Layer5 Meshery Build & Release, 24 Mar 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Meshery Build & Release Meeting (March 24th, 2022)

Description

End-to-end testing using GitHub workflows and actions in the Meshery project.

Meeting minutes - https://docs.google.com/document/d/1GrVdGHZAYeu6wHNLLoiaKNqBtk7enXE9XeDRCvdA4bY/edit#

A

And uh yeah so, okay, so I guess there is only a single topic uh today, uh although I had asked people to add their own topics so, but I guess we'll go with this one. um Let me share my screen. First.

A

hmm Okay, so today I was going to discuss uh the end-to-end test that we have uh in measuring and uh investing adapters. So specifically, although the core workflow uh that sits at the bottom of everything, is here, if you go inside the dot, github workflows and test adapters.yaml you'll find this workflow.

A

uh So first I'll go over what these workflows are doing and then I'll go over what pattern? What how we are using these workflows, so the uh a lot of events there are that allow us to trigger a workflow, uh one of which is a workflow call. So when you create any uh github workflow with this thing set as an event, it's more like a function. Definition where you know you like. You define a function where it takes these parameters and then it does this thing and that you can call later with those arguments.

A

So in the same way, you define your inputs here, like uh what pods do you basically different? You can define your own stuff, so we have defined things like what pods we are going to expect, because uh this, if you uh think of this workflow as a black box, all it does is it takes a pattern file, and you can imagine pattern file is something that deploys stuff. You know this stuff.

A

We will discuss later on what this stuff is, but this can be your pod deployment a bunch of stuff, so it takes your uh pattern file and pattern file. Just to give you an idea how patent file looks like I'll I'll show you an example, and so because everyone should be familiar with uh what a pattern file is uh to go ahead and work with this thing, uh so that it's easier for everyone so like, for example, this is booking for pattern. This is an example of a pattern file.

A

uh Sorry, my internet store yeah, so you have a bunch of services and each service does I'll come back to this particular patent file. At the end, after I go over what the workflow is and how it is uh being used, so each pattern file deploys a bunch of stuff so like, for example, this bandwidth deploys your virtual servers. It deploys your application, etc, etc, and uh when it deploys after it is done, deploying those things we expect something.

A

So when we give it a patent file, for example, when we give it a patent for this patent file, we expect something like a product, a particular container, a particular pod named product page to come up. So when we will actually call this workflow in the expected parts, we would set things like uh you know, as you can see in this particular. This is where we call it so to call it you have to inside of your needs field you have to uh so. Instead of uses.

A

Sorry, instead of users field, you have to give the path uh the path to that particular. So, for example, here it is measurey, slash mesh, and this is where the actual workflow is, and this is that workflow and uh the inputs here, like expected, pods expected namespace. These things are basically filled here. So what I mean by this thing is that uh when you deploy the patent file, we will be expecting a issued pod inside of his theo system name space, and then we will be deploying istio egress.

A

We would be expecting seo egress gateway pod in this name space, and then we would be expecting this part with this name inside of this namespace and so on. So basically there's a one-to-one relation between expected ports and expected namespaces, and then you have your deployment url, which is um I'll. Come back to this I'll, come back to this, so okay. So this is where your core workflow. Is it?

A

uh If I go over this thing really quick, it sets up your mini cube cluster and then it builds your uh adapter because we have what we effectively are trying to do here. Is we want to test the functionality of each adapter? So we want to uh what so, for example, you create a commit okay, so on your particular uh sha on your particular commit, we will create the docker image of that particular adapter. So, first you know the first check is: uh did you do something such that?

A

The image in the image uh cannot be built, so the first thing is you: it will build an image and uh after it is done, building an image. It will start your adapter. So basically it will. You know with the image that it just created uh with your changes.

A

It will start the adapter, it will start the imagery server and after uh doing all the you know, boilerplate stuff, uh it will wait for like 300 seconds and it will start expecting the points, and here there is the whole logic of uh you can go over and see what the logic is. But uh to summarize this thing, whatever the pods you expected it will you know it's like a loop, so as soon as it, for example, in this case, the expected quads are skewed as the egress gateway and it's your ingress gateway.

A

After as soon as uh it finds these three parts in this namespace it will. The workflow will exit with the message that yeah it was.

A

It was good everything worked and even after 300 seconds, if one of these parts failed to come up in this, these uh corresponding namespaces, basically, the workflow would return, but with a failure message that hey things failed, so that's the this core workflow does it takes your pattern file and the java, because you are giving it the pattern file. So you know that this pattern file is capable of deploying these particular boards in these particular namespaces. So you expect something and if that doesn't come up the workflow phase.

A

Now you call this workflow here now it's a uh now here the whole this uh adapter integrate the core workflow was defined now in each of the adapters in istio adapter and linked all the adapters. This same workflow is being called in different particular ways. So, potentially it's a three-step process. If you look go over to mishri istio and dot github workflows, e2e tests.yaml you'll find that it's a three-step process. In the first step, you actually set the pattern file.

A

Okay, now before I talk about setting the pattern phenol, let me go over and what uh talk just to give five minutes over what a pattern. Five is uh I'll, not go too deep in this thing, I'll summarize it to five minutes I'll, give you enough information so that you're comfortable working with patent pipes, because potentially that is what you have to do if you want to work with individuals. So this is your patent file. Patent file has a bunch of services.

A

uh Just like you know if you have seen a docker compose file, it's kind of similar. So here you have types you know, uh and inside of uh you, people can see my screen right.

B

Yes, it's visible.

A

Actually, I just got a message that participants can now see your screen, so I thought I was I I had lost so yeah coming back so here, as you can see that we have different types defined and uh just to hint at you, uh when your meshy server starts, uh we have a bunch of components like we have adapters and what those adapters do is they register their capability.

A

So this capability basically means that hey inside of your patent file, if you have this particular type, I can handle it so just to hint at something if it has dot istio. uh It kind of means that you can take this as a hint that is.

A

The adapter has registered this thing and it basically means that if you deploy this pattern file, you would need ski adapter to be there, because uh istio adapter has said that hey I'm going to handle it so istio adapter starts and it registers its component and if you get a message like this type of so, if you deploy this pattern file with the command message, ctl pattern deploy, and uh just like you know, you have, you know, f flag, and then you give the file path and all so.

A

If you try to deploy this thing and if it fails with the message that this is undefined, it means you should actually have your rescue adapter started up, and you will see these things in other places as well. But you'll find some types where you don't have the dot ist or dot something. That is because that, when you have dot is q, it basically means that that that's an istio specific resource, as you can see, you know this is http and uh the whole.

A

This is a virtual service spec that you would write in your uh in your manifest and if you have something like application, the there are some core components as well. That measure natively supports you don't need to have your adapters and everything so applications would work fine, okay, and if something is like service dot k8, it means that you would basically need your when kubernetes starts up. We also kubernetes also register its component, so deployment dot, k, support, dot, gates, et cetera, et cetera. So this is your patent file.

A

It deploys a bunch of stuff virtual service, dot, sto deploys a virtual service. uh This application results in uh a service and uh uh board internally. We are using our rollouts right now, but we are not using. You know. There is no things. No such thing as we are not using strategies, you know camry, you know. Basically, it just deploys a service and it deploys a pod for because that's a common pattern. You know when you deploy a pod, you deploy a service for that.

A

So, instead of deploying two things, you can just deploy an application and it will take care of it, and uh this is a machine code component. So this is your. um This is your patent file so just to go over this pattern file? How? What what things would I expect so, for example, I would.

A

I should be expecting that I should, when I deploy this pattern file, there should be a port uh that comes up with the name details uh with the prefix details and there should be a port with the name product page, and this would be pod name ratings because application internally create ports. So when I looking at a patent file, uh you can deploy that patent file, you can go over and see.

A

Cue cut will get all and see what all the ports that come up and then you can put that into your expected pods here, yeah your expected pods here. So uh coming back to our uh original, you know this adapter workflow. You have three step process. The first step is setting a pattern file right, so setting a pattern. File basically means you know this thing. This pattern file that you're seeing right now creating this thing, so every adapter has a dot every adapter. If I open uh workflows, uh it has an install.

A

If I go just a little back, it has a install directory, dot, github folder, and here we are saving the patterns uh against which we are going to run our tests. So right now we are not using this pattern. I have specifically chose this pattern because we want to use this pattern for this, because this pattern actually goes ahead and deploys your book in for service right now. If I look at the what pattern we are currently using, as you can see, we just deploy an istio mesh.

A

uh This is, as you can see, someone might ask with the kenai that why don't we have the dot something prefix, it kind of hints at the fact that this is a messaging code component again, so we know how to handle this image and what it basically does is that uh here, as you can see, the version field is empty, so uh in the core workflow we actually and not in the code works actually.

A

But if you go here and look at the set pattern file for this thing, we actually get the latest version of uh we get the latest version of istio, and then we uh basically populate this here. So we have a proper uh pattern file and, for example, in the case of deploying your sd mesh, you need to have things like you need to give it a version, and you need to give it a profile. So for people who have deployed this you're using steel cutter, you you have that high profile flag.

A

So it's basically that you have your demo profile. You have your. There are two other profiles default and minimal, so this one uses the demo profile so that we've been, and that is why you we are expecting ingress gateway and egress gateway, pods here, if you yeah, because uh in the minimal installation you would just expect one port.

A

So right now we have uh two patent files against which we test the deploy.tml and uh when we deployed this particular yaml, then we expect three parts uh uh is theod is gateway and steve ingress, and then we have one other pattern file that we try to deploy and this other pattern file actually deploys your istio mesh and, as you can see, it also deploys graphing from ethios. So again, if I uh hover a little, you know if I come here as you can see in the next three, so there are total six particular.

A

uh You know six particular workflows here, uh six particular your um what you call stages here and at each stage you're. Basically uh it can be divided into three different stages for one whole functioning thing and the first stage is you set your pattern file uh as you can see here, and you set your pattern file using this particular file that you have in your install directory. So if I look at this thing, we get the latest service service mesh version, and then we populate that set because we have left the version field empty.

A

So this is an invalid aml. If you try to deploy this thing, you will get an invalid because uh invalid error, because your version field is not specified so basically the the whole job of your first. uh This particular stage is simple: to get the latest version of istio and populate it, so you have of your functioning pattern. File available after this thing is done. Okay. That is why, in the next stage, we actually expect this is basically linearly.

A

So we first expect this to complete to go to completion and then only we go to our actual. You know meat of the matter. So now we have your uh now, you have your uh pattern file ready, as you can see here, uh github actions actually allow us to. uh If you want to communicate, if you want, you can send file between stages so using this uh particular action that we're using. So after we populate that pattern file, we upload it and here in the next stage we actually download it. As you can see here,.

A

So, as you can see here in the output file name, it has it's. The name is data.json out this output file name is basically the one which is being used in the next stage. So again, third stage is update docs. So after you are done with the test, you push all the output to the output file. Name and yeah. Here is the pattern file name, which is deploy.yaml, so whatever you set here, I set here deploy.tml. So I have to expect here deploy.tml.

A

So deploy.tml is my populated pattern file. I got my patent file and I run all my tests against these expected ports inside of these expected namespaces. The result will come up. I have to give it the file in which my results are going to go. This is data.json. My results go into this particular uh json file, and then we download that json file here in the next step last stage, which is to update the dogs. We take that particular thing and if I were the compatibility matrix, we have a compatibility matrix. uh This.

A

The job of the entire job of this particular last stage is to actually go to measuring and update the documentation, update the docs, the docs, which you will see in a minute. If my internet is okay, it's taking way longer yeah. So here, as you can see, it kind of gives you an idea of the current stage of things. So you can select your uh channels table channel and all these results that you're seeing right now the green ones are the ones tests that are passing.

A

The red ones are okay, so something is failing right now and then you have your extra information like mystery server version and uh was it an edge or what is it stable, etc, etc? So you get this data is actually populated at this uh last third stage right, so we have three stage uh as a working as a unit. First stage sets.

B

The compatibility metrics is it updated in in real time as the texts are running yeah.

A

Actually uh yeah this documentation, uh the workflow to actually update this documentation is inside of your uh message repository so whenever a commit is made there, a workflow kicks in that builds this thing. So when we create a so inside of this third action, we actually create a commit inside of your inside the mesh repository.

A

So when we, whoever merged that request or whoever merged that pull request, it would be basically okay. So let me a good question. Actually you asked because I skipped over one little thing and that little thing was this: these two stages are executed every time. You create a comment on your pull request, but we don't want to update the docs every time you create a pull request right, so this third stage is only done when the github event name is not pull request.

A

So, basically, on your mars on your release, then only the third stage will uh kick in and kicking. You know. Third stage basically means that we would take the results and we would create a commit inside of the machine repository and whenever a comment is merged. Inside of the message repository, which it would be inside of this particular stage, uh this documentation will build and you will uh within a minute or two or whatever time it takes you'll, see the results here.

A

Oh yes, so the answer is yes: in real time after the pr is merged or a release is made, so when a pr is merged, that thing is pushed onto the edge channel uh when a release is made, you see the thing in the stable channel, so if a release was made with these, for this particular release stack, so that's what you're seeing here so and it actually gives you more data. Okay.

A

So, as I showed you, these were the expected points right, so if one of which, so we don't have like a binary thing, uh that it completely failed or it completely passed. So if like three parts started but two didn't, so we actually uh have that thing registered here. So if maybe the grafana add-on and prometheus add-on failed, so it would have shown cross there, but it passed.

A

So it shows tick, and the same goes here because these all these three failed for some reason, which you know uh it is a good reason for me to go ahead and look at why what this was feeling, although I know the reason there was some other reason uh so yeah you get, the everything happens in the real time. These results are updated in real time, so in linkedin, adapter falco.

A

These are the expected pod names linked your destination, linkedin identity and linkedin proxy injector, because I know when I install linkedin these are the things that I expect to come up so yeah. Does that answer your question.

B

A

uh Yeah, okay, so uh yeah almost I'm uh done with this. uh Is there any okay? So, for example, in this particular uh repository in this particular install directory, you would see two patent files right. So each pattern file is composed of you know the test for each pattern. File is composed of three stages.

A

As I told you setting the pattern, file and testing it, then updating the docs here you have two pattern files, so the first three are for the one without the add-ons, and then you have again setting your pattern file testing that with now this time with add-on and again updating. So uh basically, we would need uh three more stages here, which would be kind of a copy paste.

A

You know you just take the uh these three set pattern: file tests and all the things that you would need to change this, because this time the pattern file would be different. So here, instead of deploy with add-ons, the first thing you would need to do is you have to create a pattern and you have to uh add a pattern file here in the install directory, and I would suggest you to go over the I'll. Actually, I should post a link inside the chat, so uh I'm not gonna see. Can someone else post the?

A

uh I think the chat, because I think I'll have to switch screens for that. So uh if anyone else uh can please post it, it's in service pattern, service, mesh patterns, and here instead of samples, you'll find a book info pattern. So you have to take this thing.

A

You have to come copy, paste it uh and create a new file inside of the install directory and paste it here, and uh once you do that uh inside of these workflows, uh you have to you know, give it that file name uh instead of uh uh instead of the uh you know the deploy with add-ons and because uh you don't you now here's the thing you won't have to. As you can see here, I'm using yq to update that you know inside of services instead of istio settings and version.

A

So, as I told you, uh if I go into one of these files, these fields are actually empty. I have left the version field empty, so I go inside of services, istio settings and version to update it while the workflow starts with the latest version. So here, as you can see, services is the settings version and I go ahead and update it. This pattern file does not have anything any such you know, version or anything, because uh this pattern file. Okay, so here's a catch. This pattern file has things like virtual service right.

A

So if you have something like virtual service, you would need your istio control train to be up. So one of the prerequisites to this patent file is uh you'd have to have your istio control plane up. So would you need another patent file? Well recently, a support was added, so I would suggest instead of using this one, you can use this one minimal big book info. Actually this is uh this pattern is quite compact and easy to use, but it doesn't give away.

A

You know a lot of a lot of fragrance of what this is actually doing. So this will take care of your uh what it does behind the scenes. Is it first actually deploys your istio and then it goes ahead and uses use the same booking for pattern that I just showed you, so I would suggest you uh so you'll have to do some manual work, so why would you do that when we have that support in patent so instead of book info? You just go ahead and uh use this minimalist.

A

Your book info.yaml you- and this is a this- is a self-reliant kind of pattern. So when you deploy a pattern, so if you have thing like virtual service, you might be expecting skill control game to be already up now. This pattern file actually will go ahead and uh deploy your list you. So of course it would need your current days to be up.

A

So that's the only requirement and there is no other requirement so because of that it would be easier to not get bugs in this case, so you should go ahead and use this one and you just copy paste it, and uh if I, if I let me just just first for the sake of an example, let me try to deploy a patent file. uh Just to you know, go over what would actually happen. So if I do patent apply, this is the command uh I have given message.

A

Ctl pattern apply and then the url to the minimal istio book info. So if I, uh if I it takes a little time, so you should go ahead and watch what's come up. uh The reason I do this is because you know I first deploy this pattern file on my system. I make sure that I'm not getting any bugs. If I get bugs. I report it to people that hey this patent file is not working.

A

There must be some issue and if this patent file is working then so, for example, as you can see first, my stew port came up because that was a requirement and here in uh just as a you know, quick note so people uh who want to know more about patent files. We have a depends on field, so that kind of allows you to. For example, here you have a booking for patent, and here you have an issue installation, so we don't do things sequentially. So what?

A

If your book info pattern kicked in first and your issue, installation kicked in lag later, so that would actually fail because you're looking for expectorative watts to be up. So in that case you can depend on this particular service. So, just like you know in your workflows, you have stuff like you actually uh like you actually depend on some other thing needs. This thing needs uh it's almost similar, so you basically uh yeah so you, basically, your istio installation would be succeeded and after it's done, your booking for app will be uh yeah.

A

So it would take a little time because my so after your istio quad was uh deployed now my uh applications would start to come up. uh So, as I told you, we use argo rollouts. So first, my uh the our product pod will come up and then that's kind of that acts like a controller, so its job is to go ahead and start your.

A

uh What so just wait for, I guess 30 seconds more uh and that's why we also have a 300 second timeout, so that you know we can allow things to come up as they do so, as you can see here, uh all the parts came up details, product page ratings reviews.

A

uh These are the ones we are expecting right. So what does this tell me? This tells me that if I deploy this pattern file, I should be expecting a port, starting with the name details in the expected team space default uh port with the product page expect in the name space default ratings default. You get my point so inside of uh translating this in terms of workflow. Just like here, you have your expected ports. uh Istio is.

A

Inside of your system, namespace, let's use the minimal installation. Why do you use egress and ingress gateway right? Just we in this case we are using the minimal one. So you only have this the old, uh so you would be expecting this theore in the system name space and after that you would be expecting reviews hyphen, v3 in default and reviews hyphen b1- and basically you get my point so here instead of expected, pods you'll have these things and expected pod names.

A

Pieces you'll have is your system for the seo and then a product page and so on and so forth uh other than that uh you'll have your pattern instead of deploy.tml you'll. Have you should have this pattern file? A minimal book is to pack file and uh because this pattern file also has a istio version, so uh you should be. What you should be doing is that if I go over here, we actually use iq services, dot, seo dot, settings dot version here the path would be if I'm not wrong.

A

The path would be here in this case is dot services, dot, steel, installation, dot, settings uh version, so so that uh and you can um and that what that would do is basically uh fill in your. You know version with the latest sd version. So basically your instead of deploy.tml use that file, uh and uh in here you just change the expected ports to the reports that I just told you. You know, details product, page ratings reviews and then updating the docs. You don't need to change anything here.

A

It's it's kind of you know it's a copy based here. uh Actually you do um this. Is this little? You know, for example, uh instead of tests, the thing the compatibility matrix that you see here is a one to one translation of it is basically a markdown. You know what you see here is a markdown, and this markdown is created here. So it's you know. Instead of echo the whole. This thing is the exact markdown.

A

What you see here, so you know you see your time stamp mesh three component, meshley server, gates, distro, you see the same things here. You know mexican component message, server, etc, etc. So what you'll have to do is that inside of uh tests, you have your, you know your istio instead of egress gateway and stu ingress gateway, you change it to whatever your expected ports were here, so in the in that case. You'll have you know product page, so you will change this line to product page and then for underscore status dot.

A

Instead of again, you have to just uh it's like a control f controller, so you have to replace your these parts with the ports that we are expecting in this case, which is you know, uh details product, page ratings reviews uh and yeah. So a change here in this file uh in this particular here instead of tests, you have to change these fields.

A

Inside of the test, you have to change the name of your expected ports and instead of your just change the file name to whatever you name your file when you uh bring your minimal bookish to uh pattern inside of your install directory yeah. So that's uh that's the gist of how we are using end to end tests uh where we have the core workflow defined in mesh three. We use that across all the adapters.

A

Does anyone have any doubts regarding how to move forward and anything? So now is the time.

B

um Everything was good, but so we shall have to write our own github workflow github actions. What but, based on this on the the best template they were the one in okay yeah that sounds okay and.

B

Okay, but I I will keep coming back to this video and refer and in case I find any other difficulties. I will contact you on slack.

A

Yeah yeah I'll, uh encourage you to you know instead of dming. It would be good if you ask your questions publicly, because I think more people will have your questions, so it would be beneficial to everyone. So if I answer you know in a in a single producer, multi-consumer way so that everyone can benefit. So if you have any doubts, ask it in the message ci channel and uh you should create a pr and we can even have a discussion over that pr.

A

You know I can go over the files you changed and I can discuss what other changes need: etcetera, etcetera,.

A

And after we are done here, we would need the exact same thing. You know in linkedin and we now have a psyllium adapter so in time there as well and uh other adapters osm everything so every because you know stu kind of gets the special attention. So so we need every adapter. So if someone takes up one, so someone else can take up another, so you can actually. This actually encourages you to even write your own pattern.

A

Five, so I told you how, if you have any doubts regarding how to write a patent file, what's the template? Actually, there is it's if you go over the service mesh patterns, therefore, there are some pointers that have it's not complete. It's not complete, but there are some pointers about how to write a pattern. File like it depends on field and what the usage.md actually gives you a little idea. You know what references are.

A

So if you look at the pattern file and it looks kind of cryptic- you should go over and uh give it this thing a look and if you still don't understand just ping me on slack- and I will I'll update this usage.md to reflect what a patent file even better, because at the core of uh a lot of things that measure it does at the core of our workflows at the core of um you know the awesome stuff we do.

A

There is a pattern file everywhere, uh so we would need people to understand how it works. uh So, if you have any doubts regarding that, I would first suggest you to go over here: use dot, md! If you still don't get it ask it on slack.

C

Yeah there's a I've got a couple of things to maybe add. If I could um and let me I'm gonna, I'm gonna I'm gonna share. If I can figure out which screen to share.

C

This this is um this is great. It's great is that there's a lot to cover uh which you did and I'm gonna go back to the meeting minutes for a second and say um that say that um yeah now is the time to consider putting some of what you've learned into action.

C

The nice thing about learning, github actions- I guess no pun intended with putting it into action. It's actually a little bit confusing the nomenclature with respect to github, workflows versus github actions. We certainly use more well anyway. We have tons of workflows that use lots of actions, and actually this community has created like five different actions like about three of which are pretty public and um gen for general purpose use and some of which are they're public.

C

um I don't know that we really promoted them for general purpose use, even though they are so I'll. Give an example um out here. There's an organization called layer, 5 labs should be a few public-facing repos. I thought yeah, here's an action, um this particular action, while it's entitled um github action for meshary plugins is actually a general purpose.

C

It's written in a general purpose way and to just to give a a quick description when measuring server and mesh redox and mesh ui and all the and mastery ctl, like all of those components, are um within the main mesh remeshery repo. When those kick off. I'm sorry, when those when a release is made, um there's a one of the workflows.

C

um In here of the many that that might potentially fire one of them or more of them, will notify um remote providers, so anyone any other um system, any other provider that wants to be notified or any other system actually like slack like this, is a great kind of if you just want to play around and.

C

Add a slack notification here to let you know like into the announcements channel that there's a new measuring release like great. This is a good place to put that um it's kind of a an announcement that there's a new message: release that's gone out, and so one of the things that happens is it notifies um this remote provider so that um it can build extensions in accordance with the specific release, and so this is open for you know any any other system, any other person or organization to to come and subscribe for those notifications.

C

But more than subscribe like be invoked. So that's what really? What happens through that action? Is that um a remote build is then invoked and that's basically reinforcing what ashish was just covering is the fact that many of these workflows are interrelated and one will invoke the next across a repo across another and in essence, what she should was describing is a self-documenting um end-to-end testing system and then testing framework of which aditya and others have helped.

C

You know organize and structure that compatibility matrix so that all of these tests are, you know, updated all the time and summarize so that people can quickly come and understand what the current um you know how healthy from a quality perspective that the project is and and where it's at, with respect to the latest version of psyllium. Is it compatible or not um so good so anyway, anyway, there's other github actions out there, I kind of digressed for a minute on github actions.

C

So the point of today's call is to is to say yep it's time for you all to write a couple of workflows for the most part. It's not it's not even desired that you would write one from scratch. It's unnecessary!

C

It's desired that you're using the existing workflows, as templates for that you're, just kind of modifying the existing workflows, as they are specific to a given service mesh adapter. So the tests that are run today only test so far. They only test that mesh readapters do a bit of life cycle management of the service mesh for each of them, it's time to take the next step and have them each of the workflows to provision a sample application, um and you can to what ashish was saying.

C

Is you can describe that sample application in a service mesh pattern in that pattern? File format um so, anyway, uh is going to follow up our meeting today with a couple of like a couple of pointers and when he does actually hopefully he'll end up pointing everybody to mesh redox, because because we've got a lot of documentation about how it is that the build and release system works and the different workflows and et cetera, the tests that are going on for adapters.

C

There's some of that um described already um the fact that there's a central workflow that kicks off basically the same description of what he was just saying, but but he'll reinforce it with this pointer. So you can go, go back and look. Do ask your questions publicly like he was saying, because as a because that way, everybody else benefits, everybody else can engage as well. Other people can answer the question as well for as fast as she is. Sometimes he actually has to sleep.

C

So um I also encourage you to like ask the questions on the forum because they get even broader use and reuse. Moreover, there's a group of contributors who are working on um leaderboards for for, like a lucky monthly winner of who's, answering the most questions, so, okay. So, having said that, going back again to the objective is to automate the test plan and uh the specific example that ashish was showing is sample, applications end-to-end testing on for each of the adapters.

C

Actually, I think it stands right now that all of the adapters, or none of them deploy a sample app just yet not in not through a workflow. All of them are completely capable of that, um and so crack open this file or drop a note in the meeting minutes or whatever you want to do, but but slap your name down next to one of these and if it isn't obvious or something like a shish, can help.

C

You identify like what, where, where the scent, where you know like where, if you want to work on any specific service mesh to just um say so, there's going to be a fair bit of noise, probably around the psyllium adapter, which is the latest adapter.

C

That project in that community is quite interested in engaging in service mesh performance and they want their adapter to be. You know, well, curated, um I don't think it's even rep like like. I don't think it's even really represented in the tests today. The test results, so so I'll be quiet. Now I just wanted to make sure that that you all were encouraged to uh grab a piece of the the action.

C

Questions comments um like for someone who is, I guess a question for me. Ashish is for someone who is to say I'd like to do an work on an end-to-end test for deploying sample apps on linker d or just like concisely. What are what are those steps like? What should they augment an existing workflow.

A

Yeah so yeah, so if you want to, uh for example, if you want to have a sample app or link id, so linker d, you go to the documentation and you have you'll find um you'll find there.

A

They'll have their games already there, okay, so, for example, if I let me screenshare so okay, so in the pattern file that you see here, uh this virtual service and the things that you see here this is uh this the if it might look familiar to you, so this is just a copy paste of what uh they already provide us. So if you want to do the same with linkardi uh you first, you have to create a pattern file and the pattern file will uh look like this. So you have your you'll.

A

Have your service uh instead of type you'll, have to specify, for example, if it's uh actually.

C

I I started interrupt. Actually I think um my sense is people are going to get hung up on this. Is that, like, um I think we can expect.

A

You with the file, then okay,.

C

A

You this particular samples repo. This only has very this- has very few patent files right now. So let us first start off with this theo and then inside of the samples. uh I think we should. Maybe you know compartmentalize something like here. Only we see your stuff, we can put it all into a seo folder. So we know that those are istio specific pattern files then I'll also create a linker d folder and then I'll push a pattern. File specific to liquidity and all you'll have to do is just use that pattern file.

A

So you'll have your in your local system. You can first start your linkedin adapter. You can start your uh message server then deploy that pattern file and look at what are the ports that come up and then you know the drill is the same.

A

You first see locally what parts come up in what name spaces, and then you augment that into the existing template that we have.

C

And by the way um it shouldn't be much longer before, like today, measuring when you deploy measuring it comes with sample applications and those applications. They're not pattern file based, um you know so anyway, in concept you could just go, deploy the applications. You can go, deploy the applications. Today there are there's, as she gets each of these sample apps. Converted to this pattern, file format.

C

Those will come out of the box when measuring is deployed. So when you write the automation test, the integration test it at some point, it can be expected that that that there's a a set of sample applications already available inside of the measuring that's deployed in the workflow, so something to bear in mind.

B

Can you please hello? Could you please help me like when I look at the mercedes plan? There are quite a number of columns and some of them are a bit confusing. Could you please just highlight what briefly what each column means like not the rules, the column is.

A

C

Yeah there's uh in the chat and in the meeting minutes there's a quick link to the test plan.

C

And I'll stop sharing, so you can walk people through.

A

A

Yeah, so what was your okay? I think me has started this.

C

Yeah sorry, I was just I just kind of you can answer the question real, quick, which is hey as a quick tour to this test plan spreadsheet. That is, uh has a lot of columns and a lot of um is well. If you, if you try to have one, that's complete, it's like wow, there's so there's you know any number of features that mesh re has.

C

Okay time is it's this combinator combinatorial mathematics like this there's a set of permutations from this that, like it's like, okay, there's, 50 features, uh there's 10, adapters or 50 features times 10 adapters times two different clients like measuring ctl and measure ui, and then times the three different os's that we would support and then times the different architectures of them, maybe they're on arm or intel or and so like. It's it's gnarly and nasty, and then you take all the service meshes that it supports and they all have versions.

C

It's like whoa, there's a significant investment, that's needed in automation of these tests so to categorize them. There's just this. um A group of tests that focus on like installing measuring for the most part, installing meshri and the use of, in this case, the component that we're really focusing on trying to test and verify that it worked well, is mesh rectil and or I guess- and we combined the bash script in here as well.

C

um There's a bash script that we'll install in s3 ctl and then consequently use measuring ctl to install a mesh and, and so that, so it's kind of organized like this.

C

What's what's the what's the category of the test, the type of test, so that another category is like well actually managing the life cycle of a service mesh? Okay? What component are we trying to test? Where are we starting to test from? What are we from measuring ctl? In this case, okay? Well, measuring will deploy to docker or it will deploy to kubernetes as a set of containers? Okay. So what platform are you testing this against?

C

Okay, um in this case uh docker, um and then was this suspect against a specific mesh, and in this case I guess no, it was just more like getting.

C

um It was running meshery and docker kind of getting it installed and so you'd use this so the command under test, like the actual like function, that's being invoked was, is is this, and this is all things that can be put into and they are actually in some um workflows and then what would we expect that the behavior? You know what would happen hey. um You know what what are the comments? You know what actually happened um and then is.

C

Is this function covered by this? This is the thing that we want to get like green and it's kind of, I guess it's sort of in this column too. We might be able to consolidate these columns because if it's covered like what, where is it covered and then like you know, and then beyond that, it's like okay, what environments are we running, those in and yeah, and so for the most part in our workflows?

C

You know: there's a ubuntu system, that's spun up, and so for the most part, the automated tests tend to focus here. At least the end-to-end tests focus there.

C

Nice, thank you, hirsch, by the way.

C

So kranda um great great question: there's a lot in here.

C

um Anyway, so my recommendation to each of you and secretary, I think, is asking the right questions is to like ignore most of the noise. Just pick pick like a super small item like uh iterating, on top of the existing end-to-end workflow for the linker d service mesh or for a different service mesh and digesting um what it's.

C

What that current test is doing, which you know, which is what she should describe and then just taking a step forward with just augmenting it to then deploy a sample app as you do it to be kind of be vocal in slack or wherever about what you're doing and just kind of talk about it as you go, so that other people learn as well.

C

So it's like okay, I see that we're installing um liquor d. I see that well, our tests say that we expect each of the linker d components to be deployed and in these namespaces, okay great.

C

This is what we're these are our test, assertions and um good, and then we um update the docs when we're done like okay, so so kind of there's an augment there's another stanza or an augmentation, or you know, or maybe there's a maybe there's a generalizable workflow that focuses just on deploying the sample app, but that's something that you guys will figure out as you go.

A

Yeah thanks for the question hamsa, uh I hope that cleared any everyone else's doubt as well and uh yeah, so we already have the the reason I was pushing for hto is because we have the pattern file ready for linker d. It will not take longer I'll get the pattern file ready for that. That one is one, but it would be good to. You know start from this theo. We have the info pattern available there uh yeah, so any other question or anything that can you know help you even more.

C

We might want to um add a link to this training in the docs. Potentially I mean it will eventually, its shelf life will eventually expire, but probably valid for another year or so about how we're doing these.

A

Yeah so uh hamza, I'm counting on you for this one so I'll. If there is not an open issue already I'll create one, but I think that actually there is an open issue. So can I go ahead and assign that to you.

B

Yes, there is no problem, I'm just, I think, left with one issue with one of these decay and yeah. I can do them concurrently.

B

And when I scroll down the missouri test plan, I see some other things like service mesh or operations like service mesh life cycle management and in in the very first column.

A

Yeah yeah, actually uh those are basically, as lee said. So we measure does a lot of stuff. So there is a service mesh lifecycle management, so you can deploy a serviceman and then you can delete your deployment, so those or the first column is giving you a hint of what uh genre of basically what thing you're trying to accomplish so service mesh configuration management is basically whether we can import a file or not. So this much life cycle management is basically when we install uh something or not.

A

So uh if you come at the bottom, you'll find a service mesh lifecycle management. uh You know with the across different service measures, so life cycle management is basically what I just described in the form of you know: github workflows that, whether you you're you can manage your life cycle, uh whether you can deploy a patent file and you can see those spots come up. So that's basically your life cycle management. So that's your first column and your second column is basically uh what component is involved here.

A

So, for example, here the mesh adapters are involved in meshes or is involved in some cases. Mr ctl is involved in and then it follows, then, in in some of the in the next next uh column, you'll see that what that thing is basically what we are supposed to do, so uh the exact action that is needed to uh that we are that will perform there. So, for example, in the service mesh lifecycle management, you'll, see provisional service mesh using pattern file and that's exactly what we did uh actually something.

A

I need to add here: uh provision uh book info app using a patent file, so that would go in the in the basically I'll add another row for that particular thing, and then uh the columns uh that follow that are describing that particular thing. So, as lee said, these are just you know, permutation and combination. So even for your configuration management, let's say even for your life cycle management. There are a bunch of ways.

A

You can do it right, so you can do it directly from machine server so uh that those would be handled in your ui test. So whether you can click when you click at that thing, you expect something to happen. So does that something happen? If, yes, then, okay yeah, that test passes, you can also expect that we should be able to do the same thing using mystic cuddle. So for the same thing you would expect uh different components. To I mean I'm basically reiterating at least these are different. Permutation combinations.

B

A

Might look, you know a little uh daunting at first, just because we are, you know, things get multiplied because we have three clients and we have 10 operations and we have 10 adapters and multiply them and it's very easy to come to 244.

B

Okay, okay, like what um what's going on in my head, will I have it's? I am. Am I going to write like a workflow for every row that has linked in it or some of them.

B

I was asking he'll write a workflow for every role that has linked inside it like with workflows. You are showing us earlier on like one to test. This message sit for folding cutting the installation, then like that, like that, with all these permutations and combinations.

A

Yeah, let's see even if you pick an adapter like, if you pick liquidy, then also not each row corresponds to a github action. So, for example, a lot of these tests are just no okay. So, first of all just to clear a misconception: no, not all the 244 thing. These 244 rows do not correspond to 245 interactions. Only some of these are you know. Other ones are some ui uh tests that go on.

A

Some of them are meshing, specific things that whether a user is manually able to you know, do meshi cuddle something something and does that something happens? Do we have the expected behavior? So if you scroll all the way down to the bottom, there only you'll find the things related to the uh the the end-to-end thing that at the root 237.

A

If you go to the messy test plan, you'll find from row, 237 you'll find those things so and there we only have one row that corresponds to lingerie and that's a provision of a service measures in pattern. Five- and we are already doing it so I'll- add uh I'll- have to replicate this particular from 237 to 244 for I'll replicate this, for you know booking for different app sample apps for each of these adapters, so that you know you can have a sample apple ingredient uh yeah. So does that on service.

B

Okay, things started that's great.

A

uh We are uh past the we have three minutes after so. If anyone else has any final things to say any questions to us. This is the time or you know you can ask it on slack as well.

A

A

Okay, I'll take that silence as a no. So thanks for coming everyone. uh This is a bi-weekly meeting, so I'll in the same meeting I'll see you next next week uh until then, until that time, good night good evening good morning, depending upon whichever time zone, you are so hopeless, you all of you, I'm, although I only see mario, I'm expecting all of the other three people to be waving at me. So bye-bye.

C

We want to make sure, there's no robots in.

B

Have a day, how would you know.

A

Well for the web come on.

B

Yeah, it can be a free image.