Building Complex Servers

Depending on the complexity of the code of your server, you might want to structure your codeone way or another. This page proposes some strategies to structure your code according to itscomplexity, adapting to its growth, while keeping it as simple as possible.

Table of contents:

• Hello World
• Defining modules
• Extracting routes
• Deployment and application.conf
• Health checks

Hello World

To get started with Ktor, you can start with an embeddedServer in a simple main function.

fun main(args: Array<String>) {
    embeddedServer(Netty, port = 8080) {
        routing {
            get("/") {
                call.respondText("Hello World!")
            }
        }
    }.start(wait = true)
}

This works fine to understand how Ktor works and to have all the application code availableat a glance.

Defining modules

You can extract the code configuring the server, also called a Ktor module, to an extension method:

fun main(args: Array<String>) {
    embeddedServer(Netty, port = 8080, module = Application::mainModule).start(wait = true)
}
fun Application.mainModule() {
    routing {
        get("/") {
            call.respondText("Hello World!")
        }
    }
}

Extracting routes

Once your code starts to grow, and you have more routes defined, you will probably want to splitthe code up instead of growing your main function indefinitely.

A simple way to do this, is to extract routes into extension methods using the Routing class as a receiver.

Depending on the size, maybe still keeping it in the same file or you can move it to other files:

fun main(args: Array<String>) {
    embeddedServer(Netty, port = 8080, module = Application::mainModule).start(wait = true)
}
fun Application.mainModule() {
    routing {
        root()
    }
}
// Extracted route
fun Routing.root() {
    get("/") {
        call.respondText("Hello World!")
    }
}

Inside the routing { … } block there is an implicit this: Routing, you can call the root method directly,it is effectively like calling this.root().

Deployment and application.conf

Once you want to deploy your server, you might also want to provide or to change the configuration of the serverexternally without recompiling it.

Ktor libraries expose some entrypoints that read an application.conf file from the resources, or by an externalfile. In this file you can define things like the entry point of the application, the port used, the ssl configurationor arbitrary configurations.

You can read more about using application.conf in the configuration page.

Health checks

Depending on your application, you might want to use different ways to create a health check.The easiest way would be to enable an endpoint like /health_check that returnssomething like HTTP 200 OK, while optionally verifying your dependant services.It’s completely up to you.

You can also use the StatusPages feature to handle exceptions.

install(StatusPages){
    exception<Throwable> { cause ->
        call.respond(HttpStatusCode.InternalServerError)
    }
}
routing {
    get("/health_check") {
        // Check databases/other services.
        call.respondText("OK")
    }
}