LiveKit Community

At a glance, it needs fixing. Am I missing something?

<@U01KZ61GV5Z> Thank you for reviewing my PR!!

of course! thanks for helping to fix the issue :raised_hands:

hey David, you can use one of the server SDKs (there's a JS/TS one) to pull information about rooms. You can put that behind an endpoint which your client hits to show the info.

<https://github.com/livekit/server-sdk-js|https://github.com/livekit/server-sdk-js>

we haven't tested that configuration before.. you can test it pretty easily with our <https://github.com/livekit/livekit-cli#livekit-load-tester|load-tester>

source code for generate is here
<https://github.com/livekit/deploy/tree/main/generate|https://github.com/livekit/deploy/tree/main/generate>

We are. Recording service is evolving to become Egress service. <https://github.com/livekit/protocol/blob/main/livekit_egress.proto>

to record a single participant, you would want something like a `TrackCompositeEgress`

Wow, I see. BTW, the way you guys start from protobuf to actual implementation is a pleasant to watch. It would be a great read for me. Thanks!

So in this case grab some audio video and transmux to into rtmp is possible?

it'll need to be a transcode, at least for audio. rtmp doesn't support opus audio AFAIK

<@U01KZ61GV5Z> how can i intercept the RTC and send to Wowza for example? using gstreamer ?

G'day everyone! Thanks for keeping up the amazing work! :clap: :fireworks:

<@U028CR98GN4> we really appreciate you saying this!

This community and the core team deserve the credit – LiveKit continues to improve because of you all.

I'll defer to <@U01KZ61GV5Z> on areas where we could use some help. Thank you for your passion and taking initiative to contribute. :pray::skin-tone-4: 

Hope you have a nice weekend!

Thanks, <@U01KZ0SCDKL>. and Hello <@U01KZ61GV5Z> :grinning::wave:

Really appreciate you offering to help! We'd definitely welcome it.

What stack are you most comfortable with? Go? JS? or mobile? We could use help all across the board here :slightly_smiling_face:

Hey guys I know this thread is a little bit old but me too would love to contribute. I have about 4 years experience with golang but can help with anything devops related. Such as the helm charts, deployment and benchmarking.

Let me know how can can help. I would be happy to dedicate 2- 3 hours / week for the project.

sometimes it could print error for messages that it doesn't quite know how to handle.

Ok.
i  restarted it then the message is gone

which version of kubernetes is your cluster running? there's this PR that removed the <https://github.com/livekit/livekit-helm/pull/19|deprecated v1beta1> syntax in favor of `<http://networking.k8s.io/v1|networking.k8s.io/v1>`

I upgraded to 1.22.6-gke.300 a few days ago. I’m not sure what it was when I originally deployed livekit-server

Sorry, I’m new to kubernetes. The 1.22.6 is my cluster version. When I look at the details of my livekit-server deployment I see <http://app.kubernetes.io/version|app.kubernetes.io/version>: v0.15.2. I’m not sure which one you’re referring to.

ah ok, which version of livekit are you upgrading to? the latest version should have addressed this.

I didn’t specify a version number so my understanding is that it would be whatever was the latest. I think it was 0.15.4

I’m using helm v3. I wonder if this line is the issue: <https://github.com/livekit/livekit-helm/blob/f0c5a5d0574300c8935a762ce9d43f99a3d0dcb7/livekit-server/templates/ingress.yaml#L28>

In helm v3 it should be .Capabilities.KubeVersion.Version according to docs: <https://helm.sh/docs/chart_template_guide/builtin_objects/>

<@U02T8UUS3J8> did you <https://docs.livekit.io/deploy/kubernetes#install--upgrade|update> the helm repo? It seems that you are using an older version. As mentioned, we no longer use `v1beta1` designation

image.png

can you run `helm template livekit-server livekit/livekit-server --namespace livekit --values livekit-server-values.yaml`

this is what I see:
```kind: Ingress
metadata:
  name: lk-test-livekit-server
  labels:
    <http://helm.sh/chart|helm.sh/chart>: livekit-server-0.15.5
    <http://app.kubernetes.io/name|app.kubernetes.io/name>: livekit-server
    <http://app.kubernetes.io/instance|app.kubernetes.io/instance>: lk-test
    <http://app.kubernetes.io/version|app.kubernetes.io/version>: "v0.15.5"
    <http://app.kubernetes.io/managed-by|app.kubernetes.io/managed-by>: Helm
apiVersion: <http://networking.k8s.io/v1|networking.k8s.io/v1>```

```kind: Ingress
metadata:
  name: livekit-server
  labels:
    <http://helm.sh/chart|helm.sh/chart>: livekit-server-0.15.5
    <http://app.kubernetes.io/name|app.kubernetes.io/name>: livekit-server
    <http://app.kubernetes.io/instance|app.kubernetes.io/instance>: livekit-server
    <http://app.kubernetes.io/version|app.kubernetes.io/version>: "v0.15.5"
    <http://app.kubernetes.io/managed-by|app.kubernetes.io/managed-by>: Helm
apiVersion: <http://networking.k8s.io/v1|networking.k8s.io/v1>```


right.. so I'm unsure why you are seeing errors.. perhaps start with a fresh install instead of upgrading

I was able to upgrade successfully after following these instructions <https://helm.sh/docs/topics/kubernetes_apis/#updating-api-versions-of-a-release-manifest>

hi <@U031GJTM8F2>, can you try using a volume mount instead of `cat` to pass in your config? I have a feeling it has something to do with windows line endings

<https://livekit.io/connection-test> is a good way to see if it’s working

<@U01KZ61GV5Z> when i did  i got `Host <https://livekit.city-hull.com> not found: 3(NXDOMAIN)`

Screenshot 2022-03-16 at 16.57.34.png

<@U01UB0LHTTQ>  it seems like the test pass

Hey alex,
1. you can still call the CreateRoom API and it'll update the max participant count
2. currently recorder is counted as part of the participants.. but I think excluding it is actually the right behavior.. (all hidden participants should be excluded)

this might work for you <https://github.com/livekit/server-sdk-go>

as an example, its used in our load tester to publish and subscribe to media tracks: <https://github.com/livekit/livekit-cli/blob/main/pkg/loadtester/loadtester.go#L96>

Yes! as mk says, our Go SDK (and server as well) is built on top of Pion so it's fully compatible with all of the Pion examples. You'd just feed the samples into our SDK and it'll take care of all of the signaling/etc for you.

Thanks for your responses. I was a bit shaky about this but i should have known. Thus i shall try to combine GoSDK and livekit. I will post my experiences. Thanks again :slightly_smiling_face: Full steam ahead now

I just realized that I haven't told you guys you're killing it in almost a week. Thanks for the great work!

You could change the code to use `ws` or you could deploy the LiveKit server using the `init_script.sh` which comes bundled with the caddy tls reverse proxy.
<https://docs.livekit.io/deploy/vm#deploy-to-a-vm>

really appreciated your explanation, I was so close to setup nginx as reverse-proxy

I will like to say thanks to the team   and every for there support,
I finally connecting my local development to my  livekit server  running at Digital ocean droplet

do you know why it's having issues connecting to zerossl?

is it blocked by the firewall by any chance?

Yes, I'm sorry to say, it is blocked by firewall, I have to proxy everything usually, like `http_proxy=<socks5://127.0.0.1:1089> https_proxy=<socks5://127.0.0.1:1089> sudo bash -vx ./init_script.sh` `http_proxy=<socks5://127.0.0.1:1089> https_proxy=<socks5://127.0.0.1:1089> sudo apt-get install snapd` `http_proxy=<socks5://127.0.0.1:1089> https_proxy=<socks5://127.0.0.1:1089> go build` `http_proxy=<socks5://127.0.0.1:1089> https_proxy=<socks5://127.0.0.1:1089> git clone <https://github.com/livekit/deploy.git>`, basically, proxy is my daily life

:wave: rajesh

You can have as many simultaneous rooms as you'd like, it scales quite well to that. 40-50 participants in each room seems pretty straight forward to handle. You can push that number much higher if you are using adaptivestream and dynacast. (they are optimizations that ensures clients are only fetching/sending what's needed)

Which sdk are you using? The recorder participant should be invisible

<https://example.livekit.io/#/> I joined the video chat here

livekit-server log shows `livekit	rtc/room.go:229	new participant joined	{"room": "room2022", "roomID": "RM_3cRhQ6SSQf2G", "pID": "PA_pDurrh4qdJHG", "participant": "RR_uUCp6fgNp6v5", "protocol": 4, "options": {"AutoSubscribe":true}}`  I think `options`  missed something

livekit-recorder log shows
```json
{
  "exp": 1648190912,
  "iss": "APISMgsXoajDrTf",
  "jti": "RR_BLoV6Fk9mzB2",
  "nbf": 1648104512,
  "sub": "RR_BLoV6Fk9mzB2",
  "video": {
    "canPublish": false,
    "canPublishData": false,
    "canSubscribe": true,
    "hidden": true,
    "recorder": true,
    "room": "room2022",
    "roomJoin": true
  }
}
```

add log at <https://github.com/livekit/livekit-server/blob/f293de054df603af5ab3c1ff7c0eda18bf73dbab/pkg/routing/redisrouter.go#L260> `{"Identity":"RR_FJF8rBJi2bGZ","Name":"","Metadata":"","Reconnect":false,"Permission":{"can_subscribe":true},"AutoSubscribe":true,"Hidden":true,"Recorder":true,"Client":{"sdk":1,"version":"0.13.4","protocol":4,"os":"Linux","device_model":"Other","browser":"Chrome","browser_version":"98.0.4758","address":"127.0.0.1:33622"},"Grants":{"video":{"roomJoin":true,"room":"room2022","canPublish":false,"canSubscribe":true,"canPublishData":false,"hidden":true,"recorder":true}}}`  looks good here

if alice is hidden, the other players should not subscribe alice, where is this logic, <https://github.com/livekit/livekit-server/blob/61655c6ff05625ecd86557711cec38a4a9f7167d/pkg/rtc/participant.go#L243> , I can not tell difference between visible ones and invisible ones.

it will trigger sending offer sdp regardless self is visible or not

mediasoup works well but it's more bare bones. You're really on your own when it comes to scaling/infrastructure bit whereas livekit has those things covered with helm and good documentation. IIRC mediasoup also leaves all the messaging up to you so you'll need to build and support some sort of websocket thing.

there's no mediasoup "react" library, but there is a mediasoup react example.

:wave: Shaoib,

You are asking some good questions, but as Jonathan mentioned, I think livekit and mediasoup have different focuses.

LiveKit is focused on being the best end to end WebRTC stack - from dev experience, client/server SDKs, to deployment and running in production. Things like running a high available multi-node setup are built in.

Most of the WebRTC features (like simulcast) are supported by both libraries. We've layered a few things on top of it (like dynacast and adaptive stream) that makes it much easier to leverage these advanced WebRTC capabilities.

In terms of performance, we strive to be the among the best. You can use the <https://github.com/livekit/livekit-cli#livekit-load-tester|load-tester> to get a sense of real-world performance. It's difficult to quote exact figures because multi-user experiences are very much dependent on your use case (i.e. publishing bitrate, how many streams would each user pull in at the same time, etc)

<@U01KZ61GV5Z> <@U02BD5MV5NX> Thank you guys for the information. Really appreciate it.

hey Alex, yup, they are <https://github.com/livekit/client-sdk-js/blob/e653a9d40673a58013d9284764277ecf978f2475/src/room/track/defaults.ts|set to true> by default. we need to be more explicit about this in the docs. thanks for bringing this up!

Hey alex, can you try upgrading to version 0.17.x, there's been many improvements to the client SDKs since then.

Bitrate should be respected.. you can try it with the sample app we've included. It will print out the current send bitrate.

another thing is.. in general Chrome will obey the max bitrate. but if it doesn't need the bits to encode current frames. it'll use a lower bitrate than set

Hi <@U01KZ61GV5Z> - I've updated to 0.17.4 but am still seeing the same behaviour re bitrates not being respected. It looks like framerates aren't respected either but that might be down to actual camera / webcam device specs.

I'm attaching screengrabs here of chrome webrtc stats for video presets h180, h216 and h1080.

• h180 target bitrate = 120kbps, actual bitrate = 600kbps
• h216 target bitrate = 180kbps, actual bitrate = 1 - 1.5mbps
• h1080 target bitrate = 3mbps, actual bitrate = 2.5mbps

Tested on 2x Macbook Pros (with Apple Silicon) and an Intel Mac.

Happy to do more in-depth testing and report back if useful to you guys, let me know?

We've previously only been able to limit or 'force' bitrates by modifying the SDP ("b=AS") - not with LiveKit - and at the time we did it only Chrome and Safari respected bitrate limits higher than 2.5mbps, Firefox would max out at 2.5mbps. As an aside, we also used SDP modification to force browsers to publish and playback stereo audio in webrtc (rather than mono).

Is this reproducible with the sample js app? there you can see the real-time bitrate information displayed on top of the video tile

Ah yes - you did mention that. I've just checked the sample js app and the lower bitrate limitations _do_ work there, so I must be doing something wrong in my implementation. Will fix. Sorry for the false report.

Out of interest, I tried the higher presets too, but the 'unable to publish _higher_ than 2,500kbps' behaviour seems to persist. I used a FaceTime camera as well as a virtual camera generated using OBS (at 1080p and 2160p source resolution with high-motion content).. might be that Chrome reckoned it didn't need to encode more than 2,500kbps but I'm not sure. Previously when I've set a bitrate of 8,000kbps in 1080p (again, not with LiveKit), Chrome has maxed out at about 7,000kbps encoding with high-motion content. This is not relevant to our current implementation, but thought I'd mention it in case useful info for anyone else.

Thanks for your pointers, David.

You can follow: <https://docs.livekit.io/deploy/vm>

hmm I'm not able to reproduce this.. which version of the server are you running? I'll double check

hi, livekit-server is `61655c6ff05625ecd86557711cec38a4a9f7167d`  client is <https://example.livekit.io/> recorder is `docker run livekit/livekit-recorder` , I can provide my env if needed

Hi, I am looking at your code responsible for detecting whether RTP packet contains h264 keyframe and I am wondering why you sometimes check if nalu_type == 5 and sometimes if nalu_type == 7 <https://github.com/jech/galene/blob/6fbdf0eab2c9640e673d9f9ec0331da24cbf2c4c/codecs/codecs.go#L129>

As far as I know type == 7 means SPS (config for decoder) while type == 5 slice of IDR. I would expect to always check if type == 5. However, I am testing my code and both solutions work properly i.e. there is no video freeze between encoding switch in simulcast :thinking_face:

I use the data channel with protobuf to send different data types.