, you would have some an entity called application.

E

That's fine, but on line three, because it's a uuid and it's probably auto generated so so, for if the user wants to customize the name they uh what they will have to do is have to add a new field below type or bidder, namespace called name in which they can actually type in a custom name of that particular thing, and that name would be reflected in the cluster. This was a new thing that was added to patents, so yeah. That was just something I had to point out because.

G

Okay, uh so it should be a sibling to type and namespace or type type.

G

Or the name should be different, like this: a.

E

Key and so name field could be anything uh by anything. I mean uh something that the user probably entered, so uh they are creating an authorization policy for seo. uh It could be admin only policy or something a a name that they actually want to be represented and represented inside the cluster, because what will happen right now? This would be a valid pattern. This is a valid pattern. What would happen is that my authorization policy would be created with this name, that is fdb86648 and yeah.

E

G

G

Okay, look into things.

C

You know I'll be. I gave just a couple of of comments in the chat that I'm hoping will make it easier to land.

C

So because kubernetes is something of a universal consideration for us, it can live outside of the mesh selector.

C

And then, let's do ourselves a favor and at least initially avoid trying to figure out icons for all the customer resources and just use some standard. You know selector like uh having you know, yeah.

G

C

Nice is everybody feeling, okay, everybody feeling comfortable or or any any question. I know, there's a lot on some of your plates.

C

uh So outside of that stress, are you feeling comfortable with what you what you need to do going forward.

A

Oh one more thing that maybe like fits in here is uh some of the work going on around service mesh performance like a performance benchmark test. So I have been looking into customizing the smp action to run some benchmark tests on github runners, so yeah that is in. That is not exactly related to this mystery release, but uh something we should get out soon.

C

Yeah but but this is the meeting that we talked about, we talk about github actions, and so um so do we need, can you do we need specialized runners, or can we just use the runners that they provide us to start.

A

Yep with we can use the runners they provide us like uh uh unless, like we, we, there are some like. If you define some a set of considerations for running the benchmark test, then it big something becomes a concentration but like.

F

uh Yeah, as far as.

A

As far as I discussed with some of the other maintenance smp maintenance like we can start with, uh get.

C

Up uh wait, you mean not, okay, not doing it. There shouldn't be any more work. I actually I just so just start, that's what I'm saying just start just schedule some tests and go and then whatever runner they give us just document what that is and and that's the test so yeah. Please don't do any more engineering in that way um up until the point that, like once, we get the test results back and then eventually we say well geez.

C

If it only had four four cores, then we could really test it and then we'll deal with custom runners.

C

Right now it's been it's hard enough to get um seasoned professionals who get paid lots of money and to be engineers that to even identify like what an appropriate test environment is like because do you know why? Because the answer is whatever the user is using, which is whatever that is, which is whatever the most popular ec2 instance in the world is like. That would be the one that has the strongest signal. So it's like the answer is they all are valid, all of them they're all valid.

C

Some are much much more meaningful to people than others, but they're all valid.

C

Is it valid to run it on your dock or desktop? Yes, that doesn't represent fraud at all, but if that's, what developers are using to try to figure out what is going to be performant in prod benchmarks on that is helpful, so.

C

I have to bite my tongue because I want to make the point like two more times cool. That's great! That's perfect! Yeah! Please I'd love to see the 5000 tests jump by a few by 10 every day.

C

Yeah that'll give us really something to talk about in s p meeting as well.

A

uh All right uh uh anything else, we missed our and does every every item on this list. Have people assigned.

B

A

B

C

Yes is the answer to your question. That's a great question, too: cool. Hey! It's been nice to hang out with a lot of you today all day long. We had a lot of meetings. Today we did some great stuff what a great presentation from jacob, obviously clearly, some interest from a variety of parties, cisco, cisco, was on the call the intel is on the call red hat was on the call multiple people from intel. uh Google was on. The call adina was on the call. So then you know.

C

uh But but really really interesting stuff, I thought that the presentation crystallized the plug-in architecture and hopefully unlocks in some people's in the rest of our minds. um You know how where we would go, write some code and what that code could potentially include it. Does the interest around that project does call it was?

C

It was I thought, heartwarming to hear jacob refer to like when anirban who's, like I think of this community, or he's only involved in this because he's in this community, when he was asking where to go file the where to go, get help that jacob had said had included layer five as a place to come to do that and um and that's nice.

C

But I guess I'm highlighting the fact that we still haven't landed, get nighthawk.dev, and so I've hinted at this with a number of you, but um that I think I think, ideally the site itself is called get nighthawk, but the logo we need to take the get out of it and just say this: green bird-looking structure is nighthawk that we're not talking about something separate, we're talking about nighthawk, and so I think after we clear up the site, remove the get.

C

As a matter of fact, I think we just removed the get and just didn't even confer with jacob and otto and uh the other guy harry, if we'd, even confer with them that they would just accept it, like, I think, in their minds. They already have they. They voted on the design they voted on the they have access rights to this repo and stuff, so yeah.

C

So as we go to build that anyway, as we get the build notes and other issues and stuff like, let's fill it like point nearby on here- let's fill it in. I know we've done some of that. So he's not.

C

Hump day was yesterday, so he should be.

C

No, no, not a wednesday hump day thing.

C

Trying to figure out yeah, okay, good yeah, we're actually smiling. Okay, all right, very nice. To talk to you all see you later navendi see you see, you see.

D

C

Thank you see you bye.