Kong User Calls, 13 Jul 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Using OpenAPI to automatically configure the Kong Gateway

Description

For our July User Call, join Sven Walther, Principal Solutions Engineer EMEA, to talk about “Using OpenAPI to automatically configure the Kong Gateway”.

We are going to automate the creation of Kong services, routes and plugin configurations only based on an OpenAPI file. The tools we will use next to Kong itself are Insomnia/inso-cli, decK and the Kong Ingress Controller.

A

Well, hello: everyone thanks to all of you who are joining us today. um My name is dalia spaceva. I'm part of the developer marketing team here at kong um and I'd like to welcome you all to our july user call. So today we have a very special presentation for you coming from sven walter, who is a principal solutions engineer here in emea he will talk to us about how to use open api to automatically configure the com gateway. So we will take all the questions in the end of the presentation.

A

Please feel free to drop them in the q, a section or in the chat. um At the end of the presentation, I will also um fix the settings, so you can unmute yourself and ask the questions if you'd like um and yeah with that, I will hand it over um to sven. Take it away. Please.

B

Thank you yeah, so hello, my name is finn based out of berlin, as already having mentioned I'm a principal solution, engineer here at kong, meaning in my daily work, I'm doing presentations and workshops with possible customers, and one of the major topics we have is there about automation, automation and automation.

B

So today's session will be about how can we automate the full configuration of the con gateway only based on an open api file without doing anything manual in between um in general? um This talk will be less than an hour, so we have on our blog, so we will take way less than an hour. We'll have plenty of time for questions afterwards and there, and if you finish early, that would also be fine.

B

So, um let's get started, and I also see questions so just ask them and you'll answer them later on. So, first of all, let's talk about one more thing. Now, the one more thing is uh we as kong. We are hosting our summit, our conference uh in san francisco, and the good news is this year, it's once again in san francisco, so it has been virtual, the last two times because well we all know why uh and it's coming back to being in persons one uh in san francisco.

B

So if you want to see customers, users talking about use cases and so on and seeing the exciting new things happening, yeah check out it here. The link has just been posted into our chat there, so you can just register there.

B

So, let's start thinking about what is one of the major use cases. Why kong is that successful? I'm personally in the ibm space since many years, and one of the reasons why I have joined kong roughly four years ago in the early days is exactly about this one statement and those having seen me in the past. I know I'm telling this every time I'm showing kong. uh Everything in kong is an api.

B

Really everything is api, so we have been built from the ground up being automated uh in every kind of aspect for you as the open source users. That's even the only way to configure it. Everything is not an api, so you would have no user interface, nothing to click. You do everything in an automated fashion anyway.

B

So let's try first working with the api and create a service um which we want to expose using our existing gateway. um So we will use this um command line here, but you can use anything to automate com. You can use curl scripts. You can use, as I will do, hd pi, which is a similar tool, but it is more fancy than that and you can use your ansible script. You can use github actions. The azure pipeline. You can use insomnia.

B

Whatever is able to make an http call is able to make a call to kong, configure it and wk.

B

So, let's do that! Here's the first hands-on thing, so we have here a running installation on my local host. So I can just say I want to create a new service, so I'm going to the services endpoint on my sim. I want to call it my service uh and there's a url, so we want to take something like an open endpoint in the internet.

B

We will just take hdbin.org, we just say we want to get that in our system, so we have already created a service in our um gateway and then we go ahead and we want to also get a route. So someone can access it. So we say for that service we want to create a route and the pass would be something like my service, for example.

B

So this automation yeah um that you can create services and routes which are then already active. So we can just verify that by just looking if this is working go into our proxy and check. If we do have the my service endpoint answering in our proxy- and indeed it does so we're getting it from kong in there, this has been ground breaking when we released com. Initially, this is the first major thing why people come to kong when they think about automation. Everything is api and we really mean everything, but times have changed yeah.

B

This is still an awesome feature, but it's the so-called imperative way of configuring things. So, where you specify exactly, I want to have this service. I won't have this route exactly now. Please do it right now, that's called imperative, so we go ahead, and nowadays we talk now more about the idea of what is about a declarative configuration yeah everything nowadays. What you do everywhere is not a rest api anymore with conversion. This is about yaml files. Think about it everything everywhere.

B

Every kind of automation tool nowadays has switched to a yaml file configuration which is called declaration, because we just explain what we want to make as a result, and then the system has to see how it comes to that to do so, there's a tool which we also write as open sourced next to kong. So you can put it into your machine where you're running your csd pipelines, this tool is called deck deck by the way declarative configuration of cong.

B

That's what it stands for is a tool where you can configure kong completely based on a yaml file. I have a very short version of a yaml file here you can see here. I want to create obviously a service to the same back-end system, and I want to create a new route in here. What you do is you use this um dac file and then configure cong using it. So let's do that.

B

I have this yaml file here on my local hard disk and you can just say um deck sync yeah based on which xml file- and they do have here this yellow file, which you are just seeing here and then you can specify if you want to right everything or if you want to say only a specific subpart should be overwritten. We can have text everywhere, as you might know, in kong, um so we say I only want to create a new definition in here.

B

What we have gotten now is another service in kong and another routing con, and it's the same instance. So we can still use the rest api and see which services do we have and we will see we do have here our yaml service and we do have the original mysteries which created using the rest api. So the same system, you can define what makes the most sense for you, configuring it maybe even doing a mixture, maybe one department that does one way the other one is the other way.

B

So it's the same system just a different way, which you can use additionally to configure kong. So how do we get such files? Yeah? We can create a manual, we can go ahead and adjust it and create that file in a manual way, because I know the format. It's quite easy, quite readable, and I can just add anything I wanted there. Plugins consumers whatever is in there. A second way to get such a file would be.

B

We can also use the deck tool and say: please dump me the current configuration of that cluster to my local hard disk. I just say deck dumb. What I'm getting now is a combined other file of everything available in my cluster, so I could, for example, see here I do have here my service. I credit this to my service to rest api yeah with this corresponding route, and then we see here I do have here the yamaha cells that just created using the yaml file and they can change things here.

B

For example, I can go ahead and say my service. uh I want to change that that it's also available using my rest created service, for example, that should have two listeners. The dactyl has more things like detective yeah. What would change if I apply that one it will say? Okay, I will update that exactly one thing: yeah, this path will be added in here and dex will send that back into my running cluster.

B

This is the way how I would say nowadays in a fresh installation and new engagement I have with customers. Everybody wants to go so the operator fail using the rest. Api is something which you can do things very quickly and try it out, but in real world csd pipeline um that happens normally based uh on the deck level format, but still yeah. If you look at our topic today, I have not spoken about open api so far, so where do we get um that open api from uh waiting an open eye into such an installation?

B

So let's go to the next step. Let's say we want now to use. Oh there's one more thing and you can do the same thing with kubernetes ingress resources. Also yaml files, yeah same idea same idea, concept services described in there and then you would use. Instead, you see an apply so whatever you prefer, uh you can do it. So let's talk about our main topic: open api.com, so this is a typical open api. I just have used it very typical pet store one to give you an example which everybody knows now.

B

It tells me where's my backend servers and that's because there is a pass pass for example. So how do we get this one into kong?

B

Well, we provide you with a library for that there's a library which you can get as an npm. So let's have a look at this one.

B

Where you get a library where's, my link going.

B

Maybe rightly just open it, let's take this one, no documentation ah dependencies at least yet here is so. We have all the documentation uh also available uh in here, where you can automate the pros. The processing of an open api into uh number file understands, and you can do that for both the yaml format from deck, but also for the english controller. So that would be a way yeah.

B

You write your own program, your own library and then say I'm using this library, which then auto translates the open api into the kong format, but we are not stopping there, so we are eating our own dog food.

B

This library is for you if you want to make it in the way you preferred, but the next step is here that we provide another open source tool to you, which is called insomnia for those not knowing about insomnia, it's most known for being an open api editor and an open api test tool, rest api and graphql test tool.

B

In the end, let me show you what insomnia looks like to give you a first feeling, and then we will look how we can use this tool to get from an open api where we have described the servers, the paths and even plugins to services router plugins in our comm world, so first of all yeah insomnia available at insomnia.dressed as the homepage. If you want to download it, you just go there and get it so insomnia gives you first of all an editor for an open api file.

B

This editor, you can use the create want to read it want to visualize everything which is available in there. um It helps you to do things right. So, for example, if you mistype something yeah, it will tell you you have mistyped something in there. That's the first use case of informnia. For us.

B

The first important use case is here on that small triangle and we have here an open api file and we have here an option to say: please export that open api to declarative configuration so we're getting here a file understood by deck, which we have used recently or we can see here. We get kubernetes increased resources, which can also be done. So we can just say I'm having your open api.

B

I'm copying that's to my clipboard uh and then putting it on my local artist and then I can use deck to sync that into my con installation that we will see in a second, huh but before we do so some more things. We are talking about automation and we do have an open api.

B

Can we do more automation than just configuring com? And the answer is yes and we can use this open api as a source for manual testing? So if you go here on the debug tab, you can see all the requests which are available from the specification.

B

So you can just click. I want to see them by text and get get find by text. I want to find it by id I'm getting here by id and I can just make calls- and I'm getting here, for example, that pet id number nine I've set it here up in my test environment, I'm getting here um kind of start assault, for example.

B

Next, to that, I can't create tests, meaning I can just say a new test huh live. I want that test being on the endpoint of deletes the pad. Then I can see getting out the right response. Whatever is in place in here.

B

So we can do tests. I do here test first up, I pass. Is there status property and is the status and this one uh in fact sold, and they say? Yes, it's starter sort, as we have just seen in our manual tests now we just saw it start or sold and that's exactly what our tests are doing here. I will come back to that. Why is that important later, but now, let's think about again, we are exporting here decorative configuration and can copy it. Nobody wants to click such things yeah in a pipeline.

B

You want to make it automatically.

B

um So we will instead of using this copied version. Do the next thing yeah we do have not only insomnia, which can do that. All we have just seen his creatures, and there is the next one, which is called enzo cli in the cli same homepage, once again, insomnia addressed is the command line version of insomnia having basically the same functionality, so I just put in here the enzo help. So, for example, it can do a lint checking or another example.

B

It can generate um a deck file or a congress resource a file in here by just using those parameters. So the same thing which we just did with clicking it in an automated flow would be lumps like that. I want to generate a configuration, then I have to define which open api file I want and then which output file I want to have. So I go ahead and go into my open api folder.

B

I do have here this pet store with kios the one we have just seen in our example in insomnia, and I just execute that command and I'm getting automatically now a new file, as defined here pet store tag import. If you look at this, let's look at pet store deck import. You can see. We are getting the now already known to us format of deck where we're getting services uh and we're getting all the different routes extracted from that open api file.

B

It's a parameter you can set if you want to create a deck file or if you want to create a con ingress resource. So we are going now with the tech example, but it's just. The flow itself is the same, just generated getting the ingress resources and then you would make a cubesat apply.

B

We instead will now use um the deck tool now to import that we have now imported that into.

B

I will now import that into our installation, so we have created that file and now we know this newly created file and import it into our running com cluster again. So what is the command yeah? It is tag sync yeah string it into my running installation, but that's the source file, and here I will also set some text so that I can later on um auto apply configurations to them.

B

So you can see, we have created one service called swagger pet store. We have here multiple routes, yeah for all the difference, find pets by status and so on, based on the open api file- and we even have one plug-in in here. So why is the plug-in in there? You might remember that we have the ability here I said on the post. Call you have this x, hyphen kong, minus plugin, um special tag. This you can use for any plugin. I see questions already.

B

I will answer them later in more detail, but you can use any plugin in here. Any plugin can be used and then you can just say, for example, here um config and then you would say um parameter one is a whatever. So anything you would configure on a plugin. You can do also here for any plugin and you can define them globally, as also here as I do here on the route level.

B

So that's what has happened. We have created all those different passes as routes and con and they should work now. So, let's test them yeah for testing insomnia. Once again, it's a nice tool. Now we can just say we do have different environments and they would have the live font at pet store which we have used before now. I can go to localhost so my base url is now in fact, switching to my localhost installation, and I can say, can I now do the same call again against localhost and yes, it's localhost.

B

You can see it's coming through kong yeah via kong in the end here, and also you can run once again the same tests if you want uh against the local one and still we get all everything is passed. So that's already a fully imported uh configuration out of an open api file. We cannot use it manually. You can just say give me on my proxy pads, the id let's say: eight.

B

Pet, that's autonomous pet. Eight pay, eight is not available, but paid. Nine, as we know, is an available one. So it's a live one proxied by kong and the only source of truth we had was the open api iphone in the beginning.

B

So we have seen that we have something about the plugin attached. Let's validate that, so we have seen it was defined in the open api on the post endpoint. So if someone wants to create something in um that, he needs to have an api key. So if I try to post something in here, our api will deny it because we say there is not the I key in place.

B

This um all is um affect everything we need to do. Yeah. We have the full flow already up and running now. So if we uh look at the full flow, how it could look like, um I do have here created a small batch script. The first thing is something I have not even shown that I can also do. For example, validations yeah. The introsomia tool commander can just do the same things as the big window desktop application. So I can, for example, check. Is this a valid open api file? If it would not be available?

B

Maybe I filed the whole flow would make no sense. We could just stop our completion here. Then we can run the same tests yeah. We have just seen where we are just seeing that we have this uh all passing. Then we are generating using indegenerate, config or deck file, and then we are striking that into our running cluster, and this is a typical example how customers are using the whole flow. So I can just do the same thing from opm to kong that everything we have just done now we are checking. Is it available?

B

Maybe I file. We are running the same tests in here and we already have imported this one, so it is in fact nothing has to be imported anymore in here. That is the pure magic. The combination of insole, cli and deck will provide you a full, automated workflow, from an open api to com.

B

As a last thing, everything I have shown you, I have created a small github repository. I will chat the link in a in a second um everything on my github, where all the examples- the open api files, the script we have just been running, the the deck commands everything machine, including the slides we have just seen here- is available. So if you just look at this one in here, you are getting on github and all the yaml files, the example yaml file. We have used for the first explanation. We see audio api file.

B

We even see an example for the invis controller, which we have not really touched on right now, but um if you want to go to english way, uh instead, you can use this one, and then we have uh all the open api files yeah one with out and plug-in configured one with the plugin configured and then the whole flow. We have just run automatically as a batch tip plus in the end, the same slides as we have just seen here on our call.

B

So that's what I wanted to show you today um for my customers, I'm talking regular on this is the end goal of their automation pipeline, because that means that in the end they can fully automate everything based on a file. Everybody understands without them needing to know about kong. At all.

B

You can just say we are creating an open api file and then you can even choose if it's really manually like written by someone in insomnia or if it's coming something from an automated flow like springboot, where you have this slash, swagger.json endpoint automatically so generated by code, where it's coming from your beta. I we don't care yeah, we just need to have it as a source and you can just have a script, curl download it and then use insole cli, convert it and make a deck um import in there.

B

So that's what's the full flow, and I see we have quite some questions, but before we do so, I will just post this link in here in the chat, all the materials yeah. And now we have a look at the questions we have gotten.

A

Thank you so much sven. That was awesome. um Let's start with the q, a section I think most of the questions are there yeah awesome.

B

So um we do have the question here. Looking at the yaml configuration makes me wonder if there are sdks for kong, are there any such sdks for kong? um There is not a direct sdk for creating the yaml files, so you have tools like that. We just have seen the open api.com, which auto generate those files, but you can also use kong, as you have just seen to generate um examples. Live customers putting something in a con gateway, a short living instance and just dumping that configuration and then use that as a basement for undead.

B

The second question is about insomnia: can manage complex comp plugins like open connect; indeed it can any plug-in can be configured I'm using uh there. um So if you want to see some more complex examples uh which I have in mind here, uh like the open connect, one which I have not shown today out of the reason we are focusing today on open source- and I do not want to show you uh enterprise plugins, but you can do really here. Config configurations like client id client secret claims, three direct urls.

B

Anything like that um urls everything which is configurable on the plugin can be attached here to this um configuration, if you just say, config and then all the parameters all plugins can be configured in there.

B

Do deck work with congress controller or only with services, routes, um deck and conquer controller are two different way to achieve, basically the same, so they do not work with each other yeah. They just do this similar things on different locations. Deck is the tool to configure kong. Regardless, where kong is running, you can use use kong to configure a used deck configure kong on an ec2 instance. You can run it on on a virtual machine locally. You can do it on a kubernetes cluster.

B

Wherever you have access to the administration api of kong, you can use deck to configure con. The ingus controller with the same idea is for the kubernetes english resources only so you can choose which one you prefer. I have even customers doing both they say they do have apis running in kubernetes for those apis.

B

They are using the ingress controller to configure everything what is already in that kubernetes cluster, but they also do have then um apis being outside of the kubernetes, where I say that we have still have some older apis, which we still want to use kong, and then they use deck to configure those. Even an extra okay, next questions.

B

Can you clarify something about dick if you save your confidence as individual file, a service with its routes and begins, you see, do you have to pass all of the files to dachshund or do you have to have everything tagged? First, um if you have a deck dump or a handwritten deck file, um and you make a deck sync with any without any text, and you can write the text both in the yaml file or, as I just did in my example, I have put them into um the zip command.

B

So if you look here um the command we have been using in here, I put it here in the select tag in here, but it can also be in the um yaml file itself.

B

If you don't do that, you have to if neither of those are set call will see this as a full new configuration and replace everything in your cluster, so never make a deck string without any text. If you don't want to re-sync everything in your cluster, having text is something crucial for you. Otherwise, if you say you want to change one thing, you have to just put one deck file in there and you have already 99 services applied in there.

B

uh If you don't have any text in place, 99 get deleted and you get one afterwards in the cluster, so be careful always have text when working with that and you can also generate them from the open api. For example, that's something many customers do or in a csd pipeline. You can also auto generate them, maybe based on the github repo name or something like that, um but have text attached. Otherwise, you get everything, gets synced and may be wiped.

B

um Next one does comp provide any tools to for fuzzy testing apis um dennis daniels. Can you go on microphone and give me some insight? What you mean by fuzzy testing here.

B

Audio check, yes,.

A

Hello, hello: I was wondering about the being able to throw fuzzy tests or fuzz testing using your tool to verify whether or not the api is going to respond properly or respond correctly. If somebody's trying to if somebody's trying to attack the api or manipulate the api in an undesired manner so fast.

B

A

Bus testing to make sure that nobody, you know that the api is resilient to people who are trying to poke it to do something bad.

B

Yeah the two answers to the question, the first one, so it has no artificial intelligence, doing fuzzy things on our side, but, as you have seen on the side, while deploying it yeah, you have those tests here uh which can be as complex as you want. Oh that's, the other one, that's the one. We have just got the time now, so we have here all those tests and they can be as complex as you want. You can inject here anything it's in india and javascript, where you can do be very creative on your tests.

B

What you would do, but there's no auto of artificial stuff in there. That's the one side of the coin that we can make sure that we test it before we deploy it.

B

The second thing and that's another thing which I on purpose, have not shown because it's an enterprise feature, and we can also make sure that everything which is incoming conforms to these standards having been defined in the open api file. So, for example, in this here I do have said- I do have your parameters and those parameters um are in. uh Where are they? They are type integer and they are query parameter or they are in the past, and all things have integer between 0 and 10..

B

This information you can provide in the open api and down here you see, I'm enabling here another plugin called the request. Validator and the only thing I'm doing is I'm activating it. I'm not configuring, I'm just saying: please do it and insomnia or in social I will automatically say okay. I know this plug-in.

B

I will generate the corresponding json schema definition for the plugin called request for the data and then every call which which is incoming will be validated against this schema and say if you are not providing an integer between 0 and 10, the gateway itself will already um deny that call before it even hits the backend system.

B

If you want, I can show that uh at the end. I do not want to show an enterprise feature yet because this recreational data plugin, is an enterprise only plugin, um so I can show it if you want at the end.

B

um Hi sven how to configure contact ins using open api specification, I think we have seen that now in multiple locations and we just say x, hyphen, login and then minus that's the name of the plugin, and then we have all the parameters uh below it.

B

Share a recent business case of how congress deck was used business cases. Oh, I like that one um yeah in general, we have many customers um which are more and more saying open. Api is the central point of truth for everything it's um for about uh this specification before they even code. It's like a contract in there, um so that gets in the direction of quality and the business case behind the set that nothing goes wrong um with developing.

B

So they say we know in advance what we want, and we don't have this typical cycle between business ones develop a little something different and back and forth. So we can just say: okay, we have agreed on that one, so it's taken in there. So the business case here is a shorter um cycle um on the creation, um the deck one then is um shrinking down. First of all, the csd pipeline. So it's way more easy because you already have that open api file. So you don't need to have special tools.

B

You need to create and people being trained. Yeah. That's also cast a cost, a block in there trained to use kong, even though kong is way easier than any other gateway in the market. You still have to learn something through this, so if you use augur vi, no training of admins is needed at all, and that brings me to the last uh and biggest cost savings thing. That means you can roll out all your apis to the con gateway. Basically, without doing anything you just need to have.

B

If you have open apis, which you normally should have already, you can just completely roll out the con gateway configuration uh from your existing, so migration use cases are very thankful things for that. We just say: if you have your api, just make a pipeline push them into kong and you have all of them running in congress. So it's in the end. It's all about agility and time to market, which are the use cases behind deck in here.

B

Is there a way to use the kong library insomnia without adding annotations to open file, otherwise between open api and conc? Yes, um there's a second way to do so, so you can say I want only have to have a vanilla, open api file, like my um the one, without a plug-in. So I have it even here as an example in that folder. So if you look here that the vanilla pet store yaml file, there is no plug-in configuration there.

B

If you want to do it in that way, yeah you need to or to generate the plug in configuration on another way from your csct pipeline.

B

Wherever it's coming from and the um the important part would be here, then you would say the pet store deck and they would say another one as a um my plugin configs dot number, where that is coming from that's up to you yeah, but you can have multiple source files in one deck soon command, and this then, would only host the conspiracy things and that would only host the ones coming from open api, but they would be seen as one input and then would be synced in one go.

B

Hi for existing ipis configured in hong kong is there any tooling that could take the config and create open api specs. um In fact, we have done that in the past, so there was a tool in our enterprise offering where we had a tooling, where we did it the other way round.

B

um We have decided to discontinue this I think a year ago than I did because, let's all be honest, all the generated um open api files which are not coming from the code yeah. The code knows a lot, the gateway knows about routes and then parameters, maybe but not really, details about it.

B

They were just not quality enough, so we have said it's there's no real values. We had them in place. We have customers using them. They got their open api files, but the real world uses was saying: okay, yeah, okay, it's maybe even easier to start writing it from scratch than starting with the skeleton where we have to anything manual. So we have this continued features because it was not the real world feature being used.

B

Could you talk about the benefits of using deck versus terraform? Well, um it's not an either or so many people you are using at the deck tool in the terraform script somewhere. um If you have an automation, flow, very auto, generating all configurations for kong in a terraform yeah. If it's already in place, go ahead with that, if you say it's, it's better! Your csc pipeline go ahead with deck. What you get out of the box is a functional tool which already has all the string capabilities in kong up and running.

B

So we don't need to take care of the contact part at all, because it's already done um so it's not a hard decision. It's what fits better. Your infrastructure and your needs.

B

Any support for schema validation- I just talked about that and I said I will show that. Maybe in the very end it's an enterprise feature. But yes, then we have the validation in place. uh Also based on the open api file, can we integrate portal department? Yes, um another enterprise feature. I see here a pattern. um Everything in kong is an api, so um the portal deployment in the end.

B

So if we look at our, um we do have our headstock file, for example the open api file and there you can use this one and just say in our comp portal. Let me just validate. I have one enabled, because I was not ready to look if I should get an enterprise feature, but it should be one there right, you see. Is there an enter portal up and running on my machine?

B

Yes, so we do have here an example. One was in there and then I can just go ahead and say I'm accessing in an automated flow, this develop portal and say: please update the specification also to documentation purposes. I will just say: put it in a path called specs: let's stay with the same name and then also get the contents from that file. So that's something outside of deck. That's something that we have everything as an api.

B

So we can update the documentation in our developer portal, and I see speaker might be a bad example because I already have speaker uh examples here have a second person, but you see there's a second one. Now on version one zero, zero one version o2, which one have we just uploaded pet store deck imports.

B

So we have just uploaded the version with one or two: that's only just uploaded uh india, so yeah, that's something we typically do in the full flow, where we go from the open api from the design phase up to the documentation phase in the same script,.

B

Next one, could you discuss role of chrome, developer portal in conjunction with obvious articles of config language such as dag yeah developer portal is about documentation, as we can see here, it's about documentation and getting the um people to know how it works and that can be based on the same api file so but the deck is configured the gateway, so it does the enforcement of routes and services, whereas the developer is about documentating due to the outside world.

B

So that is why how they work together and there is no hard linkage, so we can have uh files uh using deck for the gateway, but there is no must have in the dev portal and also not the other way around yeah. Next, one with insomnia is it possible to manage multiple spec files within the same git repository?

B

No, no, it's one workspace. So it's one gift repository per web project, but it's um one insomnia. Many different documents can be deployed, but it's different repositories so typically also one open i5 belongs to one back end and that has the same repository. Normally, um how are you handling the publicity exposed swagger? How we have just seen that a little bit there's something you need to do in between yeah? We just just have published now the hope api file just for the developer portal.

B

um The one thing you need to do is somehow exchange the host name, at least um depending on which you prefer I use for that. uh The tool called yq yq is like jq for those uh knowing that one tool, a tool which you can use to on the fly exchange values in a younger file. So I have in my csd background. One of the steps is that I take the op api file and exchange the internal hostname with the external gateway host name, and then I'm exporting it to the dev portal.

B

Next one. There are some cases where you use regex expressions in kong route to restrict the excess can that be configured using inso.

B

Yes, kind of it depends on how they look like, for example, um let's see here our pet store again, we have here all those different paths in there and there should be hopefully something here. We have here: pet id yeah pad slash pet id. If you look at to that route and I've used the comm manager to visualize that yeah, if you look at that created route on our server, so we have our swagger pet store and look at the route which one did we have.

B

It was get by deep find pets by text by start those I'm giving all of them get by id, get back pi id, and then our route automatically extracts that it needs to be here. A value in there which you can use, then for extract information. So we are using here a regular expression automatically um and extracted the data and can do them later on something like that closing once again, enterprise feature only manager next one. Where can we find documentation of what enzo can do with kong plugins, for example, the validation?

B

Again we just saw. How can we configure how to insomnia create the configuration by also going to drive the plugins things yeah? That is in fact, and I'm not sure why npm is not rendering this right now. um That is all documented here on that page, but somehow the npm.js.com page has some issues ah and we should solve. We just have to wait long enough and here's everything explained how to work with plugins here talks about the request for data plugins: how about security? How can I make all tools correct?

B

So all the things which are done are documented here also on the page. I'm not sure why mpmjs.com just did not run it, but everything is clearly explained how that works, how they, how text has handled and they are under the hood.

B

Do you have also to use your pipeline to sanitize the con conflict out of the open api?

B

um Sanitizing you mean if you want to remove the x minus com uh plug in things. The answer is yes, so we are not touching the open api itself, so you can do something like I just did with iq, which I fse would start with x, minus kong yeah just remove that in one go and the last one, and then we will as an extra bonus to the um orchestra data one hi.

B

Do customers typically expose separate developer portals for intel, and, oh that's, a really enterprise question: several developer portals for internal users, excellent users, anything in deck or other config to take what is internal versus external two questions? It depends.

B

um Customers either have one or two of them that you can do it in one dev portal, based on the permission structure that you say everybody gets as a dev portal, but not everybody sees everything so intel users just see more than that many customers uh make want to have it hardly cut just auto regulations, as I say they want to have an external developer portal. At the next delivery portal, the deck configuration has nothing to do uh with the dev portal. That configuration is about the gateway configuration of the developer partner configuration okay.

B

That was the last one and now, as promised one more thing about that. You see a live example how we can use the validation in there, and I do have this other uad generator service in here. Maybe I see we have this plug-in enabled and we do have some, for example, delay parameter between 0 and 10., so this is example I have laying somewhere else. I might also put that later on on the github for you, that you can also see that minecd.

B

And I'm doing a similar one where I'm going through publishing it to kong running all the tests. We have just seen it in the other examples and then I'm making here in the generate and then deploying everything to my kong installation.

B

So what has happened now in my con manager?.

B

Getting now for this one here, a new service called uid generator service yeah, uh with all the query, delay get and so on, and then delay yeah, the one we just looked at. We could automatically attach the request to a data plugin that is coming from this small tag. Yeah we want to have it enabled and then inso has to generate automatically schema, which can have a body schema. That was a question we had also previously. We can also do it on the body.

B

uh We can see a parameter schema, we say: there's a delay parameter, it's required. Why is it required? uh It says here it's the path, it's required true, so we extracted it from the open api um it's in the past and it should be in the chart between zero 10 and we can just try that out and we can go ahead and make some calls and say we say http on our proxy slash delay. The first thing say without the delay. We don't even say it's a valid call yeah.

B

If, okay, we say we say delay aaa now we say: okay, this is not conformed to the schema.

B

Then we can try one two: three, which is more than ten yeah, so alternate conform to the schema and then, if you go with one, we are getting a response x, so we can make all those checks based on your maybe high five, by the way, um the last um one more thing in here. If you say I want to be more helpful than that to my users, you can configure how the plugin should behave.

B

um So I can just go ahead and change the behavior of that plug-in now manually, but you could also automate it. Obviously, you can just say give them some more hints yeah and then it will tell by the way it is. It needs to be smaller than 10.. So it can it's a little bit security discussion. Do you want to give those hints to developers or not, if you want just say wrong or right, yeah, that's the one setting and the other one.

B

If you want to be helpful to developers, you can just say: okay, we tell you what you did wrong against the schema.

B

Okay, that has been a lot of questions any more questions. We will open up the microphone uh here as far as I know, so any direct interaction is now very well welcome.

A

Thank you, svan for answering all those questions so quickly and thanks everyone for all of your questions, really really awesome ones. Let's see if we have any last ones, please feel free to raise your hand. I will unmute you if you want to ask a question or drop it in the q, a section.

A

All right! Well, I think, that's all for today. Thank you very much sven what an awesome presentation thanks everyone for joining hope to see you in our august user call next month and have a great day great evening, see you next time, yeah cheers.