We’ve had several crash reports mention unhandled OkHttp exceptions (

SocketTimeoutException

,

ConnectTimeoutException

) for our KMM Android app using Ktor client 2.1.2 and the OkHttp engine factory. Is it intentional that engine exceptions aren’t caught (have I misunderstood something 😅), or is this actually a Ktor bug?

I have a curious case here. I used https://start.ktor.io/ to start a new project, I leave everything default and only added two plugins • Pebble • kotlinx.serialization Now if these two blocks here

install(ContentNegotiation) {
  json()
}

install(Pebble) {
    loader(ClasspathLoader().apply {
        prefix = "templates"
    })
}

are executed in this order, going to /pebble-index gives me an exception on the backend

kotlinx.serialization.SerializationException: Serializer for class 'PebbleContent' is not found.

As soon as I change the order how those two blocks are called without changing anything else it works as expected. Is there something I’m missing or not understanding or should a file an issue for this? To reproduce you just have to download a template with those two plugins and in

Application.kt

switch the order of

configureSerialization()
        configureTemplating()

to toggle the error. It looks like the order of those calls in

Application.kt

depends on the order you added those plugins in the web ui.

Is anyone using elastic - APM to monitor a ktor server? We are able to Use the elastic apm agent but it is not generating logs for the requests received.

is it possible to listen to the progress it takes to receive a multipart file in a ktor server? I have a ktor server running on android and I would like to display a progress bar for files being uploaded to the server via POST. with some debugging i noticed that reading each part (of the multipart) takes the longest time

Hi, is is possible to create a

StateFlow<HttpClient>

which returns/caches a client based on a

Flow<HttpClientConfig>

so the client is update/a new client is created when the config is updated, but at the same time the old clients are cleaned up automatically, for example by the GC? In our use case we need to configure all our http clients dynamically during app runtime and the configuration is a Flow which can change at any time.

hello, i use a ktor client and i currently try to supply a base URL using the

DefaultRequest

plugin as described here: https://ktor.io/docs/default-request.html#configure i tried suppling the URL as complete string:

defaultRequest {
    url("<https://ktor.io/docs/>")
}

and also tried setting the URL parameters by myself:

url {
    protocol = URLProtocol.HTTPS
    host = "<http://ktor.io|ktor.io>"
    path("docs/")
}

the base URL i try to set is:

<https://api.ws.sonos.com/control/api/v1>

during debugging, the plugin shows the correct URL including all path segments. ✅ but the

GET

request is executed without the path segments. ⛔ i receive this response (path is missing):

You don't have permission to access "http://api.ws.sonos.com/households" on this server.

for request:

httpClient.get("/households")

am i doing something wrong? is this a known issue?

Hello! Is it possible to have a custom plugin act as a “middleware” that can abort the whole pipeline(the request) ? When calling “call.respond” , we see that it still continues to other Phases in the background. (why doesn’t call.respond() call “finish()” ?) There’s no access to the pipeline.finish() method when implementing the custom plugin Hook, so we can’t even call it manually ourselves. Our use-case: After the “validate { }” function, we need to check whether our user has “profile completed” status (by checking his claims on the JWT), and if not - respond with some status. But “validate” block itself only has 2 options (if principal is null -> it returns 401, otherwise it just proceeds, so no option to have a third option to it). And if I implement a custom plugin, even if I do there “call.respond(..)” it still proceeds to other plugins instead of finishing the whole pipeline I elaborated more on it here : https://stackoverflow.com/questions/74243213/how-to-stop-pipeline-in-ktors-plugins-middleware-that-sends-a-response-and-abo

I saw this issue was fixed, but there's no explanation how? https://youtrack.jetbrains.com/issue/KTOR-2501 Is there a way to get the raw client ip address now?

Hey all, is there any way to get the path provided to the

get() {...}

route from a plugin? If not, is there at least a way where I can get the parameters from a request inside a plugin?

Hello! I’ve noticed a few spikes in response time of my microservice’s APIs. Upon further debugging, I observed that when multiple concurrent requests hit my service, then a few of them gets processed with some delay. I tried increasing the

runningLimit

of my Netty engine but it didn’t help. Did anybody else ever came across this issue? ktor version: 2.1.1

I want to pass color from my backend in ktor over to android which is my client. How can i achieve it in ktor? Previously i was doing it from android using androidx graphical compose. Any guide will greatly appreciated.here is my model in ktor

Hello all. I was wondering if anyone knows about this error:

java.net.ConnectException: Failed to connect to localhost/127.0.0.1:8080

I have a ktor server and an android app running on the same multiplatform project. Google doesn't say anything about the error except for this https://stackoverflow.com/questions/62500669/ktor-android-local-http-server-throws-error. I tried the solution and got nothing

Hello, I have a problem with the Ktor client (JS). If the addressed api returns the status 302 Found for a temporary redirect and the client was configured with followRedirects = false, the status is not returned. In this case, the status is 0. Is there a special reason for this or is this a bug?

Anyone using ktor + micrometer metrics + google cloud stackdriver? I'm finding ktor metrics are costing me upwards of $30/day on a mostly idle app. The amount of data that Google says is being received for these metrics, and what the code is apparently sending for each request seems out of whack. Anyone else seen this and figured out what is happening?

Hey is there any sort of diagram out there that shows the order of execution for the various phases in which a plugin can execute? (i.e. something like CallSetup->CallReceive->.....->ResponseSent)

Is there any way to get the response body from a plugin? I know I can get the OutgoingContet, but I can't figure out how to turn that into something usable for myself

Kindly assist on how i can query a list of taskers based on an embedded item"category_skill" inside Skill in ktor

Hi, dows anyone has a sample of a ktor server with JWT with a ktor client connecting to it? The documentation on JWT server is quite ok (at least I managed to make it work). However I don't see how to connect the client using the bearer auth. Thanks,

Does anyone have any experience using apache cassandra with ktor?

I'd like to configure one of my ktor apps for AWS Lambda deployment. I'm a big fan of http4k's serverless-lambda plugin for http apis, and am looking for something similar. I understand kotless exists, but it's far more intrusive than I'm looking for. All I need is an adapter for the AWS Lambda ApiGatewayV2 interface. Does such an adapter exist outside kotless?

Hello folks... I have a more complicated application configuration question. Imagine a carmaker application, that runs on multiple environments for each carmaker. • tesla - development, staging, production • honda - development, testing, staging, production • ford - development, qa, staging, production I think you got the idea... every "profile" can have multiple envs, each one needs its own configuration values. questions: is it possible to setup a Ktor application to fit these needs? is using

-config

via command line the best approach?

java -jar sample-app.jar -config=anotherfile.conf

Hi team, I have an OAuth/authentication best practices question. In my app, I have a few pages that need to be OAuth authenticated. On each of these pages, I check for the user session first and redirect to login (“/login” is the one under

authenticate("auth-oauth-google")

):

val userSession: UserSession = call.sessions.get() ?: return@get call.respondRedirect("/login")

But as a result of successful authentication, I redirect to one of my pages:

get("/auth/callback") {
                val principal: OAuthAccessTokenResponse.OAuth2? = call.authentication.principal()
                call.sessions.set(UserSession(principal?.accessToken.toString()))
                call.respondRedirect("/status")
            }

It worked fine when I had only one page under OAuth. But for multiple, how do I redirect to the one that invoked the OAuth flow? I’d like to avoid having a “landing” page that requires OAuth - my users most probably will use links/bookmarks to go to specific pages.

has anyone provided an instance of

NSURLSession

to ktor’s

HttpClient

? something like

engine {
      val config = NSURLSessionConfiguration.defaultSessionConfiguration()
      val session = NSURLSession.sessionWithConfiguration(config)
      usePreconfiguredSession(session)
   }

It’s hard to debug the real issue but it basically does not work. The iOS app does not crash, but no network call is made.

Hello, as Heroku free plan is being terminated, what is your best alternative to host a simple Ktor REST API server? Thanks a lot

Does anyone know how to do this, please? 🙂

hey ktor team

i am writing a higher-order framework and am thinking of adopting ktor with a module to support it

but i am curious, what is the preferred DI system with ktor? is it koin? is there a good one that leverages KSP and MPP and would be the future-safe choice?

Is there a way to change these client configurations during runtime, after they have been configured?

install(Logging) {
  this.logger = get()
  this.level = ALL
}