Kong Kong Builders Livestream Recordings, 25 Jan 2023

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Mission (Im)Possible:Building Geo-Distributed Apps with Kong and YugabyteDB

Description

🤔 Have you ever wondered how a geo-distributed app such as a Slack-like corporate messenger is architected and functions?

👂 Hear directly from Denis Magda, Director of Developer Relations at Yugabyte, on 𝗵𝗼𝘄 𝘁𝗼 𝗯𝘂𝗶𝗹𝗱 𝗴𝗲𝗼-𝗱𝗶𝘀𝘁𝗿𝗶𝗯𝘂𝘁𝗲𝗱 𝗮𝗽𝗽𝘀 with Kong Inc. and Yugabyte.

🌎 How hundreds of microservices are deployed and communicate across distant geographies? How thousands of user messages and events flow in real-time across the countries and petabytes of data are stored and accessed across continents?

⚡ Wonder no more! In this talk, you’ll learn about the fundamental design principles for geo-distributed apps using a Slack-like corporate messenger. And you'll study the principles in practice by building the messenger from scratch with Kong and YugabyteDB.

Follow Kong Champion and Dev Guru Denis Magda
on Twitter: https://twitter.com/denismagda
on LinkedIn: https://www.linkedin.com/in/dmagda/

Event page: https://konghq.com/events/mission-impossible-building-geo-distributed-apps-with-kong-and-yugabytedb

Subscribe and catch up on previous Kong Builders streams: https://konghq.com/kong-builders

== timecodes

00:00 - stream starts
03:16 - welcome and mic check
04:43 - introducing guest - Denis Magda from Yugabyte
07:05 - why are geo-distributed applications required in 2023?
14:38 - introducing a stack of technologies
24:00 - start hacking - deploy apps to US Kubernetes cluster
28:04 - testing messenger app
29:10 - ❗️attachment service doesn’t work!!!
30:00 - Let’s fix it with Kong Konnect
37:00 - configuring attachment service and routes with deck
40:00 - ❗️ more problems? How can we fix that?
46:40 - drama intensifies, and our heroes successfully fix the problem!
52:41 - send messenger app traffic though Kong
56:27 - rinse and repeat - deployment of the app but in EU environment
1:01:44 - configure Kong Runtime for EU
1:16:19 - WINNING! Successfully deployed in EU
1:17:05 - Global Load Balancer
1:23:39 - Outro

A

A

A

A

Thank you, foreign.

B

Welcome welcome. Welcome. Welcome, welcome, welcome, let's do um Quick Check of the sound. It is Wednesday. You are in Internet, somewhere you're, watching this on the Clone YouTube channel on LinkedIn or Twitter, or maybe even twitch.

B

uh My name is, as always, Victor gamof and you are welcome here at Kong, Builders call Builders, it's our regular alive show where I'm, showing you how you can use Kong Tools to tame and Rule your API jungle, how you, like my pun, let's um let's see if we have any uh in any viewers here so write down in the comments. If you see us, if you can hear us, where are you coming from what platform you coming from?

B

um Is it? um uh Is it the LinkedIn? Is it Twitter or maybe where what's your geography? Let me see if I can see um the chat in the comments um and reactions so far so far, not so many people, meaning that we might be um not many people are watching us, or maybe many people watching that too shy.

B

So when you join us, whenever you are write down in the comments where you're coming from uh meanwhile show must go on, and today it's it's great that I'm not alone today, because I today, I have my friends Dennis and the Dennis will be here. Helping me to run this stream today, Dennis it's great to have you at the Kong builders.

C

Hi Victor uh Hello friends nice to be here: let's build some stuff cool.

B

So um if you, if you follow us and you follow this channel for a while- you probably already uh seen Dennis multiple times- uh Dennis is Kong. Champion he's our um one of the Elites who are helping us to promote some of the Kong technology.

B

um We wait for a second while his internet will recover, and uh we let's see if we have a comments here so Roberto from Guatemala. Great welcome. Welcome to heavy here in uh based on the uh the icon, I see YouTube is working. Some people are actually watching us on YouTube um and uh Dennis. uh Tell a little bit about just for those of you who join us just today and don't know who Dennis is and what you do and uh how come you become a Kong champion.

C

uh So yeah a few words I'm I've been working with distributed databases and systems for the last eight years, so my zone of Interest are applications that use distributed systems, especially databases and what happened almost a half. A year ago, I came across call for two reasons, because first 10 or week and second of all, that looked like a promising technology, I will build in some application and I was looking for a Gateway, an API layer and eventually I came up with a architectural uh like not the product behind but pattern.

C

How would you deploy geo-distributed applications across continents across countries, and how would you deploy this respawn with your database with your backend, how everything can be coordinated and used uh easily so and that after that experience, we came up with an interesting topic and that's the topic of our Builder Workshop today, yeah.

B

C

Want to build a you want to build some stuff yeah.

B

Let's do build some stuff, like so the first of all Let's. uh Let's set the stage about the this, like a geography thing right, so let's talk about. Why why this is important and like does it? Does anyone need this uh in in this world? um So what is uh geo-distributed technology lens.

C

So I think that even my mom needs a geo-distributed technology because she's not a developer she's, not an engineer, but she uses uh different Payment. Systems device should buy stuff from.

D

A grocery store.

C

And the stuff that is buyed from a grocery store is uh needs to be paid for and that payment systems are usually redundant they're just distributed you never. If you use, let's say your your best like Bank, they can have an application that payment application, and usually you have multiple instances of the same application running across different data centers, so across different Cloud regions and also multiple copies of your transactions are stored. That's done for several reasons and that's an example of a geo-distributed app, but let's define it right if.

B

You Google for.

C

A distributed application you would come across two sources: one of them would be from you go by the company, I work for a distributed database and another one would be from Microsoft. Microsoft has a quite extensive uh tutorial that explains how to use Microsoft Azure to build geo-distributed applications and their definition is written on this slide. A geode distributed app is an application that spans multiple Geographic locations, and why do they do that three reasons: High, availability, performance, compliance, three plus words right yeah,.

B

Yeah, let's, let's talk about high availability, so you touched a little bit about high availability right, so your your system, when you want to go pay for the things, um needs to be highly available. So regardless, where you are, you want to make sure that you have access to your money and this this.

C

B

You can get you know, access to your system now in terms of performance, how this Geo thing helps with performance.

C

Let's say that you, the same payment system and the customers, let's take JPMorgan JP Morgan is International Bank and it has especially customers Chase or JPMorgan in the United States and in Europe at least in those two Geographic applications. So you can use Chase credit cards to buy different stuff from Amazon or from your best store, and let's say, if you buy from the United States or whether you buy from London or from there, the latency, the performance for your purchases for your transactions uh have to be the same.

C

It has to be comparable, and usually this happens, if, let's say all of the data centers, if all of the application instances and database nodes for this payment system of Chase or JPMorgan, they are located just in the United States. The performance would be terrible for those who live in Europe or who would travel to Europe, because their requests from their phones from their credit cards would need to travel to the data centers to the application instances across the Atlantic to the United States.

C

But if Chase JP Morgan deploys multiple application instances, one instance several instances in the United States, the other instances are in Europe and the same applies to the data like databases. Then the performance will be comparable, whether you live in London or in New York City doesn't matter you pay using your credit card. The latency will be comparable because.

B

So yeah by performance we mean we mean, like reduce latency in increased throughput, so we have a access to to to local and uh compliance. You know usually things that overlooked and we don't talk much but believe me or not in the recent years. The question about data locality is become more important like how data is stored where data stored. If you come for this, if you you're serving your customers who are in Europe, you need to apply certain compliance rules for for this data, and you need to make sure that is data.

B

If you performing data or your processing date of Europe in Union customers.

C

Stay there yeah exactly you're, not about the word. Jdp are your general data protection regulation? Yeah, uh that's uh I! Guess that's, probably one of the most boring things for us for application developers, but if you want to uh become a successful Financial as an engineer's right, we want to get stocks. We want our company to become successful so that we can get more money. Then you have to think about compliance. Yeah, like what you are saying.

C

Jdpr is like a usually one of the first examples we uh through in another example that I discovered a couple of months ago, uh India a Reserve Bank of India. They uh have one regulation, so if you buy anything from India or pay for anything from India, then a transaction information about the transaction has to be stored in India, like in the data centers located in that country, and you have to say several hours to transfer this information.

C

Let's say from your original location where the payment might have been processed yep so, for instance, for our financial application. Let's say that the processes payments you go to Amazon, you are Indian resident, you buy something from Amazon or some other store that is located in the United States, the compute. The transaction can be completed in the United States, but then the information about that transaction has to be stored in India within the next like 24 or 80, 24 or 88 hours. I. Don't.

B

Remember yeah so in this case uh data data replication, uh you know the products with start start getting more money right then the distributed databases like uh yougabyte also uh will get a lot of money for for this.

B

um So um we're not going to build the financial system today, right, uh we're gonna, build something more fun and uh much more cooler, I think and uh exactly hopefully, hopefully will not break um many things today. So yeah, then it's uh walk us through what kind of tools we're going to use and what gonna be doing today.

C

uh All right all right, but before that guys also to answer your question really quick. uh Is this a workshop for you, if you're, not planning in the meantime, to deploy an application that spans across several continents? Yes, the answer is still yes, because if your application is deployed just within your single Cloud region, but across multiple availability zones, that's also geo-distributed application.

D

C

Your application might be originally designed and envisioned just for the United States Market or just for the Brazilian Market or a market of your country, and then that instance, instances of your application can span across multiple regions within close proximity. For me, like I, live in the United States and for me the close proximity would be in United, States, West, Central and East Coast.

C

It's like it's fine to deploy a geo-distributed application across those locations, so that show is still for you, even if you deploy just Within A Single Girl original across multiple availability zones or within boundaries over your country.

B

So, essentially, we're going to be um also we geographically distributed because we're going to be collaborate on some of the bits um and uh we will be deploying an API layer. We will deploying the applications layer um and the database layer is also distributed right, so we're going to be using uh Cloud Technologies for all these things and uh yeah. Let us know if it's something exciting can.

C

B

We want to I want to check. A connectivity like I was told that um uh we also live on LinkedIn, but I. Don't see any comments coming through guys if you, if you're watching us in LinkedIn, just like, say hi, saying, hey I'm, uh I'm doing you know you guys doing great you guys suck just let us know um uh we need to.

C

See I have some affirmations in feedback yeah, just just let Dennis go right here and stay alone.

B

No, no I I actually enjoyed to have a um streams where I have a um the the partner crime, because in this case I would just need to be. You know, asking stupid questions in the reading. um Reading questions from from the audience which is uh okay, so Emily uh is still here. Everyone is here so connection to uh LinkedIn is established.

B

Okay, Let's. uh Let's.

C

Let's start right: let's start an architecture of Geo distributed application, so building blocks or components. Folks are exactly the same as you would have for your standard application. How do I Define the standard application? That's an application that runs in a single availability Zone and it's in a single data center. Remember.

B

Remember that your times when you were in in Java world- and we call it like a three-tier application, we will follow um Java, Java 2 ee, um a blueprint where you have a front end or your um kind of like a front-end part. You have a business logic. Obviously you know we're doing business here. It's not the it's not like. We, we slacking or whatnot, uh speaking of which we're going to be building slack, clone so kind of slack in here, and there will be data access layer.

B

There will be database layer so we're following this uh blueprint steps from a long time ago, right.

C

Exactly exactly so, and generally like those steps remained, the same. Architectural suggestions remain bad with the geode distributed app all of those components like backend data layer, API layer they have to be distributed. They.

D

Have to function.

C

Across multiple availability zones or regions, okay, yep, and also to give you an example like why even a database has to be distributed. Let's get back that example with our JPMorgan Chase payment system, all right, they're, smart enough. They deployed application instances in London and New York.

C

But let's say: if they have data only in the United States, then it doesn't really matter that an application instance is on the is also in London, because that application instance in London has to send requests under the ocean through the Atlantic Ocean to the United States, just to get the data and that's the wall. A proper architecture. Fast architecture should be designed in such a way that customers who are in London, who travel to London, they can get data from London.

C

They don't need to go to the United States and that's their Challenge and requirement for geo-distributed application.

B

And the thing is with this: uh like with the API layer, you have to hear that, like a synchronous communication, you want to have a response immediately from your API server, like when you're getting data or you're trying to get something out of it, and you want to get this like immediate response.

B

um Where is data replication or how the data would actually travel? um And then it also came from this? We came from the same world of distributed caches as well like we we a lot in in the past.

B

We use this pattern called uh write through cache, um so in this case the the the the the the it writes, Into Cash, faster because cash, usually in memory or stored somewhere here that faster than disks and after that, the cache system will be responsible to kind of flush it into appropriate long-term storage, because the cash um you know architecturally is think that not uh not gonna.

C

uh That's right, so, let's uh Let's see we are not going to talk about the database today, just for the reference guys uh that deserves another conversation. There are multiple deployment patterns: how to arrange your database across different regions. It depends on how far your regions are from each other, but today we are going to talk about corn right. Just where is my t-shirt? Yay.

B

Yeah so um Dennis also known as a as a Kong champion, and the Kong Champion also receive some some cool swag. Meanwhile, uh we start getting uh more more viewers. uh Portugal is here, Marcos welcome. Welcome to uh Kong Builders great.

C

B

C

One one so what we'll design our blueprint for today we are going to use a true Cloud environment. We are going to use Google Cloud, but don't be discouraged all of these architectural patterns and blah blah blah. They work the same for AWS Azure, IBM, Cloud, Oracle Cloud. Whatever is your preferred vendor? We will deploy two kubernetes clusters. One kubernetes cluster will be running in the United States East somewhere in Virginia, and the second cluster will be in your West Somewhere in Germany yeah.

B

Most likely in Frankfurt somewhere.

C

Probably you should know better.

B

No, like I, know that Europe also they deploy in Dublin somewhere right and.

C

Danny, like they have uh so generally, they haven't Google. At least Heather has a region in Amsterdam in somewhere in Netherlands. They have also in Germany. They have in some other places. It's it's. It's really.

B

It's a geographically distributed cloud system, so yeah so.

C

Like was it a cloud cloud vendors they already come with abundant resources. You can literally deploy your application in any, not at it in any, but pretty much everywhere. Cloud regions are everywhere fast, Network interconnects, all of those Cloud regions, but we would take the simplest example. We would have two different regions, your request and United States East to kubernetes clusters in each region and every kubernetes cluster will be running our application. So.

D

C

Victor decided to reinvent the wheel. We are going to build the select killer that messenger. That manual was used daily to.

B

Caliber or maybe not the killer, let's not be that it's like a better we want to when we want to develop another like a corporate messenger. um That also would be Geo distributed and uh also would be nice and cool. So that's.

C

What you want to be doing.

B

C

But if to just to give you an idea like why? Why slack in fact slack that the real slack is an example over geo-distributed application, you can go to the slack engineering block and find articles on how their application is designed. It operates across a different uh Cloud regions, different continents, and that also was one of the reasons like how, let's let's try to build it using these architectural patterns. So our applications like like uh messenger, it will have two micro Services. One microservice is the messenger micro service.

C

It's basically the one that allows us to exchange uh text messages and the second one is attachments. So if you let's say decide to send the picture uh from your Resort, if you're on vacation or from your office, and then you can do this and that would be the responsibility of the attachments, microservice cool microservices. Actually they are not tightly coupled they communicate to each other. Using cons or corn is our Prime.

C

We are going to use cone Gateway can't connect today, but you certainly can improve this architecture with other uh services and libraries of the.

B

Course, yeah we're not gonna We're Not Gonna focus on some specifics of the of the Quan would rather show you bigger picture like uh many times on the stream we're going into the specifics of certain features, and of course we can do this. It's called Builders, um but I think uh one of the reasons why I'm excited about this particular use case is that uh we we can show you um bigger picture so I think we were already like 20 minutes in yeah.

C

Yeah, let's do this, then yoga by DB keeps our application data messages, profiles, workspaces channels, Google Cloud Storage is for attachments it's an equivalent of Amazon S3 and multi-class your Ingress, that's our Global Cloud load balancer! That's the entry point for all of the users like the load. Balancer will decide what kubernetes cluster. What application instance we'll be handling the user request based on the user location, all right, yeah.

B

C

Let's get get started with.

B

C

Get let's get to work yeah, so let me show you what you've done. We already have uh our clusters running kubernetes clusters. For that, let me quickly uh execute this command. They have the notion of the fleet.

C

Memberships uh list here is I. Have one cluster is in the east coast of the United States another one is in Europe West the Clusters are running. Do you have any application instances if you should not if I removed everything, uh no, that the kubernetes cluster is uh uh empty and just to show you what could build latest cluster? My tool is connected to right now it should be connected to the cluster in their Easter.

B

C

States that so generally, what we'll do first, we will deploy the whole solution in the United States, and then we will do the same in Europe and then we will show you how to use the load balancer just who can coordinate between different instances, yeah so deploying yeah.

B

So, let's uh first of all we're going to deploy this like a messenger, app um and uh yeah you can, you can start deployment, but meanwhile we can take a look on the.

C

B

C

A special script yep and that's.

B

So, while the script will be running, we can we can we can allocate to uh into deployment. So let's take a look on the Manifest for, for those um and.

C

The scripture start: what the script does it's a series of different kubernetes commands? We are switching to one of the kubernetes clusters right now we are deploying in the east coast of the United States. Then we are deploying attachments. We deploying attachments, microservice basic kubernetes commands. Then we're deploying the messenger micro service. Again, we need to have some Secrets blah blah blah deployment service, nothing that requires our attention right.

B

Nothing really special right, yeah.

C

Yeah, so let's see if the deployments are ready, yeah looks like yes, so let me open the locks, that's a Java application. uh Folks, but again it's up to you what language to use doesn't really matter.

C

We are Java developers.

B

Yeah we came from the very long um um lineage of Experts of Enterprise applications and job is our our drug, um so essentially Let's uh we're gonna do this in a in a steps, and uh you know one of the things that you folks can do. You also can um you know follow along, so Dennis will send me this.

C

Yeah, so let's do it this way, so folks, uh what this screen shows is that I, don't see any exceptions. uh My application instance, the messenger microservice managed to connect to my Geo, distributed yoga by DB deployment right now, I I'm, using a multi-region cluster in the United States Central Zone, East and West I, also have a different question that spans across several countries, but again another topic for today. Let's make it simple, so the connection is established and we are ready, can.

B

You uh can you send me IP address.

C

Yep I'm sending you the IP address and folks, if you also can try to connect, but week, will be our user. I will be showing you Deluxe and Nick. Let's try to send the first message: yeah.

B

So I'm going into into the chat, and so that's how this UI looked like yeah.

D

B

Hello from Kong builders.

D

B

The message went out and we should see this message appear in the log that.

C

Would be weird, as you can see, database yeah the low level messages that my backend in the east coast of the United States hibernate, is a java oil RAM and always find the changes have been written. The database keeps all of our data and distributes it across several regions. Next, we try to send the picture. Send me.

B

Your okay, let's photo, let's, let's try to find something: uh we'll just do upload file and we're going.

C

Into here we need to see your screen yeah.

B

Of course, we're going into my screen upload file, something like this and it says upload failed.

B

Was it the? Was it intentional and we see there's nothing, nothing really here: Happening Here, Yeah.

C

So in my back end, my messenger microservice reported that issue. It says unresolved address exceptions, so the messenger expects to communicate to the attachments microservice through corn, but the con Gateway is not running yet and that's why the exception yeah, you can easily fix it. So it's time for corn yeah.

B

So I think there's a bunch of bunch of patterns that uh we can show in the future. For example, uh right now we're going to be using Chrome gateway to route or all our apis.

B

um But in future we can talk um on I, have a I have a talk about the.

A

B

Know how to use the service mesh, for example, now, so what we're going to be doing now with this is to.

B

Yeah wait a quick second, while I um show you this, so um we have uh we're gonna deploy kwank in kubernetes, but uh we're gonna be performing all the uh management Jazz from this console. It's called con connect and uh we will have our instance of our application. uh We'll be you know running here inside.

B

um We will see the traffic we'll see everything that goes into this. um So in order to do this, um Dennis will be executing set of commands.

B

Essentially, um Quant will be running in this so-called hybrid mode, where uh con only will be um the counk that runs inside the Danish cluster only will be serving traffic but commands all this configuration and services config They will receive from uh from colonic connect console so I'm switching back to Dennis, and so one of the things that he needs to do is to just like update him charts on his uh his local environment and in.

D

Order to add this.

B

Control plane data plane in in order to disconnect will be communicated to konk. We need to create the pair of keys that will be stored as the secrets inside kubernetes cluster.

B

um So those things are, uh it's just again the part of mechanics there's nothing super exciting here.

C

um So, okay, no no hold on. That's that's my issue. uh I need to go to uh this directory, and now from here we are installing did.

B

You create the secrets.

C

It was created yesterday.

B

C

Yeah I have the secrets, so all right looks like the deployment went. Well, let's get sports right. We want to see the ports in their namespace cone.

C

Okay, let's see the looks cute blocks still now. Let's, let's try to refresh it's initial lighting, Cube CTL logs from this spot and they're in corn. What is the initializing? We need to wait. Do we see anything here in Euro and.

B

um Not yet so one of the things that it says out of sync, meaning that it's not up and running yet.

C

Yeah but there yeah, the corn proxy, uh the service that is going to be used as their for external communication is already up, and all we are waiting for is our port.

C

Is waiting to start potentializing?

C

B

Can you open K9s, so we can have like a visual view to what is going on there.

B

And you can go to con namespace.

B

Yeah, so the the beauty of the coin Builders is that you know something will definitely won't work and we will see how we can fix that. So that's the that's the cool part, so yeah namespaces.

B

uh Just go numb spaces enough that select cone here, so let's go to uh describe uh you can do D and see why why it is stuck.

D

B

And go all the way down to see if there's a uh on the way like there's, usually status, all the way.

B

That says like what is going on there, what it waiting for to create new cluster or you just like reuse, the the previous one.

C

um Now we use the same cluster scheduled.

B

Yeah yeah, it looks like it looks like you are, you secrets are not found, so that's what I said.

C

Like just create.

B

The secrets again.

C

But I could ah sure I have the files locally on my laptop, but they are not loaded to the yep.

B

C

Distribute me all right: yes,.

B

So the go back to this, uh the command with the creating secrets.

C

Yeah, let's, uh let's drop, let's drop no.

B

You don't have to drop, the kubernetes will uh will eventually um pick.

C

B

Up yeah, it eventually will pick up. So we know this like a crash, Loop uh back off um and the secrets was created, so uh the board will you can you can actually balance this deployment? um I.

C

B

You should just do a good control and delete delete deployment, and, after that it deployment will be. You know, recovered.

C

Yeah, let me do this. I would.

B

Or or like this, but it's it's like very nuclear.

C

Okay, it's dead; okay,.

B

C

In case, let's repeat, once again, the secrets is created and.

B

C

Yep we need, all we need to do is to deploy the Gateway.

B

C

C

One is terminating and another one is starting.

C

Let's check the locks.

C

C

D

C

At least we see all.

B

Right so live here, yeah, and uh what I see here right now in my uh in my instance, I see that it's just connected- and uh this instance is uh uh is not is- is connected and Let me refresh the screen one more time, so why it is important, uh it will synchronize the configuration it says like if I see what kind of version is up and running is here, so all good, that's um proxy. Just now and I we can do multiple cool things here.

B

So um so first thing that we need to fix this um attachment problem right. So the with the attachment. What we we. What we want is that the the service that would be available uh this service needs to go uh through the quantum.

C

You need to, we need to have roads right. We need to to create a road from that, so that the messenger microservice can go to the attachments microservice using the cone, and we need to.

B

Have you know correct so and in this case, what we're using here uh we're using this um configuration, so we we can go and manually, create the service I'll go to my service cup uh like overview service, Hub and I create service, but we like to be like more productive, because we're going to repeat this multiple times, we're going to repeat this in Europe and for this we're going to be doing using this deck tool.

B

Dac is the con declarative, config system, and what that can do actually is um a deck can operate with uh connect. So if I'll use this this ping um and I, it will show me that it's connected to connect so the way how it works once again um from my computer I will be talking to connect and since connect already have a connection to quonk, um this service will be created and connect, and this configuration will be shipped into the Danish con.

B

So here's your um deck sync so um I also specify a runtime group, so I can distinguish from one Management console I can distinguish multiple instances of conk and they are running in the different places and the next step. I'll show you how we can do the same steps in in Europe connect it to Europe based control plane, so it will be closer to Dennis um con Gateway instance that will be deployed in Europe so and right now.

B

There's no services and I want to go here and I will just do this, so it will take this file configuration file that has the two configs one is that we need to have attachment service. That's the location inside my kubernetes cluster, so con can find the service and we created two routes upload in ping.

B

So uh after I run, this I should be able to go my messenger and just do uh remove this file. Let's do it one more time, upload and saying here you go file from Kong builders so and when I click here send we will see that file is uploading.

C

No, no try try to remove that one. You, the file.

D

C

Select a new one, just.

B

C

B

C

Fail so it's still it's, it still says unresolved address exception, so.

B

Let's see if, let's see if our configuration came in came in through so if I'm going here so, let's see inside my service, Hub I have attachment service and inside my attachment service I should have. um Thank you.

B

Service is not published, it's the it's okay, so we have our upload route and we have a ping route. So, let's see, if my um this instance will have a Gateway service. Yes, attachment service is published.

C

It's published okay, so let me double check on my site. Yeah.

B

C

This you deployed it in the okay. So let's do this Pub CTL! First of all, config current context. We are in there United States good. What's about Services get services, my services, uh messenger attachments- they are running so far so good and for the corn.

C

Keeps your tailgate Services namespace cone, no resources found in Clone namespace.

B

C

Where am I deployed.

B

uh Deployed inside the Conch, let's see.

C

um Hold on hold on hold on something is strange. It should have been deployed in the United States East right.

B

Yeah, just uh can you open the um um uh K9s I'm operating on K9 is much better much faster, so um a shift shift semicolon and you can go to service write the service.

B

So this is the services we deployed here. So they'll go to namespace conk you can do uh you, can click the number one and inside the Quan and you can do Services here.

B

um That's that's. Okay! That's.

A

That's totally expected.

C

Yeah? Why? Because.

B

uh Because we don't, we don't expose the Chrome to outside world. Yet.

D

B

um So because uh let me try this one more time, maybe it's just a um upload. Let's do this one more time, upload failed. What what the log says.

C

So the log says the same and resolved address exception.

B

Let's, let's scroll up which, which, uh which exception.

C

No, what's no no, what should happen? I'll show you uh so the messenger it connects to Corn using this AP address: yeah, that's within the kubernetes cluster, so my local kubernetes cluster in the United States East and yes, the name of the service, and there is no service. For some reason we have the port, but we don't have this service.

B

uh Did you kill the Quan deployment and everything in it Let's? uh Let's take a look on on the values um on the values yaml, one more time on the yeah on the values yaml.

C

So this is what I used right.

B

Yeah so um database off uh it uses uh yeah connected to this one. It's all good certificates are good and all this and uh scroll up um we disabled admin. Api. We don't need this um yeah, so it looks like the configurations actually shipped to conch, but Kong service is not um is not available. Did you create the account service like manually before um in order to this service will be available for for.

C

Let's, let's try to reproduce okay, so.

B

C

Do this um I was deploying here so we executed just this command and before that we killed the corn in the common namespace. That's all we did.

B

Yeah, um let's kill; uh let's: let's kill this one more time.

D

B

um You can even do kind of like a purge thingy.

C

Like this, okay.

C

Should be deleted.

C

We can we can even let's delete the namespace. You want to delete the namespace.

B

uh Namespace yeah yeah. Why not, but do not forget to create this Secrets afterwards.

B

Now this is exciting, like something is not working. It's it's always exciting. Here, it's much more much more exciting.

C

C

Okay, deleting everything done so what's next, let's do this.

B

Creating the space.

C

And I have the rep just in case. Let me do what we did. Con already exists with the same configuration: okay, it exists and.

B

Then, which is okay, that's.

C

The second command this one yeah and just okay,.

D

C

The repo, after that we are creating the secrets created and final one. We are installing everything from there chart and we have the values file. Let's do it.

C

So, let's check here uh my code, namespace deployed uh right. We have the namespaces ETC, get.

B

The uh yeah get this uh everything what the Kong says in terms of like services get.

A

Services inside the content.

C

B

C

D

C

Okay, okay, sometimes sometimes you folks need to start from scratch. It's.

B

A it's a it's called eventual consistency. Let me try to upload the file one more time uh distributed.

D

B

uh Here we go kubernetes; okay, okay,.

C

And also yeah yeah go back to me. I'll show you.

B

C

The message succeeded and I have a reference to this file. Yeah.

B

You see I think I.

C

Think to Google storage.

B

C

Let me get it from there: a.

B

Little bit drama a little bit drama.

C

B

Emily excited to to see this yeah, that's what uh that that's, what they call that's, what we're doing on Builders right! Okay! So, while we're still here uh while we're still here what things I, don't like and I, don't like to see, um can you do um uh cook control, get Services uh dash, dash all name spaces.

C

B

Yeah, so there's a messaging service is deployed as a load. Balancer, that's not! No! No, oh okay! So um that's not not cool! So let's do this uh something and something different and we will um route this messaging service through the clock as well. So in this case you don't have to expose this through this IP address and we're going through a quantity address which is, let me pick up. um Can you do cool control, get Services, there's, there's all um or all name spaces.

B

uh All namespaces or Dash a I, don't remember, Dash a.

A

Try to do yeah.

B

No! No! No, not like this, like dash dash all Dash namespace, something like this yeah.

C

B

C

B

C

Okay, I can do it this way. Man, it's yeah,.

B

Like Yay, zero press, zero, zero, okay: here we go so, as you can see here, we have a Quant uh service deployed as a load balancer, and we also have a IP address for this um I, don't see how you can see the full version. Can you click here.

C

B

It's internal it's inside the cluster, but we're looking for the 35.99. uh That's why you know you your your window is too too small.

A

B

We go here we go so uh if we do this, let's just copy this IP address. Let me also copy this IP address.

C

Yeah no roads, okay, yeah.

B

Perfect, that's that's the clock response that this is where we need to um configure Quant now um so.

C

All right, so what I.

B

Want you to change this so okay, so let me show you cool uh cool thing that you can do in um in the uh in the K9s, so switch back to K9s.

B

Okay, you go into your uh messaging service and type e like edit, okay, so now scroll down and uh instead of type load, balancer change, the type cluster uh cluster, IP scroll down, scroll down, type load, balancer change it to Cluster IP folks. If something cool breaks, I take full responsibility for this.

B

It's apparently it's the.

C

It's improvisation, impromptu cluster.

B

Ip yeah IP, all capital yeah and the do space between yeah and I. Think I, don't know what stages there's. This is the veeam um we will see. Oh, oh, oh.

B

C

B

Of editor is this like: what's the.

C

System, UA type Q, a Escape.

B

Changes Escape semicolon Q, it's veeam; okay, yes, okay, q, a how to save uh QA.

C

I, don't know how to save it. So, let's, let's do it my way? Well, let's do it my way: okay, The, Brute, Force way. What I will do? That's fine.

B

uh It was uh like w I, don't I, don't know, because I always have this right. Wq WQ, not not q. A WQ do.

C

You want to try one again once.

B

Again, yeah: let's do it one more time, foreign.

C

Messenger uh Ops no wrong direction, I know edit, and then we go to the many.

B

Times, write down in the corner in the comments how many times you stuck in beam space, put the space because I don't know how the this yaml? Okay, no, no, no Escape, W, Escape, uh semicolon WQ!.

B

Here we go okay, now cluster IP, so I should lost the connection to this.

B

um To this service, but now inside my.

B

I can go and create another service. I can call it.

D

B

B

And description go on builders.

B

And listen: Messenger Service.

B

Now the we need to add service uh service version, name V1.

B

um This is where we're selecting the runtime group, where this service will be deployed, because we have only one runtime group: we need to add implementation service name. So um what was the Dennis? What was the URL? It's kind of like a messenger service right.

C

um Yeah, it's a different one yeah. Let me.

B

It's gonna be mess and sure.

C

It's yeah, it's you need to replace, swept a messenger.

C

Messenger, no, it's a little bit different. Let me yeah messenger check the spelling.

B

Can you can you send me, what's the correct spelling messenger with e Messenger right yep.

C

Messenger service I'm, looking at their K9s, okay.

B

So I'm going there and I create this service. So now uh Dennis you have an IP address for the calling that you open this, and it shows that there are no routes and stuff like that.

D

B

D

B

So wow, oh of course that's the that's the that's the beauty of uh connect. So let me show you well: I'm uh Dennis will be clicking on putting some data in the background. I'll show you one of the things that we can get immediately here. So, for example, for the uh for, for for our attachment service, we can immediately see the how this API is used.

B

For example, we use this once because I upload this file once, however, if we go in through this messaging service, we start seeing that this API was used 41 time because Dennis is clicking around each request will be generated. Some of the um you know. If the Dennis will be clicking inside this uh UI, let's do less less Tower, so we can have a more exciting graph for the last 50 minutes. So um can you send me, did you you send me this IP address.

C

The IP address here in uh Telegram.

B

Yeah and I'll just do a test at gmail .com.

B

D

B

C

That's password yep.

B

Okay, so I still hears nice dope.

D

B

So messaging works now next stuff is next, stop is to deploy uh in Europe, um and you know we can. We can post this somewhere like where's the IP address guys you can grab this IP address. um You can grab this IP address and play around with chat, test test and generate some traffic.

D

B

Exciting 35 99 40, 37. I'm, switching back to Dennis and.

C

Now, let's, let's do this in Europe.

B

Yeah, let's do this in Europe and I will create. While you will be deploying your your part of application in Europe, I will create a runtime groups in um in Europe as well.

C

So folks, what you can see uh the same script that starts my microservices and this time the parameters have changed. I'm, instructing the script to deploy the same application instance, the same micro service, messenger, blah blah blah, but in Europe West one uh looks like the flight is normal, also uh I. As long as we were deploying in Europe, my Cube control tool. Right now is looking into that kubernetes cluster. Let me double check that current context, yep I'm looking to Europe and let me check the services. uh Okay, the services are running and deployments.

C

uh Deployments: okay, the messenger instance is studied. Let me check her. The logs yeah I can check the logs here and.

B

Remember you don't you know you can create your service for now as a service, not uh cluster IP. You don't have to create this as a load balancer now, because we're gonna.

C

Drive it anyway, like I, need to change it once again. I can change it uh by default, because in my scripts it started as a load balancer.

B

Yeah, that's fine.

C

You want to change it, I can change it at least I I can practice with it or yeah.

B

I sent I sent you um I, sent you two certificates against I, sent you um certificate and key for Europe instance.

C

The first one is certificate right, yeah.

B

First, one is certificate, just create something like uh folder Europe and create the same file, so you can use the same command.

D

C

So give me a second: uh where is my deployment? It's in here.

B

Yeah, meanwhile, so we will not yeah.

C

I would create USA I.

B

Would yeah I would not spot the the keys, so that's why I walked through this uh config on my site, so we didn't show. We didn't show this before, but uh I'm, creating this runtime group. That will consist one runtime that will be running inside the column uh inside of Dennis kubernetes cluster in Europe and, uh as you can see here since it's Europe we're going to be using Quantum deployment in Europe.

B

uh So the connection between uh my control plane without management configuration and then is runtime, uh it would be a local, so the communication would be much much faster um and after that, it's it's a value files from um from Helm charts. You've seen this multiple times on this uh live stream. I hope! um If you, if you follow this for a while.

B

um Meanwhile Dennis we have uh Gabriel from Argentina, is testing our system and it says it works. So that's awesome. Cool.

C

Cool actually Gabriel keep doing that. Once we deploy an instance in Europe. ah No, it will Argentina yeah. The United States is still covered.

B

You yeah, um and uh so what we did is uh generated. The this is like a the certificate um certificate key and those certificates will be deployed again as a secret and, as you can see, the the Quan uh connect is a walk you through the process. What you.

C

Know Corporation I think I'm ready.

B

You can deploy, you can deploy your runtime group yep yep, so.

C

But make sure your right.

B

C

C

Europe, wonderful, so creating the namespace and their.

C

Creating secrets.

C

B

You can see see. Latency is higher because you're hitting this cluster from your um from your place uh on the east coast to um yeah. So let's see.

B

So, depending for external IP address yep.

C

C

On this, one I hope I did not mess up, because we are literally running out of time. Oh good, lucky me so uh man, what do we need? We need a.

B

uh Nothing so essentially, we need to have an IP address, but this IP address for conk. Only so now you need to go and change your service, so you can have a load balancer and meanwhile, I should show you some some of the cool stuff on the Conch side of things. So again go change. Their service and I will show you so now now I'm in Europe and my console is in Europe, as you can see here on top over here, it's Europe uh and one of the things that we just just now.

B

That then, is uh connected to Dennis instance. So now I will be able to control is right now, as you can see here on my there's. Nothing here, there's nothing here on my service, so what I can do? I can go back to my North America, my North America runtime group, and just simply a dump configuration what I have here. So what I can do deck dump, uh there's a command that does this and it will bring all the configuration. Oh I'm, sorry, um runtime group.

C

B

C

A second connect.

B

I just do dump.

B

Okay, okay, terminating so did I change it in the wrong place.

B

uh Okay, it's the leftovers! Let me kill this.

B

uh Cancel, let's do it one more time, let's jump, uh overwrite, yes and now. I should have all this configuration that we have from this Messenger Service and from um attachment service. So what I can do now? I I can do deck sync and instead of us I said West one one, so I I downloaded this configuration for your assist and now I apply this configuration to this one.

B

Okay, so inside file we are going here, I'm just saying it's: okay,.

B

um Let's see why it is that connect on time group runtime groups not found Geo, ah it's not us West yeah. My runtime group named, is.

B

Us West one: it's not us, West one. It's a Euro, sos1 right.

D

B

Like just like, let's I'm, not sure, if I will be able to change this right now we run off time but hey.

B

Let's do this one.

B

uh Okay, let's do.

B

Runtime group not found how I can specify, ah apparently, for uh for this instance, I need to create for Europe No. It should work, um should work with the same global uh Europe.

B

um It should work with my why it's not finding my runtime group, let's see deck sync.

B

Sue, let's see if we need to specify kind of like a geography somehow um uh to to do config uh to to connect address token, the color.

B

D

B

The address ah okay, so I need to point this to different uh different API, uh so in this case it needs to needs to hit my your European API service, um where I can find this.

B

Let's see Docs connect because by default hitting the um the European.

B

um Connect configuration service Hub, let's see um time.

B

C

So you cannot connect from the from the con connect to my cluster.

B

No no I need to um region specific. You see, um I need to have this foreign.

B

Can I do this? Let's see um deck ping.

B

And I can do.

B

um It's it's just my local thing. So just let me do this. One.

B

Connect address.

B

By default, it needs to be something like this, but in this case I assume this is going to be our.

B

uh It's our region.

D

B

404, let's see if it's a global.

B

B

um Connect identity, runtime groups, API documentation, let's see us or Eau API con V2. Okay, we do you API conch, V2,.

B

B

B

That's weird, um so in this case let me do this. Let me do this manually right, so we can do this manually.

B

Let me try this one more time, so I will remove this. This runtime group from the file we're not going to use any transform and, let's see.

B

The sync verbose.

B

um Connect runtime group, Geo messenger US1 runtime group not found.

B

Let's create this manually, I need to figure out, maybe I'm forgetting some flag. um We go into the service hub, new service display name, messenger.

B

B

uh We're gonna save it.

B

We're gonna do a new version.

B

V1 runtime group here.

B

Implementation URL, we do have it.

B

Http next in route name here, which is, do create.

B

B

And path excels this one create so now, what's the IP address, what was the European IP address? Dennis.

C

Send it to you in Chrome, proxy right, yeah, service, okay,.

C

Okay, in telegram.

D

C

B

Can try to open it? Oh you're still setting up no I think so that's my.

C

That's my United States AP address.

B

Perfect box perfect and.

C

This is the European one, wonderful, it's loaded so right now, I'm from the United States East Coast, connected to the instance in Europe West Somewhere in Germany, yeah.

B

And let me see where's the network console so I can show you that we actually go and everything goes through conch and if I will see any request and I see headers and it served via um via quantity price. So and also in this case, I will be able to see.

B

um I should be able to see this guy right, so the traffic is going in uh what, for some reasons, 302 um early direct. It's.

C

I I think that we have not fully set up their uh attachments, microservice.

D

But it's secondary I can send messages.

C

To Europe from Europe but I try to load. That's.

B

C

There is something not ready, yeah.

B

Would you say application name attachment foreign and we can create this one create new version. uh We call it V1 a runtime group here create and implementation. You need to have a two two um right. You need to have a two uh two routes. One route with.

C

I, don't need pin at least the upload route. The pin route was used. You know just for testing just the upload and.

B

D

C

So yeah, when I hit the trout now.

B

Go to the try it again.

C

It's right in here: okay, so.

B

C

Going that's my that's my instance in Europe. Let me just refresh it and uh well: oh nope, nope nope nope uh not found 404 error, not found path. We messed up with their.

B

um With the with route configuration, let me see.

C

Let me check the locks for the attachments micro service.

D

C

While you're get receives anything, get uh okay locks uh for their deployments, attachments.

C

B

Runtime manager we're going into we're going in.

C

No, it has not received any. It means that, like the messenger microservice here, something is wrong with the route.

B

Let me figure out service version edit Implement edit implementation protocol host.

B

um Routes upload name up; okay, let's see.

B

B

C

Four or four, probably the the name, is upload.

D

C

Url check it, the URL is this: the name of the route is upload route, but probably the.

B

C

String path, string path needs to be sold to false.

B

Okay, three path, yeah. That makes sense yes strip, that, uh let me show you one more time what I have here so upload is my route, a Street Path here.

C

False it has to be false.

B

It has to be false: okay,.

B

Now, try try to do upload.

C

Right man, one more time.

C

Okay, deleting.

C

B

Winning winning okay. So, let's see if we have a API hit over here in our um let's do refresh one more time um we're getting to upload route in Europe. uh There's one 404. There should be.

B

One hit in Europe: there should be one at.

C

Least, attachments or I can load more I.

B

Mean let's try them one more time. Maybe it's not not super? No! It's it's here! It's here! You know this is one 404 that we failed in now. It's a strip path that will continue to do upload to the service beautiful now last thing. Last but not least, Global load balancer, let's go let's.

C

Do it the load balancer Fox is already ready uh here is because it takes usually 10-15 minutes to deploy it and propagate the IP address to all of their regions. That, where you would like your application uh to be deployed- and here is the command I- will use Cube control. True get the IP address of the load. Balancer, okay, I have to be the load. Balancer is deployed through the United States, which means that I need to change. My uh cluster use context, G key Us, East, 4.

C

I, switch to the United States and give me the IP address. Okay, it's not very important important person, it's virtual AP address I'm, sending it to you and also I will open it on my end. So right now we both live in the United States and, if I open it to my end, it has to select the instance in the United States as well. So if you open the attachments, I can also show The Source Code, the logs all right load, balancer load balancer forwarded the request to the United States East.

C

Let me show the logs I will stop this one, because those were from their instance in Europe and I need to open this okay. So the last time was 18 15th.

B

And it's in time time in Europe, somewhere in Germany, I.

C

Guess, no, that's! That's! That's United States.

B

C

Just sent a message: yeah, let me try to load a picture through the global load balancer. uh What else I have? Oh I? Have the architectural diagram load it going back to the logs it's here so.

B

C

That's the beauty of the load balancer! That's the picture, I've just loaded! That's the beauty of the balancer Fox! You can connect to the load. Balancer you have a single IP address that IP address can be mapped to your public DNS name. You can give this IP address to your front-end developers to your mobile developers to anyone who is building, let's say some UI on top of your apis on top of your microservices and that's it then, when you send the request to that load, balancer the load balancer decides, it sees all right.

C

You live in the United, States and I have and I have live instances of corn and their back end in the United States. Those instances will be serving my requests.

C

Let's do this, let's just let's disrupt this I want to show that load. Balancer also can forward the traffic automatically to Europe. If my instance in the United States is down just two more minutes and we are closing right.

D

C

Good, so what I do um Cube CTL get deployments?

C

I will kill you. This deployment I will kill the main messenger instance, the main backend in the United States. This is how I emulate outage in the U.S east for region. So let's do it group CTL delete deployment.

C

This one done so now, yeah the lock is down because there that is down. That was the IP address. The direct AP address, I think yeah. It's not going to load yeah the instance in the United States is down, but that's the IP address of the global load. Balancer. Let me try to type something you see the reject. You know what happened.

C

Load balancer detected that the instance in the United States is down and it's down this one is not available and the load balancer started forward in my requests to Europe, because that instance is up and live. Let's connect.

C

So I want to see the logs to do that. Let me switch to the European cluster.

C

Config use context, jke Europe West, one I'm in Europe and I'm, seeing the logs okay, that's the instance in Europe. Sending the message through the load. Balancer goes to Europe, okay, let's pre-load the picture at least I can what else I have I I should show you something else uh uh yeah. Let's do ssds.

C

Load it- and here is in the logs of the instance in Europe I- have can find this file. The file is in Google storage and just the final check yeah, that's the picture. So that's how it works. It took us a little bit more time to replicate this architecture to Europe, but once you have all the moving pieces ready, you just start using the lower balancer and the magic comes through.

B

So I am confirming that we were doing everything right with the commands and I'm trying to see why our commands is not uh didn't working, um but uh otherwise I think it was a huge success for, for this show we improvised a lot as a matter of fact, like some of the things that we you changed on the Fly uh and it turns out to be great. In my opinion, I I, like the I, like the the the this collaboration. uh Well, you can always find again.

B

uh You can always find um Dennis where's uh where's, my where's, my window. You can find a dentist in uh his uh Twitter I guess uh you can, or you can find him somewhere in uh in LinkedIn, and ask the question about the hugobyte and distributed microservices. You can find me. You know where to find me. You can find me on Twitter. You can find me YouTube. You can find me everywhere um and ask questions about Quan and all this kind of stuff like next time we need to.

B

uh We need to move this application to the service mesh and uh so, in this case, we'll have a um much much cooler use case, because you know service mesh is cool, it's cool, a cool thing to to have and uh Dennis. Thank you so much for being part of Kong Builders. It was great having you, um it was great. Show I really enjoyed this collaboration, uh any last words for our audience.

C

Yeah, the last word actually folks, you first step and I can actually I have a slide just one one, one last minute: let's.

B

C

It I have a slide. I just want people to see it. I want people to see it. uh Where is my slide this one? Because how do you start with all this mess that we just showed you if you're creating? Let's say the new application? You have existing one and you're thinking all right guys.

C

You showed me multiple regions, different countries, but it's not for me, yet it might be for you at some point in time when you build your next application, even if you are not looking to deploy across the country deploy at least your application instances. Your code, your database across several availability zones within a single region, and this way what we will achieve at least we will achieve High availability and performance within one region. It's good right.

C

If your availability Zone goes down, you are up and running and second, you will already practice how to develop this geo-distributed applications. Then well, let's say some month from the management comes and says all right. We are successful in the United States right now. I want to spend your solution to Europe, to Asia to South America, and you kind of you can respond. I'm ready, you're ready, because your architecture was already appropriated.

C

It was executed within a single region across multiple exists and now what you need to do just to deploy it to other regions in the cloud. So that's my advice, build and design in a way so that your application runs across multiple azs or regions. There are a lot of good reasons why you would want to do that.

B

And with this I think uh um just a quick confirmation that um I, like I like when this happens uh like when the stream is over I use wrong command. So, instead of like a clock, address, I need should use, connect address and after that everything works so like I didn't I'm, not total screw up. So that's that's all good Dennis again. Thank you. So much for being part of connect. Builders everybody uh subscribe to this YouTube channel where's. My subscribe button where's my magical, subscribe button. Please um leave your comments in the show.

B

um Leave the comments. What you want to learn uh the next. The next episode is going to be in in two weeks. We probably will talk about open Telemetry, so stay tuned. It's gonna be awesome, and with this my name is Victor gamof and, as always, have a nice day.