LiveKit Community

are there initialization steps that I might be missinng

does the EG_xxxx participant disappear after a while or does it stay there indefinitely?
The egress participant is a “hidden” participant by default, so it should never show up inside the room.
Can you tell me which versions of the livekit packages you’re using? (everything from clients, egress, server)

<@U02EU2ZPTQU> thanks for getting back. It says until the end of the call. Some clues : when it happens, it appears at the moment an ingress (ogg/ivf via livekit-cli) leaves. livekit/egress:v1.5.1, livekit/livekit-server:v1.3.1, livekit/livekit-cli:v1.2, “@livekit/react-components”: “^1.0.3", “livekit-client”: “^1.5.0", “livekit-server-sdk”: “^1.1.0". Happy to open a ticket if needed.

cc <@U02CM28PB4Y>
I think the client shouldn’t even be aware of hidden participants, right?

Yes they should not be aware. Looks like a bug on server side.

Can you please file an issue at <https://github.com/livekit/livekit/issues> with repro steps? We will have a look. :bow:

<@U02CM28PB4Y> <https://github.com/livekit/livekit/issues/1178>

Hi Everyone!!
My application is not audio-only. I want to export a mixed audio file containing audio from all participants in the room.
I tried to stream only audio to rtmp using egress RoomCompositeEgressRequest as below, but it doesn’t work. Could i know how?
```request := &amp;livekit.RoomCompositeEgressRequest{
		RoomName:  roomName,
		Layout:    "speaker-dark",
		AudioOnly: true,
		Output: &amp;livekit.RoomCompositeEgressRequest_Stream{
			Stream: &amp;livekit.StreamOutput{
				Protocol: livekit.StreamProtocol_RTMP,
				Urls: []string{
					streamUrl,
				},
			},
		},
	}```

Hey, I am running a similar setup successfully. It might not be the best setup for performance but it works. Here are the <https://github.com/livekit/livekit-helm/pull/53|changes> required, pending a review and hopefully a merge.

Thanks for the reply. What I fail to understand is how to correctly exchange IP/Connection info for clients if nodes can't open ports as they like. Do you tunnel everything through nginx/LB?

<@U04BW8J8C64> correct: <https://kubernetes.github.io/ingress-nginx/user-guide/exposing-tcp-udp-services/>

Would you happen to have the Service/Ingress config you use available? I'm aware of this nginx functionality but wonder how to correctly implement it with livekit

with the PR, this should be enough for the service, you only need to add the exposing to your ingress-nginx config

then something like

```apiVersion: <http://networking.k8s.io/v1|networking.k8s.io/v1>
kind: Ingress
metadata:
  name: livekit-server
  namespace: ${module.config.values.namespace}
  annotations:
    <http://cert-manager.io/cluster-issuer|cert-manager.io/cluster-issuer>: letsencrypt
    <http://nginx.ingress.kubernetes.io/proxy-read-timeout|nginx.ingress.kubernetes.io/proxy-read-timeout>: 3600
    <http://nginx.ingress.kubernetes.io/proxy-send-timeout|nginx.ingress.kubernetes.io/proxy-send-timeout>: 3600
spec:
  ingressClassName: nginx
  tls:
    - hosts:
        - ${module.config.values.domain_live}
      secretName: live-tls
  rules:
    - host: ${module.config.values.domain_live}
      http:
        paths:
          - path: /
            pathType: Prefix
            backend:
              service:
                name: livekit-livekit-server
                port:
                  number: 80```

Thanks for the info, I'll see if I can get it done

Hi <@U04BW8J8C64> we have been playing around with STUNner and Livekit to get rid of the "direct reachability" problem in k8s, and we used LB service and nodePort. However, we took some steps to test it with metalLB (paused right now), feel free to open an issue here if you feel like it could be useful feature for you:
<https://github.com/l7mp/stunner/tree/main/examples/livekit>

Hey Ire, have you taken a look at this?
<https://docs.livekit.io/oss/benchmark/|https://docs.livekit.io/oss/benchmark/>

you would use it like this:

```lksdk.ConnectToRoom(url, token, roomCB, lksdk.WithAutoSubscribe(false))```

Thanks for your answer.  I am still hitting a wall trying to receive  something via onDataReceived. Both variants, ConnectToRoom and ConnectToRoomWithToken have the same issue.

The minimal code seems to be:

```room, err := lksdk.ConnectToRoom(url, lksdk.ConnectInfo{
   APIKey:              "A....DYA",
   APISecret:           "....So0yCfxc",
   RoomName:            "TestRoom001",
   ParticipantIdentity: "Robot007",
}, &amp;lksdk.RoomCallback{
   ParticipantCallback: lksdk.ParticipantCallback{
      OnTrackSubscribed: onTrackSubscribed,
      OnDataReceived:    onDataReceived,
   },
})```
Now the onDatarReceived method should be called when another participant sends something via data channel. For testing i left out the target info (empty array) and thus all participants should receive the data event. Browser based client does, but not the client using the code above. Wierdly enough, if i send something, basically with

```err = room.LocalParticipant.PublishData(payload,
   livekit.DataPacket_RELIABLE,
   []string{})```
The same go code running on the server side is receiving the data. Also the web client.

Thus 2 out of 3 participants are receiving but one participant stays in the dark. the two go implementations behave differently.



``````

I used the file transfer example, added my onDataReceived callback and it worked. it is clearly an issue in my code.

The simple code to test data channel is:

```func createAgent(roomName string, callback *RoomCallback, name string) (*Room, error) {
	room, err := ConnectToRoom(host, ConnectInfo{
		APIKey:              apiKey,
		APISecret:           apiSecret,
		RoomName:            roomName,
		ParticipantIdentity: name,
	}, callback)
	if err != nil {
		return nil, err
	}
	return room, nil
}

func TestDatachannel(t *testing.T) {
	pub, err := createAgent(t.Name(), nil, "publisher")
	require.NoError(t, err)

	var dataLock sync.Mutex
	var receivedData string

	subCB := &amp;RoomCallback{
		ParticipantCallback: ParticipantCallback{
			OnDataReceived: func(data []byte, rp *RemoteParticipant) {
				dataLock.Lock()
				receivedData = string(data)
				dataLock.Unlock()
			},
		},
	}
	sub, err := createAgent(t.Name(), subCB, "subscriber")
	require.NoError(t, err)

	pub.LocalParticipant.PublishData([]byte("test"), livekit.DataPacket_RELIABLE, nil)

	require.Eventually(t, func() bool {
		dataLock.Lock()
		defer dataLock.Unlock()
		return receivedData == "test"
	}, 5*time.Second, 100*time.Millisecond)

	pub.Disconnect()
	sub.Disconnect()
}```

Hi, currently candidate logs and rtc stats details are not json. Is there any way to  do that?

hey <@U03FECB3Q7N>,You can try to modify the code like this `activeVideoTrack != null &amp;&amp; !activeVideoTrack.muted ?  …`

that’s a bit out of my expertise, but I’m not seeing that error in our codebase

here we go, I think its coming from pion/TURN: <https://github.com/pion/turn/blob/40ee5ed2f3ebf77582e41336b90be73cb90ee283/internal/server/server.go#L74>

<https://github.com/pion/turn/blob/74d6899337250c0165abddc9cc797fe301e75da2/server.go#L184>

which gets called here <https://github.com/livekit/livekit/blob/master/pkg/service/turn.go#L128>

are multiple participants trying to connect with the same participant name?

Hi, multiple participants have different names in our app. I reproduced the issue load testing with 500x livekit-cli instances tasks: total=500x rooms and 5/5 pubs/subs per room with ~10m ramp up. It seems there's some overlaping in dynamic port allocation, Does this makes sense?

Interesting… I’ve never seen this issue before. But its possible we just haven’t tested the TURN component as thoroughly. We’ll have to look into this some more. (cc <@U01KZ61GV5Z> in case you have a better sense for why this is happening)

<@U03KYURETMF> Is this actually causing any client failures for you? It's possible this is happening only due to the frequent connection/disconnections between load tester clients

Hi, currently no customer impact. We are preparing for football matches that induce high demand in short time. The issue is hapening in peak load testing. I see new rooms beeing allocated to already congested nodes and that is something we seak to avoid. That may be a reason to see this error.

<@U03KYURETMF> you may want to set up <https://github.com/livekit/livekit/blob/master/config-sample.yaml#L197|node selector limits>

Hi Kornel, I think these days we use the `livekit-cli load-test` command to do load tests that simulate subcribers. Check it out here: <https://github.com/livekit/livekit-cli>

what react-native-webrtc version do you have in your yarn.lock file?

can you provide the console logs from the error?

<https://reactnative.dev/docs/debugging#accessing-console-logs>

LOG  rn-webrtc:pc:DEBUG 0 ctor +0ms
 LOG  rn-webrtc:pc:DEBUG 1 ctor +9ms
 LOG  rn-webrtc:pc:DEBUG 1 setRemoteDescription +74ms
 LOG  rn-webrtc:pc:DEBUG 1 setRemoteDescription OK +16ms
 LOG  rn-webrtc:pc:DEBUG 1 addIceCandidate +1ms
 LOG  rn-webrtc:pc:DEBUG 1 addIceCandidate +1ms
 LOG  rn-webrtc:pc:DEBUG 1 addIceCandidate +3ms
 LOG  rn-webrtc:pc:DEBUG 1 addIceCandidate +0ms
 LOG  rn-webrtc:pc:DEBUG 1 createAnswer +2ms
 LOG  rn-webrtc:pc:DEBUG 1 createAnswer OK +11ms
 LOG  rn-webrtc:pc:DEBUG 1 setLocalDescription +8ms
 LOG  rn-webrtc:pc:DEBUG 1 setLocalDescription OK +9ms
 LOG  rn-webrtc:pc:DEBUG 1 setRemoteDescription +104ms
 LOG  rn-webrtc:pc:DEBUG 1 setRemoteDescription OK +48ms
 LOG  rn-webrtc:pc:DEBUG 1 createAnswer +4ms
 LOG  rn-webrtc:pc:DEBUG 1 ontrack +7ms
 ERROR  TypeError: undefined is not a function, js engine: hermes
 LOG  rn-webrtc:pc:DEBUG 1 ontrack +18ms
 ERROR  TypeError: undefined is not a function, js engine: hermes
 LOG  rn-webrtc:pc:DEBUG 1 createAnswer OK +6ms
 LOG  rn-webrtc:pc:DEBUG 1 setLocalDescription +4ms
 LOG  rn-webrtc:pc:DEBUG 1 setLocalDescription OK +107ms
 LOG  rn-webrtc:pc:DEBUG 1 getStats +354ms
 LOG  connected to  <wss://whalesnet.livekit.cloud|wss://whalesnet.livekit.cloud>

it happens as soon as someone starts there stream be it audio or video

hmm, does this happen if you run the example from the repo itself?

no it doesn't, but it's weird cause I've tried exactly the same code also a code with very minimum functionalities both gave same error

if it's not too much trouble, could you upload the minimal example repo project that repros the error?

hey, can you check if the react-native-webrtc is actually resolved to 106.0.0-beta-5? the code in your screenshot looks like beta-6.

I've tried both beta-5 and beta-6... and regarding the code if you don't mind I'll send it in 2-3 hrs.

<@U04BR8DR6CT> following up on this, do you still have issues with this?

oops! I'm really sorry, I forgot to inform I fixed it by making and a clean build and updating versions, thank you for your responses <@U01NZKTNE1H>

IMG-20221120-WA0004.jpg

cpuload is probably the better metric to use. sysload is slower reacting.

what are you setting the threshold to? how quickly are these rooms being spun up?

node_selector:
    kind: cpuload
    sort_by: cpuload
    cpu_load_limit: 0.5

I've some initial pods in eks and some placeholders to speed up new pods when necessary. When I start stress testing around 30% avg cpu, the hpa launches more livekit pods. I was expecting to see the initial pods(nodes) going up only to 50-60% due to nodeselector but they go all way up to 100% cpu!

how many users are joining per second? and how is your HPA set up?

If the pods are spinning up slower than demand, you'd need to spin up more capacity ahead of the time. there are probably more tuning necessary in the node selector stack too. Can you log the CPU load over time (i.e. one log entry every 2 seconds) ? It'd be good to see how they are responding to the traffic increases

Hi, please take a look in the pic attached. It shows the initial and subsequent created nodes. The initial nodes keep receiving new rooms and respective participants all the way even when the cpu is above the defined limit.

this is showing me number of rooms per node, without information about
• CPU load on these nodes
• amount of traffic each room is doing
If rooms are allocated, and then they keep getting larger, that would also push the CPU utilization above the limit previously set.

Hi, The number is room allocated per node at that minute in the initial ~10m(Numbers are not cumulative). After the room is created it handles 5 subs/pubs for 25m, then all participants disconnect. I'l work on getting the cpu load and bandwith data integrated in logs for better insight.

<@U04BRS9FHB5> This is a common error in the flutter plugin, you can try the following methods to fix it

```flutter clean &amp;&amp; flutter pub get
cd ios &amp;&amp; rm -rf Pods Podfile.lock
pod repo update &amp;&amp; pod install```
try to recompile

Restarted simulator, fixed the issue. I see lots of issues with connecting

Not getting any helpful errors from the `Room.connect()` method.
```flutter: Could not connect Failed to connect```
Fresh token created with CLI, using the URL provided from LiveKit cloud
```wss://{my_app}.livekit.cloud```
Using snippets of code from the flutter client sdk example.

Can you connect successfully with the example app?
Maybe you need to print out some flutter console logs to help determine the issue

What other information could I be logging? It’s logging the exception currently

Using `flutter run` on the terminal will output flutter/dart logs.

okay, back to the first step, Can you connect successfully with the client-sdk-flutter/example?

I can download and run your code to confirm. But I copied and pasted the room and connect screen to my POC.

&amp;auto_subscribe=1&amp;adaptive_stream=1&amp;protocol=8&amp;sdk=flutter&amp;version=1.0.0&amp;network=wifi&amp;os=iOS&amp;os_version=15.5&amp;device_model=iPhone14%2C6#' was not upgraded to websocket

So, using the example app can’t connect correctly? :thinking_face:

I can create another token, but doesn’t look like the issue

Then we should check whether the token has expired, Before this, we encountered a similar issue

There should be a js playground in the cloud, you can verify whether the token is valid.

Tried it all again with a new token and it’s working now. Thank you for the patience

I got the logs from running the application in debug mode

all participants must be in the same room (channel) to be visible to each other. since the livekit-server should not have a preset default room, I think `room` is necessary.

<@U04BRS9FHB5> can you thread your messages pls? :pray:

livekit-react-native or just the web js client?

cc: <@U02EU2ZPTQU>, probably capping scaleResolutionDownBy to 1 is in order?

<@U0444GJ7CLC> would you happen to know the width/height of the screenshare source?

thanks for the mention <@U01NZKTNE1H>

<@U0444GJ7CLC> can you post your room options settings here?

```connectOptions={{
        autoSubscribe: false,
      }}
      roomOptions={{
        adaptiveStream: false,
        dynacast: true,
        publishDefaults: {
          videoEncoding: {
            maxBitrate: 1_500_000,
            maxFramerate: 20,
          },
          screenShareEncoding: {
            maxBitrate: 6_000_000,
            maxFramerate: 30,
          },
          audioBitrate: 20_000,
          dtx: true,
          simulcast: true,
          videoSimulcastLayers: [VideoPresets43.h480, VideoPresets43.h240],
          screenShareSimulcastLayers: [ScreenSharePresets.h1080fps30, ScreenSharePresets.h720fps15]
        },
      }}```

thanks! it’s possible for this error to get triggered if the original resolution is smaller than the one specified in the preset. we have a fix for it that will be released with the next client-sdk-js release! <https://github.com/livekit/client-sdk-js/pull/513>

What do you mean by “upload thumbnails”? Still frames to use as a preview image?

`startedAt` and `endedAt` should be nanoseconds since the Unix epoch (<https://pkg.go.dev/time#Time.UnixNano|our backend is in go>)

`duration` is also nano-seconds (it just subtracts `endedAt - startedAt`)

got it

Then the first 10 digits are the unix epoch in seconds:

<http://Time.at|Time.at>(startedAt[0..9].to_i)

And to get the time in seconds:

duration / 1_000_000_000

Thanks!

right, you can also divide it out when constructing the time like `<http://Time.at|Time.at>(<http://startedAt.to|startedAt.to>_i / 1_000_000_000)`

as for a preview image thumbnail, I don’t believe we generate one, but I’ll let <@U01UB0LHTTQ> answer that question

we don’t have any plans for uploading thumbnails

Hi, it should be 1.0.7.
<https://github.com/CocoaPods/Specs/blob/master/Specs/d/f/f/LiveKitClient/1.0.7/LiveKitClient.podspec.json>

We tried again and we have the latest version now. Thanks <@U02MAJLGN65>

Hi Shaan! If it’s actually critical for you to avoid packet loss, you can `DataPacketKind.RELIABLE` to cause LiveKit to resend packets until they are received.

Probably, though, you’ll want your game engine to just deal with packet loss.  I’m not a real gamedev :sweat_smile: but you can find an intro to the basic strategies here: <https://arstechnica.com/gaming/2019/10/explaining-how-fighting-games-use-delay-based-and-rollback-netcode/>

Oh also, if you implement a DataChannel transport for LiveKit, you can use the Unity netcode for free: <https://github.com/Unity-Technologies/com.unity.netcode.gameobjects>

(we plan to include that in our SDK in the future)

Thanks <@U02MC5V3C2H> for the detailed overview! We wanted to confirm that there was something native in livekit we were not missing.
adding this to SDK would be good addition to get on par with other solutions like photon and mirror :slightly_smiling_face:

It wouldn't matter which TURN instance ends up handling the request. Any TURN server could proxy the data to the media node hosting the room

I am unable to figure out how to us `buff` in Unity

Are there examples or docs where this would be illustrated?

Maybe they were supposed to be <https://github.com/livekit/client-sdk-unity-web/tree/main/Documentation~/api|here>?

Is this how I would access it in C#?

```JSNative.GetNumber(JSNative.CallMethod(NativePtr, "getBuffer"));
    }```

Hey, this blog explains how we did our bridging interface. If you want to communicate between C# and JS, follow this <https://docs.unity3d.com/Manual/webgl-interactingwithbrowserscripting.html|documentation >:slightly_smiling_face:

Thanks Theo!
Ok makes sense, using native JS &lt;--&gt; Unity bridge