xiujunma/kotless

Kotlin Serverless Framework

Kotless

Kotless stands for Kotlin serverless framework.

Its focus lies in reducing the routine of serverless deployment creation by generating it straight from the code of the application itself.

So, simply speaking, Kotless gives you one magic button to deploy your Web application as a serverless application on AWS!

Kotless consists of two main parts:

• DSL provides a way of defining serverless applications. There are three DSLs supported:
  • Kotless DSL — Kotless own DSL that provides annotations to declare routing, scheduled events, etc.
  • Ktor — Ktor engine that is introspected by Kotless. Use standard Ktor syntax and Kotless will generate deployment.
  • Spring Boot — Spring Boot serverless container that is introspected by Kotless. Use standard Spring syntax and Kotless will generate deployment.
• Kotless Gradle Plugin provides a way of deploying serverless application. For that, it:
  • performs the tasks of generating Terraform code from the application code and, subsequently, deploying it to AWS;
  • runs application locally, emulates the AWS environment and provides the possibility for in-IDE debugging.

One of the key features of Kotless is its ability to embed into existing applications. Kotless makes super easy deployment of existing Spring and Ktor applications to AWS serverless platform.

Getting started

Kotless uses Gradle to wrap around the existing build process and insert the deployment into it.

Basically, if you already use Gradle, you only need to do two things.

Firstly, set up the Kotless Gradle plugin. You need to apply the plugin:

//Imports necessary for this example
import io.kotless.plugin.gradle.dsl.Webapp.Route53
import io.kotless.plugin.gradle.dsl.kotless

//Group may be used by Kotless DSL to reduce number of introspected classes by package
//So, don't forget to set it
group = "org.example"
version = "0.1.0"


plugins {
    //Version of Kotlin should be 1.3.72+
    kotlin("jvm") version "1.3.72" apply true

    id("io.kotless") version "0.1.6" apply true
}

Secondly, add Kotless DSL (or Ktor, or Spring Boot) as a library to your application:

repositories {
    jcenter()
}

dependencies {
    implementation("io.kotless", "kotless-lang", "0.1.6")

    //or for Ktor (Note, that `ktor-lang` depends on Ktor version 1.3.2)
    //implementation("io.kotless", "ktor-lang", "0.1.6")

    //or for Spring Boot (Note, that `spring-boot-lang` depends on Spring Boot version 2.3.0.RELEASE)
    //implementation("io.kotless", "spring-boot-lang", "0.1.6")
}

Please note that if you use Ktor or Spring Boot you will need to replace existing in your project dependency with a special Kotless *-lang dependency. Also, after that you will need to align version of dependent libraries (like Spring Security) with version bundled in *-lang (see this paragraph)

This gives you access to DSL interfaces in your code and sets up a Lambda dispatcher inside your application.

If you don't have an AWS account — stop here. Now you can use local task to run the application locally and debug it. If you want to continue — create AWS accounts following simple instruction by Hadi Hariri.

If you have an AWS account and want to perform the real deployment — let's set up everything for it! It's rather simple:

kotless {
    config {
        bucket = "kotless.s3.example.com"

        terraform {
            profile = "example"
            region = "us-east-1"
        }
    }

    webapp {
        //Optional parameter, by default technical name will be generated
        route53 = Route53("kotless", "example.com")
    }
}

Here we set up the config of Kotless itself:

• the bucket, which will be used to store lambdas and configs;
• Terraform configuration with the name of the profile to access AWS.

Then we set up a specific application to deploy:

• Route53 alias for the resulting application (you need to pre-create ACM certificate for the DNS record).

And that's the whole setup!

Now you can create your first serverless application with Kotless DSL:

@Get("/")
fun main() = "Hello world!"

Or with Ktor:

class Server : Kotless() {
    override fun prepare(app: Application) {
        app.routing {
            get("/") {
                call.respondText { "Hello World!" }
            }
        }
    }
}

Or with Spring Boot:

@SpringBootApplication
open class Application : Kotless() {
    override val bootKlass: KClass<*> = this::class
}

@RestController
object Pages {
    @GetMapping("/")
    fun main() = "Hello World!"
}

Local start

Kotless-based applications can start locally as an HTTP server. This functionality is supported by all DSLs.

Moreover, Kotless local start may spin up an AWS emulation (docker required). Just instantiate your AWS service client using override for Kotless local starts:

val client = AmazonDynamoDBClientBuilder.standard().withKotlessLocal(AwsResource.DynamoDB).build()

And enable it in Gradle:

kotless {
    //<...>
    extensions {
        local {
            //enable AWS emulation (disabled by default)
            useAWSEmulation = true
        }
    }
}

During local run, LocalStack will be started and all clients will be pointed to its endpoint automatically.

Local start functionality does not require any access to cloud provider, so you may check how your application behaves without AWS account. Also, it gives you the possibility to debug your application locally from your IDE.

Integration with existing applications

Kotless is able to deploy existing Spring Boot or Ktor application to AWS serverless platform. To do it, you'll need to set up a plugin and replace existing dependency with the appropriate Kotless DSL.

For Ktor, you should replace existing engine ( e.g. implementation("io.ktor", "ktor-server-netty", "1.3.2")) with implementation("io.kotless", "ktor-lang", "0.1.6"). Note that this dependency bundles Ktor of version 1.3.2, so you may need to upgrade other Ktor libraries (like ktor-html-builder) to this version.

For Spring Boot you should replace the starter you use ( e.g. implementation("org.springframework.boot", "spring-boot-starter-web", "2.3.0.RELASE)) with implementation("io.kotless", "spring-boot-lang", "0.1.6"). Note that this dependency bundles Spring Boot of version 2.3.0.RELEASE, so you also may need to upgrade other Spring Boot libraries to this version.

Once it is done, you may hit deploy task and make your application serverless. Note, that you will still be able to run application locally via local Gradle task.

Advanced features

While Kotless can be used as a framework for the rapid creation of serverless applications, it has many more features covering different areas of application.

Including, but not limited to:

• Lambdas auto-warming — Kotless creates schedulers to execute warming sequences to never leave your lambdas cold. As a result, applications under moderate load are almost not vulnerable to cold-start problem.
• Permissions management — you can declare which permissions to which AWS resources are required for the application via annotations on Kotlin functions, classes or objects. Permissions will be granted automatically.
• Static resources — Kotless will deploy static resources to S3 and set up CDN for them. It may greatly improve the response time of your application and is supported by all DSLs.
• Scheduled events — Kotless sets up timers to execute @Scheduled jobs on schedule;
• Terraform extensions — Kotless-generated code can be extended by custom Terraform code;

Kotless is in active development, so we are currently working on extending this list with such features as:

• Support of other clouds — Kotless is based on a cloud-agnostic schema, so we are working on support of other clouds.
• Support of multiplatform applications — Kotless will not use any platform-specific libraries to give you a choice of a Lambda runtime (JVM/JS/Native).
• Versioned deployment — Kotless will be able to deploy several versions of the application and maintain one of them as active.
• Implicit permissions granting — Kotless will be able to deduce permissions from AWS SDK function calls.
• Events handlers support — Kotless will generate events subscriptions for properly annotated events handlers.

Beta release

Kotless beta releases include cutting-edge features that are under testing right now.

Since Kotless 0.1.7-beta-4 release we are introducing support for GraalVM. With one configuration variable you can convert your existing JVM applications into ultra-fast native applications.

Since Kotless 0.1.7-beta-5 release Kotless is migrating to Kotlin 1.4. Unfortunately, due to restrictions of current Kotless architecture it means that now Gradle 6.8.1+ is required to use Kotless.

Just specify the GraalVM runtime and let Kotless do everything for you.

kotless {
    webapp {
        lambda {
            runtime = Runtime.GraalVM
        }
    }
}

Examples

Any explanation becomes much better with a proper example.

In the repository's examples folder, you can find example projects built with Kotless DSL:

• kotless/site — a site about Kotless written with Kotless DSL (site.kotless.io). This example demonstrates @StaticGet and @Get (static and dynamic routes) usage, as well as Link API.
• kotless/shortener — a simple URL shortener written with Kotless DSL (short.kotless.io). This example demonstrates @Get ( dynamic routes), @Scheduled (scheduled lambdas), Permissions API (for DynamoDB access), and Terraform extensions.

Similar examples exist for Ktor:

• ktor/site — a site about Kotless written with Ktor (ktor.site.kotless.io). This example demonstrates static {...} and routing {...} usage.
• ktor/shortener — a simple URL shortener written with Ktor (ktor.short.kotless.io). This example demonstrates routing { ... } (dynamic routes), Permissions API (for DynamoDB access), and Terraform extensions.

And for Spring Boot:

• spring/site — a site about Kotless written with Spring Boot (spring.site.kotless.io). This example demonstrates usage of statics and @RestController.
• spring/shortener — a simple URL shortener written with Spring Boot (spring.short.kotless.io). This example demonstrates usage of @RestController (dynamic routes), Permissions API (for DynamoDB access), and Terraform extensions.

Want to know more?

You may take a look at Wiki where the client documentation on Kotless is located.

Apart from that, the Kotless code itself is widely documented, and you can take a look into its interfaces to get to know Kotless better.

You may ask questions and participate in discussions on #kotless channel in KotlinLang slack.

Acknowledgements

Special thanks to:

• Alexandra Pavlova (aka sunalex) for our beautiful logo;
• Yaroslav Golubev for help with documentation;
• Gregor Billing for the help with Gradle plugin and more.