LiveKit Community

I have in my notes to add/expose a style prop from the VideoTrackRenderer component, but until then, you can use the className prop, or a blanket css rule which sets `border-radius` 

I do this here:
<https://github.com/livekit/hq/blob/23a9f1c3ef139f5b795443b529e21a60f4bba175/client/src/index.scss#L64|https://github.com/livekit/hq/blob/23a9f1c3ef139f5b795443b529e21a60f4bba175/client/src/index.scss#L64>

I haven't x-browser tested, but works in Chrome

I've never used Flutter, maybe someone else can chime in here? ¯\_(ツ)_/¯

Hi <@U02SFES0N9Y>  can i have a screen shot? What do you mean by mixed results ?

Container does nothing and ClipRRect the corners are black not transparent.

I'm not sure if this is related to the renderer being gpu accelerated but I will take a look later :slightly_smiling_face:

```child: ClipRRect(
              borderRadius: BorderRadius.circular(20),
              child: VideoTrackRenderer(p.videoTracks.first.track as VideoTrack, fit: RTCVideoViewObjectFit.RTCVideoViewObjectFitCover)));```

Also this is Flutter Web. I haven't tested other platforms yet.

<@U028SSAKLQ1> Which client and which version of server are you seeing this in?

<@U02CM28PB4Y> Client: v0.16.1 (Chrome/macOS Monterey) / Server: v0.15.3

Thank you <@U028SSAKLQ1>. Do you have `dynacast` enabled. With `dynacast`, we disable simulcast layers for which there are no subscribers. If you are sitting alone in a room, it will take effect 10 seconds after you join. Wondering if that is what is happening and a refresh/reconnect resets that 10 second of full resolution video.

`dynacast` is disabled and the situation can be observed in rooms with 2+ particpiants as well..

Got it. Would you be able to check the available bandwidth in webrtc-internals?

Unfortunately, the session has long been closed, but: Both clients have been connected via 1Gbps uplinks (ethernet/fiber) to the (very close) server (10Gbps uplink). So I‘m almost entirely certain that bandwidth was not the issue here.. 

Thank you <@U028SSAKLQ1> . Will try to repro

The thing is, we could observe almost the same behavior with the same setup just a few days apart. But both times it happened about 40-60 minutes into the call – so it‘s really not easy to reliably reproduce this.

As asked above: Are you aware of any reason why a publisher could be stuck in low quality mode and stop reevaluating its conditions? And/or is there a way to programmatically trigger this reevaluation?

Especially the second part could be interesting to debug the issue when it happens the next time.

<@U028SSAKLQ1> we are looking at an issue that's similar.. where subscriber is not getting their video upgraded after having their available bandwidth fluctuate. will keep you posted on our investigations

<@U01KZ61GV5Z> that‘s good to hear – thanks for the update and your continued support! :heart:

<@U028SSAKLQ1> are you using JS SDK by any chance?

<@U01KZ61GV5Z> yes I am. Currently on v0.16.1.

thanks for confirming. I think this should fix it: <https://github.com/livekit/client-sdk-js/pull/142>. We'll merge tomorrow and will cut a release for you after validation.

Excellent, I'll deploy to stage ASAP and try to reproduce. Thanks a bunch!

<@U01UB0LHTTQ> do you know why end_recording isn't working?

this is usually caused by <https://github.com/livekit/livekit-recorder/issues/22>, either because there is no video, or because it wasn’t able to connect to the room in the first place

The interesting thing is, I still got valid video files at the end, so it seems there was both video, and that it could connect to the room.

I'd be happy to share any config files or debug info if it would be helpful

<@U02L67PHKKN> can you try upgrading to v0.3.13? The frozen pipeline issue should be fixed

<@U01UB0LHTTQ> Thanks!  I'm cautiously optimistic, but it seems to be working now.  Recordings start immediately and I can stop them when I want as well.  I get what appear to be valid mp4 files.  I'll give it a thorough workout today and let you know my findings.

Once all clients disconnect, however, the recorded file is saved properly and the recorder pod becomes available again and can make another recording.

So it seems only the manual `end-recording` is failing.  Allowing the recording to stop when the call terminates works.  Has anyone encountered this?

FWIW, I'm running server version 0.15.3 and the recorder version is 0.3.12

Screen Shot 2022-02-07 at 7.48.31 PM.png

I’m trying with a cluster that has nodes with 8 vCPU and 32 GB RAM, and autoscaling enabled, <@U01KZ61GV5Z> sorry for the tag but i’m stuck here for 2 days, an hint about the cluster spec, the issue?

Ok, I have moved past this issue after creating a cluster with 3 noes, each has 16 vCPU and total of 48 CPUs, is this the minimum requirement?

Thank you:pray:

update: it’s all working fine now, I had to enable all ports in GCP firewall (tried the GCP config here: <https://docs.livekit.io/deploy/ports-firewall#firewall>) but that’s my homework to figure the ports out

hey <@U02UMFKPC6Q> , sorry for the late reply. I'm glad to hear you got it all working!!

<@U01KZ61GV5Z> Thank you so much, its working with a 3 nodes cluster, each node has 16 cores, I guess this might be a bit costly at GCP as a start, is there a way to lower the cluster specs?

unfortunately not.. livekit uses very little resources, but Kubernetes has their own min requirements (mostly for # of distinct nodes). You could probably use less cores for them. Typically they recommend having 3+ nodes for a kube cluster

<@U01KZ61GV5Z> Thanks again, I will play around with the config and let you know how it goes, i REALLY like how livekit is built and already invested in, chat soon :smile:

<@U01KZ61GV5Z> by changing these values, i should be able to deploy to lower specs cluster? just trying to lower my GKE cost as it seems skyrocketing now :smile:
```resources:
  limits:
    cpu: 15000m
    memory: 10240Mi
  requests:
    cpu: 14000m
    memory: 5120Mi```

yeah, definitely.. it doesn't actually need that much.. you can try ~4 cpus for now and see how it goes

ok, as an update i was able to update the values.yaml to
```replicaCount: 1

livekit:
  rtc:
    use_external_ip: true
  redis:
    address: redis-master.default.svc.cluster.local:6379
    # db: 0
    # username: root
    # password: FZCiDcJJ22
  keys:
    key: val
  turn:
    enabled: true
    domain: <http://turn.mydomain.com|turn.mydomain.com>
    tls_port: 3478
    secretName: tls
    loadBalancerType: gke

loadBalancer:
  type: gke
  tls:
    - hosts:
        - <http://live.mydomain.com|live.mydomain.com>
      secretName: tls

autoscaling:
  enabled: true
  minReplicas: 1
  maxReplicas: 5
  targetCPUUtilizationPercentage: 60

resources:
  # limits:
  #   cpu: 15000m
  #   memory: 10240Mi
  # requests:
  #   cpu: 14000m
  #   memory: 5120Mi```
i have changed the replica count to 1 and commented the resources section and it’s now running on a 3 node cluster with e2-medium VMs which is  $25 / month nodes

I might actually submit a PR to update the docs and have a newer version of the GKE values as anyone will try with the current version will have to create a big/expensive cluster unless they edit the values as above, anyways Thank you <@U01KZ61GV5Z> very much for your help

glad to hear it! the default in the sample config is using 4000m as the request. this is really up to the user, since everyone's deployment situation/use case is different.

Hi <@U01KZ61GV5Z>. Is there a minimum of nodes when deploying to k8s?

hey Juan, welcome and it's good to hear you are finding it to be useful!

currently permission updates aren't relayed to the client itself (though it probably should).

as a workaround, you can set that piece of information in the metadata field and it'll be broadcasted to other clients in the room.

Ah that’s a good one, thanks for the response :slightly_smiling_face:

hello Jelle! thank you for the feedback!

you are bringing up a great use case. and I agree the current behavior is not ideal.

With WebRTC, we are not able to unpublish existing tracks without client being involved. so that may be necessary in order to accomplish this. I'll file an issue for this and will discuss with the team on the best way to proceed:

<https://github.com/livekit/livekit-server/issues/423>

Do you have any info on what you're doing when the crash happens?

also, which livekit library version is this?

Hi <@U01NZKTNE1H>, Server version is 0.15.3. Crash happens on Android using LiveKit 0.8.0 or 0.8.1. Seems fine on some devices, but I'm seeing the crash on a Pixel 4a.

The crash seems to happen on `localParticipant.publishVideoTrack()`

We do not use the `localParticipant.createVideoTrack()` because we are not using the phone camera video, but we have our own stream coming from a different device. So we create our own LocalVideoTrack and videoSource etc. I don't get any crash when I use localParticipant?.createVideoTrack and use the android device's camera.

``` room = LiveKit.connect(
                            CGCApplication.applicationContext(),
                            url,
                            token,
                            ConnectOptions(),
                            RoomOptions(),
                            this@LiveKitManager
                        )

                        val localParticipant = room?.localParticipant

                        val useHardwareAcceleration = true

                        val peerConnectionFactory = PeerConnectionFactory.builder()
                            .setOptions(PeerConnectionFactory.Options())
                            .setVideoEncoderFactory(getVideoEncoderFactory(eglBase, useHardwareAcceleration))
                            .setVideoDecoderFactory(getVideoDecoderFactory(eglBase, useHardwareAcceleration))
                            .createPeerConnectionFactory()

                        val videoSource = peerConnectionFactory.createVideoSource(false)

                        val webRTCLocalVideoTrack = peerConnectionFactory.createVideoTrack(
                            "video0",
                            videoSource
                        )

                        val surfaceTextureHelper =
                            SurfaceTextureHelper.create("CaptureThread", eglBase.eglBaseContext)

                        videoCapturer.initialize(
                            surfaceTextureHelper,
                            CGCApplication.applicationContext(),
                            videoSource.capturerObserver
                        )

                        videoCapturer.startCapture(videoWidth, videoHeight, 30)

                        val localVideoTrackFactory = FakeLocalVideoTrackFactory()
                        val localVideoTrack = LocalVideoTrack(
                            videoCapturer,
                            videoSource,
                            "video",
                            LocalVideoTrackOptions(
                                false,
                                null,
                                CameraPosition.BACK,
                                VideoPreset169.QHD.capture
                            ),
                            webRTCLocalVideoTrack,
                            peerConnectionFactory,
                            CGCApplication.applicationContext(),
                            eglBase,
                            DefaultsManager(),
                            localVideoTrackFactory
                        )

                        localParticipant?.let {
                            it.publishVideoTrack(localVideoTrack, VideoTrackPublishOptions())
                            liveKitIsConnected = true
                        }```


livekit-abc5727e537-m97debug.aar

livekit-abc5727e537.aar

hey andrew, not able to repro on my devices. hope you can help me run a quick check here:

one theory I'm curious about is if the use of your own eglBase/PeerConnectionFactory runs into any problems when interacting with ours. I've exposed the eglBase/PeerConnectionFactory we use on the room object (accessible through `room.eglBase`/`room.peerConnectionFactory`); can you try using those in the code instead of creating your own?

I've also attached a version of the livekit sdk with an updated webRTC m97 build. Please try using that as well to see if this is fixed by that (this build also should hopefully have better debug symbols if it does still crash so we can diagnose better.)

Hi <@U01NZKTNE1H> For the livekit-abc5727e537.aar, I'm getting the following crash on the LiveKit.connect() line. I'll do more testing mid to late next week. Thanks again for your help working through this.

2022-02-11 16:38:32.298 29469-29469/com.ourcompany.ourapp.debug E/LiveKitManager: LiveKit: error in setupLiveKit(): java.lang.NoSuchMethodError: No static method checkNotNullFromProvides(Ljava/lang/Object;)Ljava/lang/Object; in class Ldagger/internal/Preconditions; or its super classes (declaration of 'dagger.internal.Preconditions' appears in /data/app/~~qJhUJgsk_RJDsic_JC5OVw==/com.ourcompany.ourapp.debug-1n88ukd_DzcJdx13cbV_hA==/base.apk!classes2.dex)
        at io.livekit.android.dagger.WebModule_OkHttpClientFactory.okHttpClient(WebModule_OkHttpClientFactory.java:33)
        at io.livekit.android.dagger.WebModule_OkHttpClientFactory.get(WebModule_OkHttpClientFactory.java:24)
        at io.livekit.android.dagger.WebModule_OkHttpClientFactory.get(WebModule_OkHttpClientFactory.java:10)
        at dagger.internal.DoubleCheck.get(DoubleCheck.java:47)
        at io.livekit.android.dagger.WebModule_WebsocketFactoryFactory.get(WebModule_WebsocketFactoryFactory.java:25)
        at io.livekit.android.dagger.WebModule_WebsocketFactoryFactory.get(WebModule_WebsocketFactoryFactory.java:11)
        at io.livekit.android.room.SignalClient_Factory.get(SignalClient_Factory.java:49)
        at io.livekit.android.room.SignalClient_Factory.get(SignalClient_Factory.java:13)
        at dagger.internal.DoubleCheck.get(DoubleCheck.java:47)
        at io.livekit.android.room.RTCEngine_Factory.get(RTCEngine_Factory.java:31)
        at io.livekit.android.room.RTCEngine_Factory.get(RTCEngine_Factory.java:9)
        at dagger.internal.DoubleCheck.get(DoubleCheck.java:47)
        at io.livekit.android.room.Room_Factory.get(Room_Factory.java:48)
        at io.livekit.android.room.Room_Factory_Impl.create(Room_Factory_Impl.java:23)
        at io.livekit.android.LiveKit$Companion.create(LiveKit.kt:41)
        at io.livekit.android.LiveKit$Companion.connect(LiveKit.kt:75)
        at io.livekit.android.LiveKit$Companion.connect$default(LiveKit.kt:66)
        at com.ourcompany.ourapp.djivideo.livekit.LiveKitManager$startLiveKitStream$1.invokeSuspend(LiveKitManager.kt:62)
        at kotlin.coroutines.jvm.internal.BaseContinuationImpl.resumeWith(ContinuationImpl.kt:33)
        at kotlinx.coroutines.DispatchedTask.run(DispatchedTask.kt:106)
        at android.os.Handler.handleCallback(Handler.java:938)
        at android.os.Handler.dispatchMessage(Handler.java:99)
        at android.os.Looper.loopOnce(Looper.java:201)
        at android.os.Looper.loop(Looper.java:288)
        at android.app.ActivityThread.main(ActivityThread.java:7839)
        at java.lang.reflect.Method.invoke(Native Method)
        at com.android.internal.os.RuntimeInit$MethodAndArgsCaller.run(RuntimeInit.java:548)
        at com.android.internal.os.ZygoteInit.main(ZygoteInit.java:1003)

ah, looks like sharing aars directly doesn't work. I've published the builds on jitpack: <https://jitpack.io/#davidliu/livekit-custom-builds>

they can be found in the branches tab, under the names `abc5727e537-SNAPSHOT` and `abc5727e537-m97debug-SNAPSHOT`

Thanks <@U01NZKTNE1H> for providing those builds. Using the room.eglContext and the room.peerConnection factory doesn't seem to make a difference in the abc5727e537-SNAPSHOT build.

But I don't get the crash at all with the abc5727e537-m97debug-SNAPSHOT build. Seems to work fine. Is that with the updated WebRTC?

thanks for checking it out, should have a new release within a couple days (or can use the current livekit public snapshot build)

I can confirm this is fixed for me on 0.9.0. Thanks!

image.png

hey <@U02SFES0N9Y>, this may be an issue with how the video renderer is configured in Flutter. Could you file a an issue for this in `client-sdk-flutter`? in it please include a layout that could reproduce it.

I resolved this. Thanks for the help <@U01KZ61GV5Z>!

```console.log(props)
console.log(props.participant)
console.log(props.participant instanceof LocalParticipant)```

Logging them shows that the types are correct :thinking_face:

instanceof should return the right results there. we use this within our React example. I suspect what's happening might be state related. You'd want to use the `useParticipant` hook to ensure state changes are triggering refreshes on your component: <https://github.com/livekit/livekit-react/blob/master/src/components/ParticipantView.tsx#L72>

But the instanceof check isn't working. Is there a different way to check the types of the participants?

I found this PR that will add an event when the room state has changed.
<https://github.com/livekit/client-sdk-js/commit/5bdb4319ded3cfeb4dd1b917e36b4a139360bb69>

Not sure if this would be a bit late.
Either way, having an event when the publisher/subscriber are created might be helpful

Hey <@U02JTUS785V>, welcome!

This is a pretty advanced use case, may I ask what you are doing with the peerconnections?

I think given the advanced usage, it's best to keep that event on `engine` instead of the room itself. feel free to submit a PR for it!

I would like to watch, or pass those connections somewhere else. Yeah, this should definitely be an engine event

:wave: Hi Sandeep, this might be helpful:
<https://docs.livekit.io/deploy/vm>

hmm, does this happen regularly if you tried to use `Camera.open(cameraId)` in a separate sample app? the crash seems to lie mostly with the underlying android os here. perhaps a permissions issue or even a samsung specific issue of some sort?

what’s weird is that around half the time it works fine and I get the camera feed

accessing the camera directly from the `Camera1Enumerator` video capturer and create the track manually work fine without crashing

hmm. could you try doing multiple Camera.open(cameraId) in a row (without releasing in between) and see what happens? maybe there's a race condition going on here.

I’m going to try setting the `autoSubscribe`  connect option to false

ah, sounds like a race condition then; what os version is the s2 tablet on?

does the multiple `camera.open` crash consistently, or intermittently like before?

sorry for the late reply, multiple `camera.open`  do consistently crash, and I just got the crash again even with the `autoSubscribe` set to false in the connect options, the android version on my tablet is 7.0

hey, there should be a fix for this in 0.9.0 now. should avoid the extra camera.open now, fixing this crash

hey, sorry for the late reply, so, I’ve tried the new version and in fact it doesn’t crashes anymore :slightly_smiling_face: but, for the call I’m using a tablet with front and back cameras and another device which only has a back camera, the `setCameraEnabled`  failed due to a crash when trying to get the enumerators since the default is to use the front camera, in order to avoid that I’m manually creating the video track for the back camera with
```val videoCaptureParams = if (sendHDStream) {
    VideoPreset169.HD.capture
} else {
    VideoPreset169.VGA.capture
}

val videoTrack = localParticipant.createVideoTrack("video", LocalVideoTrackOptions(
    isScreencast = false,
    deviceId = null,
    position = cameraPosition,
    captureParams = videoCaptureParams
))
videoTrack.startCapture()
localParticipant.publishVideoTrack(videoTrack)```
that was working fine before the update, but afterwards I got no crash but the video track from the device with just the back camera never gets published, so I assume it doesn’t gets created :disappointed:

that code is run instead of the `localParticipant.setCameraEnabled(true)`  for the device I have with a front and back cameras in it

this code runs fine on the device with the front&amp;back camera?

it works for me, but my devices have front&amp;back

for the video track not getting published, can you check the device logs and see whether the publish request is sent? Ensure `LiveKit.loggingLevel` is set to VERBOSE, and then look in logcat for something that looks like this:

```SignalClient: sending request: add_track {
      cid: "0b6359d0-e75e-4741-82db-0cb514e9706e"
      name: "video"
      type: VIDEO
      width: 640
      height: 360
      source: CAMERA
      layers {
        quality: MEDIUM
        width: 640
        height: 360
        bitrate: 400000
      }
      layers {
        width: 320
        height: 180
        bitrate: 125000
      }
    }```

the same code does work in the device with both front and back cameras but not in the one with only back

and I do see the request to add a track
```sending request: add_track {
      cid: "b8409e99-6c3a-403c-aaed-f8af27ec52de"
      name: "video"
      type: VIDEO
      width: 864
      height: 480
      source: CAMERA
      layers {
        quality: MEDIUM
        width: 864
        height: 480
        bitrate: 800000
      }
      layers {
        width: 320
        height: 177
        bitrate: 125000
      }
    }```

it just never gets published to the other side

it is being generated as I can play it in the device :thinking_face:

the other device doesn't receive any update about the track? could you post logs from the other device? curious to see if any `response: update`  messages are missing the track

hey Abbas, you can use the server binary to generate it: `livekit-server generate-keys`

using docker would be:
`docker run --rm livekit/livekit-server generate-keys`

Pat, which version of the server are you running?

take a look at docker compose logs. <https://docs.livekit.io/deploy/vm#troubleshooting|here> are the steps to getting them.

First, docker-compose did not install as part of the cloud_init file. Odd.  After installing it, and collecting the logs again I get:
ERROR: Version in "./docker-compose.yaml" is unsupported.  You might be seeing this error because you're using the wrong compose file version.  Either specify a supported version (e.g. "2.2" or "3.3") and place your service definitions under the 'services' key, or omit the 'version' key and place your service definitions at the root of the file to use version 1.

Now when I run docker-compose --version It populates that is is version 1.25.0 build unknown.  The version number in the docker compose yaml is
"3.9".  the "services:" section has nothing after it.

you would need docker compose 2+, which the included shell script should install. usually package managers contain a really old version that doesn't work with newer compose files

Odd that it did not install it at all.  Guess onto trying to update docker compose then.

So, I now have version 2.2.3.  Same error.

this is what I get:
```root@dz-2:/opt/livekit# which docker-compose
/usr/local/bin/docker-compose
root@dz-2:/opt/livekit# docker-compose --version
Docker Compose version v2.2.3
root@dz-2:/opt/livekit# docker-compose up
[+] Running 3/3
 ⠿ Container livekit-redis-1    Created                                                                                                                                                                                         0.1s
 ⠿ Container livekit-caddy-1    Created                                                                                                                                                                                         0.1s
 ⠿ Container livekit-livekit-1  Created     ```

When making the cloud file. does docker compose need to be up to date on THAT machine?  Meaning the machine creating the cloud file?

no, the machine that creates it doesn't require compose. I'd suggest double checking compose version again.. since the error you are seeing is the same when running with compose 1.x

I run "docker compose version" and get version 2.2.3.  Maybe the old didn't completely remove.  Investigating now lol.

Alright, so the old was still there.  That is fixed!  But, more issues have arisen.
Caddy errors.  When I run the curl http(mydomain) i get port 80 connection refused.  But, when I run firewall list all 80/tcp as per the guide is listed.

Local host file issue.  Resolved.  My server is up! WOOT!

previously <@U02CM28PB4Y> has solved this problem but in latest version I can see problem still exist. Main porblem if this happen after that the user can't share webcam until disconnect that user.

I don't believe that issue has been merged yet. It's dependent on <https://github.com/pion/webrtc/pull/2107|this PR> getting merged first.

Did you check your firewall settings? <https://docs.livekit.io/deploy/ports-firewall>

all of the ports that need to be exposed should be open on the firewall

Yeah, I checked that, all are open.
I have get "OK" from the url but cannot skip the test.
Error code :"Internal error"

found this <https://livekit-users.slack.com/archives/C01KVTJH6BX/p1641750448074300?thread_ts=1641749849.074200&amp;cid=C01KVTJH6BX>

trying now ..

that worked, do we need to update the docs? I will open a couple of PRs once my deployment is over

hey Ahmed, this is a good catch!

the examples should really give a full example of how to use it. if you are up for making a PR, that'd be fantastic!

<@U01KZ61GV5Z> here is a PR for the sdk repo, i will open another one for the docs <https://github.com/livekit/server-sdk-go/pull/41>

docs PR: <https://github.com/livekit/livekit-docs/pull/21>

I think this is going to happen with mute, because when you are writing, it's not using an internal clock. I'm not an expert in this, but I think you'd have to keep filling the file with blank frames with the right timestamps to indicate no data.

GStreamer should be able to handle this, if you feed this into it

:wave:Thx, David, will take I look on GStreamer and If will be succeed, inform you:slightly_smiling_face:

hi there, assuming you are using our Helm chart, you would be declaring your API keys within the helm chart. Once deployed, you will not be able to change the keys without redeploying.

I would recommend starting out with a single VM instance, as it's much simpler to get it to work end to end.

VM is tried already David. Works fine. I using helm chart but please confirm in Helm chart am i create earlier or helm chart will create any keys ?

I'm not following.. your API keys are specified in your customized `values.yaml`  like <https://github.com/livekit/livekit-helm/blob/master/server-sample.yaml#L26|here>

David sorry for my confision.
Please tell me how i create API key ( i know with docker run ) on EKS directly ?
If i run docker run ncommand on my local machine that created keys not for my EKS nodes
I couldnt exactly understand that part if you clarify for me so appreciate

you won't be able to do that.. you'll want to generate the keys locally and fill into the yaml config. the keys will be known by you anyways.. without them it's not possible to generate access tokens

Hi, is this a Flutter SDK question ? I recommend to unpublish the local tracks when no longer needed which should release the camera .

Sounds good. I'm disposing of the room and listener when the widget disposes and I'm running into what I think is an edge case here.

I'm using vrouter and if I navigate to the same page with different parameters the dispose runs but it does not release the camera.

can you make sure you are `awaiting` the dispose ? before you proceed the flow.

I am. I'm also using riverpod for state managment and I'm wondering if that's messing it up.

I will need to check the code but, can you try calling `disconnect` then `dispose` ?

I have a life cycle issue due to riverpod.

got it, riverpod is great. But it's tricky to get used to at first :slightly_smiling_face:

we print logs into stderr/stdout, so it depends on where your docker output logs are being redirected do

ok. docker logs &lt;containerid&gt; gives me all the stdout

I have set up livekit on docker with redis and I have tried with 5 people in a room with audio and video. After some time video starts to freeze for random participants. sometimes it catches up later and sometimes all videos freeze. Any suggestions what I should look at to understand what is going on?

which version of the server are you using?

I pulled livekit/livekit-server:latest around two weeks ago. How can I check what version this refers to?

it prints its version at startup, something like `starting livekit...`

so for logs, see if you can find logs like these:
`state change` from XXX to XXX
`committing channel capacity(bps)` from XXX to XXX

they will help in understanding if there's a perceived bandwidth constraint

I see a lot of state changes: "from":"STABLE","to":"DEFICIENT"

and coming from redis I see a lot of: channel is full for 1m0s (message is dropped)

are you using simulcast for your streams? Stable to deficient would indicate there is insufficient downstream bandwidth for the subscriber for all of the streams they are subscribing to. Using simulcast would be helpful to resolve issues like this.

you can also try to pllay around with <https://github.com/livekit/livekit-server/blob/master/config-sample.yaml#L52|congestion control> config.

The deployment it is working now, but I have problem now with turn and LB the GKE

Reading the helm something is strange, because the service expose TCP but the recipes on deployment only expose UDP

<https://github.com/livekit/livekit-helm/blob/master/livekit-server/templates/deployment.yaml#L65>

<https://github.com/livekit/livekit-helm/blob/master/livekit-server/templates/deployment.yaml#L71>

<https://github.com/livekit/livekit-helm/blob/master/livekit-server/templates/service.yaml#L24>

there is 2 domains with 2 different services:
• establishing session and peer connection, this uses websockets over HTTP and needs a sub domain + ssl certificate, for example: <http://live.mydomain.com|live.mydomain.com> 
• turn server: internal turn server and this uses a separate sub domain and SSL certificate, for example: <http://turn.mydomain.com|turn.mydomain.com> 

If you face problem with turn then can install trun server (coturn) separately. New livekit version support external turn server. <https://github.com/livekit/livekit-server/blob/c534099e3a3abae2b645d4b2f03faaeb156a519c/config-sample.yaml#L43>

hey <@U033FFDFN0K> I know this is pretty confusing, but the ports are correct. WebRTC utilizes both TCP and UDP ports, but we don't have to declare the UDP ones due to use of host networking.

See <https://docs.livekit.io/deploy/ports-firewall|ports docs> for more on the ports that livekit uses

Hey <@U01KZ61GV5Z> just after open and set the udp pors on kube service that turn works

Yeah I know <@U02UMFKPC6Q> I setup in different services and domains. I put a exclusive domain only for turn

I will put here the screen shot related to the information, I will add the tcp as well like the documentation 