K
Kord4mo ago
vyfor

[Kotlin/Native] Slow connection issues

i'm experiencing noticeable connection issues with Kord (branch feature/native on Linux). about 90% of interactions fail, i can't even defer them to respond later. so far i couldn't upload files too, and it takes 10 seconds for a message to be sent (and retrieved afterwards). what's strange is that it always takes 10 seconds. here's the code for the last part:
val (message, restLatency) = measureTimedValue {
ctx.message.reply("Calculating...") // this is sent instantly though
}

message.edit {
content = buildString {
append("Gateway: ")
appendLine(ctx.kord.gateway.averagePing ?: "unavailable")
append("REST: ")
append(restLatency)
}
}
val (message, restLatency) = measureTimedValue {
ctx.message.reply("Calculating...") // this is sent instantly though
}

message.edit {
content = buildString {
append("Gateway: ")
appendLine(ctx.kord.gateway.averagePing ?: "unavailable")
append("REST: ")
append(restLatency)
}
}
no such issues occur when i run on JVM. could this be related to Ktor, or maybe the curl engine itself?
No description
32 Replies
vyfor
vyforOP4mo ago
Kord won't let me use the CIO engine either: TLS sessions are not supported on Native platform.
viztea
viztea4mo ago
well its a ktor thing not kord send trace logs
vyfor
vyforOP4mo ago
though the first 2-3 requests were fine
No description
vyfor
vyforOP4mo ago
i dont even know whats going on anymore
No description
viztea
viztea4mo ago
you can try manually instrumenting the http client to rule out rate limiter problems but it’s most likely a curl problem
vyfor
vyforOP4mo ago
how would i do that?
viztea
viztea4mo ago
custom plugin or i think Http call logging is an official one
viztea
viztea4mo ago
is mine would need to remove the engine name thingy
vyfor
vyforOP4mo ago
did i just need to install it as a plugin? if so, here are the logs:
[] [DISCOVERED]:[BUCKET]:Bucket discovered for 3df15bae86f6647dd4dfcbd5c6949480
TRACE: [] [Http/Ktor: Curl] <- POST https://discord.com/api/v10/channels/1191449585458094180/messages 200 OK - 10.772144703s
DEBUG: [[R]:[KTOR]:[ExclusionRequestRateLimiter]] [RESPONSE]:200:POST:https://discord.com/api/v10/channels/1191449585458094180/messages body:{"content":"Calculating...", ...}
DEBUG: [[R]:[KTOR]:[ExclusionRequestRateLimiter]] [REQUEST]:PATCH:/channels/{channel.id}/messages/{message.id} params:[{channel.id}=1191449585458094180,{message.id}=1274360378347884575] body:{"content":"Gateway: 39.631ms\nREST: 10.773614199s"}
TRACE: [] [Http/Ktor: Curl] <- PATCH https://discord.com/api/v10/channels/1191449585458094180/messages/1274360378347884575
TRACE: [] Received raw frame: Frame BINARY (fin=true, buffer len = 91)
TRACE: [] Gateway <<< {"t":"MESSAGE_UPDATE","s":4,"op":0,"d":{"content":"y.ping", ...}
TRACE: [] [DISCOVERED]:[BUCKET]:Bucket discovered for 3df15bae86f6647dd4dfcbd5c6949480
TRACE: [] [Http/Ktor: Curl] <- PATCH https://discord.com/api/v10/channels/1191449585458094180/messages/1274360378347884575 200 OK - 10.153825131s
DEBUG: [[R]:[KTOR]:[ExclusionRequestRateLimiter]] [RESPONSE]:200:PATCH:https://discord.com/api/v10/channels/1191449585458094180/messages/1274360378347884575 body:{"content":"Gateway: 39.631ms\nREST: 10.773614199s", ...}
[] [DISCOVERED]:[BUCKET]:Bucket discovered for 3df15bae86f6647dd4dfcbd5c6949480
TRACE: [] [Http/Ktor: Curl] <- POST https://discord.com/api/v10/channels/1191449585458094180/messages 200 OK - 10.772144703s
DEBUG: [[R]:[KTOR]:[ExclusionRequestRateLimiter]] [RESPONSE]:200:POST:https://discord.com/api/v10/channels/1191449585458094180/messages body:{"content":"Calculating...", ...}
DEBUG: [[R]:[KTOR]:[ExclusionRequestRateLimiter]] [REQUEST]:PATCH:/channels/{channel.id}/messages/{message.id} params:[{channel.id}=1191449585458094180,{message.id}=1274360378347884575] body:{"content":"Gateway: 39.631ms\nREST: 10.773614199s"}
TRACE: [] [Http/Ktor: Curl] <- PATCH https://discord.com/api/v10/channels/1191449585458094180/messages/1274360378347884575
TRACE: [] Received raw frame: Frame BINARY (fin=true, buffer len = 91)
TRACE: [] Gateway <<< {"t":"MESSAGE_UPDATE","s":4,"op":0,"d":{"content":"y.ping", ...}
TRACE: [] [DISCOVERED]:[BUCKET]:Bucket discovered for 3df15bae86f6647dd4dfcbd5c6949480
TRACE: [] [Http/Ktor: Curl] <- PATCH https://discord.com/api/v10/channels/1191449585458094180/messages/1274360378347884575 200 OK - 10.153825131s
DEBUG: [[R]:[KTOR]:[ExclusionRequestRateLimiter]] [RESPONSE]:200:PATCH:https://discord.com/api/v10/channels/1191449585458094180/messages/1274360378347884575 body:{"content":"Gateway: 39.631ms\nREST: 10.773614199s", ...}
viztea
viztea4mo ago
yeah curl moment yeah this is happening to me as well
viztea
viztea4mo ago
lmao
No description
viztea
viztea4mo ago
it's weird though, some requests go through instantly like that 'Connected to message'
vyfor
vyforOP4mo ago
hmm i'll go ahead and test ktor itself
viztea
viztea4mo ago
jesus fuck kotlin/native regex parsing is SLOW
vyfor
vyforOP4mo ago
how slow are we talking lmao i didnt notice anything strange, only the first requests seem to be a little slow due to warmup that aside, my bot also does http requests here and there, and uses supabase-kt to communicate with the database with no connection issues whatsoever
viztea
viztea4mo ago
yeah it might be discord only lol 38.561442049s
vyfor
vyforOP4mo ago
goddamn
viztea
viztea2mo ago
DOT_MATCHES_ALL has shit perf bit of a necropost but @vyfor can you try using a noop request rate limiter?
object NoopRequestRateLimiter : RequestRateLimiter {
override suspend fun await(request: Request<*, *>): RequestToken = NoopRequestToken()
}

class NoopRequestToken : RequestToken {
override var completed: Boolean = false

override suspend fun complete(response: RequestResponse) {
completed = true
}
}
object NoopRequestRateLimiter : RequestRateLimiter {
override suspend fun await(request: Request<*, *>): RequestToken = NoopRequestToken()
}

class NoopRequestToken : RequestToken {
override var completed: Boolean = false

override suspend fun complete(response: RequestResponse) {
completed = true
}
}
requestHandler {
KtorRequestHandler(token, NoopRequestRateLimiter)
}
requestHandler {
KtorRequestHandler(token, NoopRequestRateLimiter)
}
this fixes the problem for me so if it does for you too then it would narrow down the problem cc @LustigerLurch i should try using the parallel rate-limiter doesn't seem to happen with parallel either
LustigerLurch
LustigerLurch2mo ago
gotta look into rest ratelimiters at some point
vyfor
vyforOP2mo ago
sorry for the very late reply, i totally missed this message. i'll give it a go rn
vyfor
vyforOP2mo ago
of course kord wont build for me
No description
viztea
viztea2mo ago
windows moment
vyfor
vyforOP2mo ago
yup that was the case
No description
vyfor
vyforOP2mo ago
wish they had setup an action to republish kordx.emoji whenever kord is republished yup that seems to be working so far
LustigerLurch
LustigerLurch2mo ago
pulling all into one repo would also solve this, been having this in the back of my head for some time
vyfor
vyforOP2mo ago
hmm in case you dont end up merging them into one, you might wanna look into repository_dispatch not sure why it's happening in the first place, having to republish every time is a pain
SchlaubiBus
SchlaubiBus2w ago
I am still against putting everything into one repo
LustigerLurch
LustigerLurch2w ago
i forgot why, so why?
SchlaubiBus
SchlaubiBus2w ago
Esentially I don't want to make another androidx The biggest problem with monorepos like that is: 1) slower build time 2) slower import time 3) Slower CI runs, since it's very complicated to only trigger the CI for the one part that changed 4) If we do a snapshot, we always need to update kordx.emoji as well, otherwise it won't compile, even if we just want to do a prototype of the actual feature first 5) Either we write a convoluted detection logic mentioned in 3) or kordx (and probabbly cache if we go that route) have identical versions to kord, which would end up in "empty" releases for thoes projects, which is exactly the problem with androidx's jetpack
Want results from more Discord servers?
Add your server