LiveKit Community

either that or create the room via server SDK and start recording immediately

which platform (web/ios/android) are you seeing this on?

iOS. it’s weird.. it’s like a normal frame loads, then it loads a very darkened version of that frame (or perhaps the next one), then gets stuck on that frame

interesting.. is this consistently reproducible? Do you see this with the included example app?

it’s reproducible in my app but i highly doubt in the example… as i made a simplified version of my code and it *did not* reproduce there

i have quite a complex UI and the video renderer is nested v deep

inside animation widgets, an Offstage, a TickerMode, BackdropFilter, stacks etc etc

i’m flummoxed, i’m sure it’s something i’m doing..  whether this should cause the TrackRenderer to fail is another matter

ah ha.. ok, try disabling `optimizeVideo` in Room Options <https://docs.livekit.io/client-sdk-flutter/livekit_client/RoomOptions-class.html>

it might be an issue with our visibility detector pausing the subscription due to it thinking that the element is invisible

this is before connecting to a room, jsut the local track

it's the same.. optimizeVideo (on by default in this version) turns on visibility and size detection.. though it shouldn't be running on LocalTracks.. this is more for remote

yes this for a LocalVideoTrack only, before room creation

:thinking_face: I'm not sure then.. but sounds like some odd interaction between your UI and something deeper in rendering. if you manage to isolate the combination that breaks it, we can take a look at how we might be able to work around it.

yep i am doing a UI monte carlo simulation as we speak

well it’s triggered by a rebuild of the FutureBuilder that is listening to the localtrack created by createCameraTrack()

Hi <@U02Q46USG4D> can you try if this happens when you *first run the app*? since hot-reload/ hot-refresh may not clear internal states of WebRTC(flutter_webrtc)

also can you try temporarily turning off(removing) `BackdropFilter` widget since that might interfere with rendering

yes i think i may have found the rather embarasssing solution h

i had some old agora code in my app i think was stealing the camera frames

it happens Mark! glad you figured it out :raised_hands:

are you running nginx on the same server on 443? if so.. TURN cannot use the same ports.

<@U01KZ61GV5Z> , but shouldn't TURN server be exposed as port 443, the docs mentions that's how it is advertised to clients.

yes.. so does the primary signal server HTTPS.. I think Kezy has nginx to terminate SSL for the signal connection on the same node.

trying to set up everything on the same instance is tricky without using some sort of SNI, I think it's do-able... but just need to test and update docs

so, TURN and SIGNAL should operate on different ip addresses? I'm asking as I had the same issue, and decided not to use TURN  for now

yeah, that's the current workaround.. or moving to terminate HTTPS for signal via an external load balancer

<@U02HN8GPUBY> what do you mean by different ip addresses? I have only one server with one ip, and two domains(one for LK and one for turn) that points to that one ip.

<@U01KZ61GV5Z> how can I terminate HTTPS for signal via an external load balancer ?

I would recommend not worrying about TURNS for now, and focus on getting your application right. Only a tiny % of traffic requires TURNS.

<@U02SUCZ6TTN> I mean having two public ip addresses on the same server, so you can bind port 443 for TURN on one IP, and bind 443 for Signal on another IP. I decided not to use TURN for now, until I scale up and have more than 1 server

<@U01KZ61GV5Z> so what is the fastest way to set up turn server on one node? Will you update the docs? It is very urgent for us! Thank you in advance!

<@U01KZ61GV5Z> I have set up the SNI with nginx:
```server {
    listen       443 ssl http2;
    listen       [::]:443 ssl http2;
    server_name <http://turn.com|turn.com>;
    ssl_certificate "/fullchain.pem";
    ssl_certificate_key "/privkey.pem";

    location / {
      proxy_pass <http://127.0.0.1:444>;
      proxy_http_version 1.1;
      proxy_set_header Upgrade $http_upgrade;
      proxy_set_header Connection "upgrade";
      proxy_read_timeout 86400;
  }

}```
And the `config.yaml` looks like:
```turn:
  enabled: true
  tls_port: 444
  domain: <http://turn.com|turn.com>
  cert_file: /fullchain.pem
  key_file: /privkey.pem```
After visiting <https://www.livekit.io/connection-test|this> website, I am getting the following error:
```Can connect via TURN
Error: could not connect after timeout```
May you tell me what I am doing wrong?
Thanks!

<@U028SSAKLQ1> maybe you will know, beacuse I have seen that you had similar issue

<@U02SUCZ6TTN> please note that TURN is not using the HTTP protocol so you can‘t proxy it with the standard nginx proxy_pass. In this thread here I‘ve outlined how to use HAProxy to serve both HTTP and TURN on the same interface and port: <https://livekit-users.slack.com/archives/C01KVTJH6BX/p1637174009227500|https://livekit-users.slack.com/archives/C01KVTJH6BX/p1637174009227500>

<@U028SSAKLQ1> in <https://livekit-users.slack.com/archives/C01KVTJH6BX/p1627899576007300?thread_ts=1627898083.006900&amp;cid=C01KVTJH6BX|this> thread, you had paste Nginx snippet that allows us to use only it. But it is not working :(

<@U02SUCZ6TTN> sorry that link doesn‘t work. However, since I‘m no longer using nginx, I can‘t really help you with that anyway. If you want a working single-interface solution, you should really try out the HAP way.

<@U02SUCZ6TTN> did you manage to solve this issue?

To answer your questions:
1. It uses Chrome to composite the video and thus could be resource intensive. It needs to decode + render + composite + re-encode into a single stream.
2. Are you running the server with a redis instance configured?

Thanks for your reply <@U01KZ61GV5Z>.
1. I presume that if we were to have a Go service that joins the room with the identity "recorder", we won't have to use Chrome and that can consume less resource ? I'm new to the video space, so I'd really appreciate it if you can correct my understanding
2. Not currently as I'm running it locally for development. The production deployment will have Redis

To clarify, our use case is as follows:
• We don't want to composite the videos
• We want to save each stream separately (of each participant - these are not users, but drones).
• Save each stream with a unique filename and push to S3 (which is easy enough) 

livekit-recorder is kind of overkill for your use cases. without compositing, using livekit-sdk-go is the simplest way forward.. We do want to offer a built-in way to export those files, but haven't got to that yet. (currently it requires a bit of work on your end).

service mode isn't supported without redis.. so that'd be the thing to set up, even locally

curious: how are you ingesting the streams into livekit? is your drone Android based?

We thought so as well, and I don't mind working on it. I would appreciate some pointers if you don't mind. My understanding for our use case is as follows:
1. The recorder service is essentially a bot that joins a room, and downloads each participant's track (not sure if goroutines can be used for concurrent processing - I'm still learning goroutines).
2. Once the track is downloaded, we can then upload to S3
What I don't understand is how to download each track. I saw this in the docs: and I presume this is where the download takes place in the callback ?

```func trackSubscribed(track *webrtc.TrackRemote, publication lksdk.TrackPublication, rp *lksdk.RemoteParticipant) {}```

If I'm running LiveKit via Docker, does this count as service mode ? I followed the example on <https://docs.livekit.io/guides/getting-started>

`webrtc.TrackRemote` lets you read packets from RTP stream

<https://github.com/pion/webrtc/blob/master/track_remote.go#L159>

I'm not entirely sure what stream ingestion is (I just joined the project). The drone is DJI and we have a custom Android app (iOS incoming) that talks to the drone. I think LiveKit is merely a video broker, not necessarily a manager

My manager has given a pass on open-sourcing our solution if I manage to get this working. Thanks for the hint for track_remote!

fantastic to hear! seems like a cool use case for this. Let us know how we can help

Hi <@U01KZ61GV5Z> , sorry to trouble you, I'm currently struggling to understand how to get continuous stream of each video. I was able to write from the RTP stream to file using my `room.Callback.OnTrackSubscribed` handler. However, this only gets fired once for each participant, so my video file is always 0 seconds long. I tried printing "data printed in {event handler}` but only `OnTrackSubscribed()` gets triggered. I'm not entirely sure how to "leech" off the video track apart from using callbacks. Any pointers on this ? Thanks!

you have to read the packets in a goroutine.. in a loop.

no worries, we've all struggled at some point :slightly_smiling_face:

Hi <@U01KZ61GV5Z> , hope you're doing well! I think I've made quite a lot of progress on the recorder; it's catching the RTP packets and dumping them into the file as bytes. But I can't open the file afterwards and I'm convinced that I can't just dump the bytes into a file with an `.mp4` extension. Can you recommend any libraries to deal with video processing in Go ? I've scoured quite a few, from `ffmpeg` bindings to `gocv`, but none that fit this use case. Any suggestion is appreciated. Thank you and looking forward to hearing from you!

unfortunately dealing with media is a quite complicated, especially getting them into the same container file. Here's an example of doing something like that for .webm: <https://github.com/pion/example-webrtc-applications/blob/master/save-to-webm/main.go>

Gstreamer is a good option to look at, here's an example of feeding read bytes to GStreamer for further procesing: <https://github.com/pion/example-webrtc-applications/blob/master/gstreamer-receive/main.go>

Thanks for your suggestions, I think we were thinking of GStreamer as well, so I'll give it a go and update you on further progress.

Hi <@U01KZ61GV5Z>, just wanna give you a quick update on my progress. A bit slow, but I've managed to record one video stream into `.mp4` . Currently working on extending it to download each participant's video and optionally, audio stream separately. We're planning to test the recorder on Monday locally, and if our CEO approves of open-sourcing our solution, I'll let you know sometime next week :slightly_smiling_face:

that sounds awesome! I know at least a few people that had been waiting to be able to record individually!

That's good to hear! Since I'm still a junior, I'm not my API design is the best or most idiomatic. I presume that it's best to release it first and improve on it based on the feedback from the community ?

I'm happy to work with you to improve/iterate too. Would be great to build file output capability into the SDK itself

Yeah that sounds good! I'll update you once it's ready for release :slightly_smiling_face:

Hi <@U01KZ61GV5Z>, our CEO has given a pass. I’ll do 2 things:
• Submit a PR in livekit/server-sdk-go
• Publish the RecordBot service (not on Dockerhub yet) on our company GitHub. 
The service is a webserver where given Livekit url, api key and secret, users can send HTTP requests to “/recordings/start?room=…” and “/recordings/stop?room=…” and it will record in background. Only S3 upload is supported as of now

<@U02SGSET1RT> , that's awesome to hear! Having a separate service for it makes a ton of sense. I'm excited to see it!

<@U01KZ61GV5Z> I've made a PR (<https://github.com/livekit/server-sdk-go/pull/32>). Please let me know of any issues or questions :slightly_smiling_face:. Any feedback is appreciated!

I should add that currently the error handling is mainly through Panics or just return prematurely in `onTrackSubscribed` and `onTrackUnsubscribed`. I'm not sure how to handle errors in a closure that does not return an error. I was thinking of using something like `chan error` , but I'm open to suggestions.

This is incredible work <@U02SGSET1RT>! I'll try this later today and respond in the PR.

Thanks for your feedback <@U01KZ61GV5Z> , I’ll work on it over the weekend and hopefully it’s good to go into the SDK early next week!

hey <@U02SG8NA0DU>,  what does your environment look like? Is it on firebase? Here's a workaround: <https://github.com/livekit/server-sdk-js/issues/9>

yes I am on firebase and what you suggested is exactly what I did :slightly_smiling_face:
The server SDK is so far perfectly running now!

My biggest (and stupid) problem now is a single docker container based on your liveKit image running on a gcloud vm, or even on my localhost. I currently don't manage to get the ports open from outside the container for connecting the Server-SDK calls properly. Port 7881 is the only one that is connected, so the Angular Frontend is happy, everything fine here. But the Server SDK trying to connect to Twirp is not reaching the container, even via port 80?!
To be honest and sorry for that: I am not a DevOps guy at all so docker is not my core competency, but maybe you can give me a hint on that..., any help highly appreciated.

when changing the imports like this, transpilation errors disapeared:
import * as _m0 from "protobufjs/minimal";

did I do anything wrong here, of course its not optimal to change your SDK files :thinking_face:

Hey Travis, thanks so much for the kind words and welcome to the community! Happy to hear you’ve got things up and running quickly. :slightly_smiling_face:

1. Tokens encapsulate a few different pieces of information and are used to effectively authenticate a user to a room. You should use your API key and secret to generate them on the server side. The iOS SDK cannot generate tokens at this time — this is for security reasons. We don’t recommend embedding your API key and secret on the client. So, the best practice is, add this package to your Node server: <https://github.com/livekit/server-sdk-js> and then set up an endpoint for generating tokens. Here’s an example for you: <https://replit.com/@RussdSa/livekit-node-token-example>
2. Yes! Screenshare on iOS is supported ootb. I’ll defer to <@U02MAJLGN65> on if you can use RPScreenRecorder to create a custom track.
Let us know if you have any other questions or need more guidance!

Hi <@U02S5MWAW85> :slightly_smiling_face: Which type of screen share are you trying to do,
1. Screen share of the app’s screen only (in-app screen share).
2. Screen share of all apps (background screen sharing) while your app is in background.
If all you need is *in-app screen share* it’s very simple, you just need to call `LocalParticipant.setScreenShare(enabled: true)` and it will take care of creating `RPScreenRecordeder` internally.

Thank you both, that works for me Russ. And I’m using in-app screenshare so that's good to know Hiroshi

Hey! We don’t have a specific ETA, but it’s on our radar for this year.

Could you tell me a bit more about what you’re building and your needs? It helps figure out how to prioritize improvements and new features. :pray:

seems like you are running your code on firebase? If so, it will not be able to reach a local IP: `<http://192.168.178.125:8080>`

I am running everything on my local dev environment which is normally able to reach my localhost, e.g. for using the firebase emulators..., but I will check that.
Can you confirm that twirp and/or nginx are started along with running up the docker container as described above?

BTW: so far the same issues with the same docker container on gcloud...

I think it depends on how you've set it up.. the server API endpoint is the same as the one clients connect to. My guess is something firebase emulator is doing that makes it behave diffferently

<@U02BBTRGF98> is pretty pro :slightly_smiling_face:

fwiw got it running using delve in vs code - seems to work

It should be possible to debug code through a container, but there's no benefit as far as I can think of. You'd have to compile it and copy it into the container with every change. And you'd have to expose delve’s port

You're doing the right thing :sunglasses: 

as an fyi had to comment out the “Commands” section of main.go - it didn’t recognize generateKeys etc

also - any tips on how to size the vm for a ~6 people conference videocall?

I don't think we have any Jitsi-like web app clients, do we <@U01KZ0SCDKL>? Closest thing would be our demo app, or JS example app. 

LK can handle 6 users pretty easily, check out our benchmark results <https://docs.livekit.io//deploy/benchmark/|https://docs.livekit.io//deploy/benchmark/>

<@U02T7E5L1NE> <@U02BBTRGF98> is right, we don’t have a _full-featured_ client that can be deployed and self-hosted. Closest thing would indeed be our example app here:
<https://github.com/livekit/livekit-react>
Let us know if you end up putting together something more polished!

Typo in that URL: <https://docs.livekit.io/deploy/benchmark/>

it's surprising that I don't see projects on github already. How old is LiveKit?

also <@U02BBTRGF98> I had seen those, however the hardware used is pretty big, and the test is also on a pretty big scale. I was wondering if there's any data about smaller deployments, just to understand how much it would cost me in hosting

<@U02T7E5L1NE> quite young, certainly compared to jitsi, we launched the OSS project in July 2021

I see. Things sould start popping up then

6 users isn't that many so I would think you could use a small machine. My guess (without having tested it) would be a 2 vCPU server should be more than enough. But let me know how your testing goes!

I guess disk and RAM aren't big factors here

image.png

I run a livekit cluster on <http://fly.io|fly.io>'s smallest server types. 1 shared vCPU and 256mb of memory. Here are the server stats for one of my nodes at a time when 15 clients were joined, all running video and audio feeds. I wasn't in on any of these meetings, so I can't speak for how it was working - but I don't have any customer complaints, so I assume no issues :slightly_smiling_face:. One caveat is that with my use case, these are all low-def video feeds (320x240)... but still, LiveKit sips CPU and memory compared to other systems.

Maybe try a `go mod tidy` and try building again?

you can set the health_port on your recorder (it's an http service) to check the status of your recorder, you can also check the logs of your recorder docker

you can run it in service mode
<https://github.com/livekit/livekit-recorder#service-mode>

Ah thanks guys! I'll give it a go and report back

That kinda works! It says it started a recording, but the last message in the log is "Failed to load cuda library". When I don't run it in service mode it's exactly the same, except that after the "failed to load cuda library" I get a "state changed" and then it actually works.

Do you have any idea what might cause this?

The url in the "request validated" msg does work, I'm seeing the video stream when I open it in the browser

cuda is for gpu acceleration, that’s expected/safe to ignore

Yeah I read that in the docs. It seems that the service mode version isn't getting the `START_RECORDING` log, it's just logging:
```{"level":"debug","ts":1641676164.4720886,"logger":"livekit-recorder","caller":"logger/logger.go:22","msg":"chrome console error","msg":""}```

The speaker template should stay focused on screen share tracks. For the issue with your 4s files, there’s currently <https://github.com/livekit/livekit-recorder/issues/22|a bug> where it fails when all videos are muted

<@U02K46F1YLW> the recorder issues around muted video have been fixed in v0.3.13

Awesome :star-struck:. All I need todo is update the recorder or do I need to update LiveKit proper also?

just the recorder, no protocol changes this time

in golang?

```import (
   "<http://github.com/livekit/protocol/auth|github.com/livekit/protocol/auth>"
   ...
   ...
)

keys := map[string]string{
   os.Getenv("LIVEKIT_KEY"): os.Getenv("LIVEKIT_SECRET"),
}
provider := auth.NewFileBasedKeyProviderFromMap(keys)
data, err := webhook.Receive(r, provider)```

good call.. we should document this. Thanks <@U02HN8GPUBY> !

assuming you are talking about the Go SDK. it returns a `webrtc.TrackLocal` with `LocalTrackPublication.TrackLocal()`

i.e. `localParticipant.Tracks()[0].(*LocalTrackPublication).TrackLocal()`

thank you. if there are multiple local tracks how might i locate it? doesn’t seem any of the fields like “kind” have usable info?

i’m rather new to go and it’s usage of interfaces so forgive me if this is obvious

TrackPublications contains metadata about the tracks.. such as user supplied name, kind, and source

you'll want `setMicrophoneEnabled` . See: <https://docs.livekit.io/client-sdk-js/#handling-common-track-types>

Thanks for the reply, but what I want is not actually turn on or off the camera or microphone. I want to assign isMicrophoneEnabled to a variable, but I think we don't have access to this value directly (as far as I can see, there is only a get method in the Participant class. So even though I wanted to set it to true, I guess it's throwing false because it couldn't find it.).

ah I misunderstood earlier. You are trying to reference the value in a variable. Unfortunately that isn't possible with a getter. it only returns the value at the moment when it's checked

so is there a way to check if it is enabled? I check it after call enableCameraAndMicrophone() method

did you `await` for the promise to complete?

when I console log it after that I can see that it is already there and true. But cannot assign it

Here there is only 1 line between them. If I look participant, it is true but if I look for the participant.isMicrophoneEnabled it is false.

hey take a look at gstreamer, it's pretty powerful in terms of the set of things that you could do with it

so i got the docker file to run and it says done but when i try to launch the site using <http://example.livekit.io|example.livekit.io> or via my domain i can't

on example.livekit i get LiveKit Video
Show Stats


0
error Failed to construct 'WebSocket': The URL '
<http://video.clarkeverson.com/rtc?access_token=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJleHAiOjE2NDQ2Mzc3MjQsImlzcyI6IkFQSW16SDhwZE1zVXZqMiIsImp0aSI6ImNsYXJrIiwibmJmIjoxNjQyMDQ1NzI0LCJzdWIiOiJjbGFyayIsInZpZGVvIjp7InJvb20iOiJteXJvb20iLCJyb29tSm9pbiI6dHJ1ZX19.nD3-YmNDlgIMff5zpJjry47aZvLcWUGiOYIJlgAcDZc&amp;protocol=5&amp;sdk=js&amp;version=0.15.2|video.clarkeverson.com/rtc?access_token=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJleHAiOjE2NDQ2Mzc3MjQsImlzcyI6IkFQSW16SDhwZE1zVXZqMiIsImp0aSI6ImNsYXJrIiwibmJmIjoxNjQyMDQ1NzI0LCJzdWIiOiJjbGFyayIsInZpZGVvIjp7InJvb20iOiJteXJvb20iLCJyb29tSm9pbiI6dHJ1ZX19.nD3-YmNDlgIMff5zpJjry47aZvLcWUGiOYIJlgAcDZc&amp;protocol=5&amp;sdk=js&amp;version=0.15.2>' is invalid.

i re-ran the docker cause i realized i missed a space but still get LiveKit Video
Show Stats


0
error server was not reachable

this is my updated docker compose
```version: '3'
networks:
   caddy:
      external: true
services:
   livekit:
      image: "livekit/livekit-server"
      restart: always
      environment:
         LIVEKIT_KEYS: "API: Secret"
         dev: null
         node-ip: 23.95.166.25
      expose:
         - 7880
         - 8000-10000
      ports:
         - "7880:7880"
         - "7881:7881"
         - "7882:7882/udp"
      networks:
         - caddy
   caddy:
      image: caddy:2-alpine
      restart: unless-stopped
      ports:
         - "80:80"
         - "443:443"
      volumes:
         - /data/caddy/Caddyfile:/etc/caddy/Caddyfile
         - /data/caddy/data:/data
         - /data/caddy/config:/config
      networks:
         - caddy```

and my caddy file
```{
    # Global options block. Entirely optional, https is on by default
    # Optional email key for lets encrypt
    email <mailto:myemail@email.com|myemail@email.com>
    Optional staging lets encrypt for testing. Comment out for production.
    acme_ca <https://acme-staging-v02.api.letsencrypt.org/directory|https://acme-staging-v02.api.letsencrypt.org/directory>
}
<http://video.clarkeverson.com|video.clarkeverson.com> {
    reverse_proxy localhost:7880
}```

sorry, not too familiar with caddy here. letsencrypt cert should work ok, but I'm unsure if the staging cert is accepted by browsers as being valid.

Yeah but if a cert wasn't accepted wouldn't it say unsafe site can't access? Not that it can't find the server?

okay so i tried this without caddy and it said it wasn't secured on <http://example.livekit.io|example.livekit.io> but going to the IP it said: OK and i tried with caddy production and <http://example.livekit.io|example.livekit.io> not available as did the website

are there any examples of doing it in docker compose with nginx or some other method to make it work?

as I mentioned, you need a CA issued cert for https/wss to work.

majorities of connection issues that folks see is because of that.

you can test it by going to `https://&lt;yourhost&gt;`

i got it working on caddy something very small was missed but i can write some documentation for it later for it to be shared...very easy to use the caddy side, hit a different issue but its unrelated to this so will do new thread

do you mind posting the thing you changed to get it to work? for those that might find the thread in the future :slightly_smiling_face:

Where I put localhost I just had to put the name of the container. So in this case livekit

Here is same. If room was created by API &amp; no user joined then metadata update won't work. Would be helpful if we could set metadata during create meeting. Waiting for resolve <https://github.com/livekit/livekit-server/issues/196>

hey <@U02LCP104CQ>, that's still on our roadmap. we'll look at it shortly

Thanks <@U01KZ61GV5Z>.. as a work around I'm updating information to redis directly.

have you tried to delete the room with livekit-cli utility to see if the same thing happens?

Just tried &amp; same problem but in log

```2022-01-13T16:17:41.689Z	INFO	livekit	logger/logger.go:30	Could not find room	{"room": "test01"}```

OK, I have found the problem now. If I create room using server API then it don't get ready until someone joined the room. So, delete or update metadata won't work. After one user join it seems those started working but for my application I need metadata to be set during room creation. That's why I tried to send in second request to update metadata.

ah, interesting finding.. do you mind filling an issue for it with your findings? we'll take a look

Thanks. I've opened here: <https://github.com/livekit/livekit-server/issues/338>

Hi, I searched and came across this thread. Is this issue fixed? I'm using version 1.0.1 and I can confirm that I have no issue when updating room metadata but the issue with DeleteRoom seems to still exist. When I create a room using server api, I cannot delete that room using server api unless at least one user has already joined that room. I use postman to directly manage rooms. I can list, create rooms without any issue, but when I want to delete a room it returns http error 500. if I connect to that room once and delete the room later then there is no issue and the room is deleted successfully. the returned error response is:

{
    "code": "internal",
    "msg": "operation cannot be completed",
    "meta": {
        "cause": "*errors.errorString"
    }
}

i concur, I am experiencing the same bug using <https://github.com/livekit/server-sdk-js> v1.0.1

also, is the property `cause: "*errors.errorString"` a value to expect?

Does this property have other values?

<@U03FWUU3P7C> This was fixed in <https://github.com/livekit/livekit/pull/721>. That change made it into 1.1.0 release. Would you be able to try that release?

<@U03KRLGTXTK> Wondering if you are hitting the same error reported in this thread (in which you reference this thread :slightly_smiling_face: ) - <https://livekit-users.slack.com/archives/C01KVTJH6BX/p1655299691819099>. Would you be able to try 1.1.0 as well?

Hi <@U02CM28PB4Y>, sorry for the late reply. I tested the scenario and I can confirm that it's fixed in 1.1.0 release. I can now create a room using server api and delete it using server api before any user has connected to it. thank you.

Thank you <@U03FWUU3P7C> for the confirmation.

A couple of questions:
• is it recoverable automatically? or do the participants need to rejoin to recover?
• do you know is if it's on the publisher or subscriber side. (i.e. if everyone is seeing an issue with a single user's streams, then it's likely on the publisher's side)
• is it happening to everyone regardless of their platform/network/etc?

• it doesn't seem to auto-recover, sometimes after 15 minutes or so it may recover.
• the black screen seems to be the publisher, but the stream not being updated seems to be on the receiver, as some people see it and some other don't
• everybody is experiencing these issues, I'm testing it on a mixed environment, all flutter apps, on android, ios and web


hmm really odd, we haven't heard/seen this reported by anyone. Can you think of anything you might be doing differently with the integration and/or setup?

<@U01KZ61GV5Z> I actually have something particular on my implementation

I set autosubscribe to false, then:
• on room join I manually subscribe to each publication
• on TrackPublishedEvent I subscribe to every publication
• on TrackSubscribedEvent I create an object that stores the publication
• on TrackUnsubscribedEvent I delete the object 
then I render all object's publication's track with VideoTrackRenderer

also I disabled the optimizeVideo when connecting
on web build I disable simulcast
on mobile I enable simulcast
on mobile and web I set VideoQuality.LOW  for the subscribed publications

<@U01KZ61GV5Z> any ideas what to test on my side?

Hey Alex, I'm not quite sure what might be happening, but wonder if you would have a way for me to reproduce? either with the sample app or I'm happy to try your app too.

thanks <@U01KZ61GV5Z> I'm trying to make a reproduction of it, it's hard as it is very random,, I'll let you know

thanks Alex. I know how some of these things are. We'd like to get to the bottom of it either way.

if i am reading this correctly, i assume due to the docker network, the ip address is 172.18.0.3 for host? if so then that's what i need to change...as it's not the ip I ssh into it would have to come from docker

yes! precisely.. if you are running on Linux, then you should be using host networking with Docker. and if it's deployed, you should enable <https://github.com/livekit/livekit-server/blob/master/config-sample.yaml#L30|use_external_ip>

so i rebooted my docker container using
```docker-compose up --force-recreate --build -d --remove-orphans```
even after updating my docker compose to use the docker ip address its timing out
```version: '3'
networks:
   caddy:
      external: true
services:
   livekit:
      image: "livekit/livekit-server"
      restart: always
      environment:
         LIVEKIT_KEYS: "APImzH8pdMsUvj2: hMy34yUgIFRyDQhwQfUprjdDH5pvKzFV5jeZ9YXv5Il"
         dev: null
         node-ip: 172.18.0.3
      expose:
         - 7880
         - 8000-10000
      ports:
         - "7880:7880"
         - "7881:7881"
         - "7882:7882/udp"
      networks:
         - caddy
   caddy:
      image: caddy:2-alpine
      restart: unless-stopped
      ports:
         - "80:80"
         - "443:443"
      volumes:
         - /data/caddy/Caddyfile:/etc/caddy/Caddyfile
         - /data/caddy/data:/data
         - /data/caddy/config:/config
      networks:
         - caddy```
so i know caddy is working but livekit something isn't working, what is my docker compose missing?

I am okay restarting container right now because I am just working on all the correct files and documentation to make this easily deployable

these are the logs
```docker logs 96faf4476b95
{"level":"info","ts":1642121112.2290483,"logger":"livekit","caller":"logger/logger.go:26","msg":"using single-node routing"}
{"level":"info","ts":1642121112.2303796,"logger":"livekit","caller":"logger/logger.go:26","msg":"starting LiveKit server","addr":":7880","nodeID":"ND_1VYgB3l5","nodeIP":"172.18.0.3","version":"0.15.2","rtc.portTCP":7881,"rtc.portICERange":[50000,60000]}
{"level":"info","ts":1642121182.0775025,"logger":"livekit","caller":"rtc/room.go:214","msg":"new participant joined","room":"myroom","pID":"PA_bePATd8zeLCV","participant":"clark","protocol":5,"room":"myroom","roomID":"RM_Czd5ZFNM6Qks"}
{"level":"info","ts":1642121182.0785022,"logger":"livekit","caller":"service/rtcservice.go:174","msg":"new client WS connected","room":"myroom","participant":"clark","connID":"clark","roomID":"RM_Czd5ZFNM6Qks"}
{"level":"info","ts":1642121187.1334035,"logger":"livekit","caller":"service/rtcservice.go:155","msg":"server closing WS connection","room":"myroom","participant":"clark","connID":"clark"}```

hey clark, your IP is still wrong: `172.18.0.3`  it needs to be a publicly accessible IP in order for this to work

did you checkout what I posted earlier about use-external-ip?

I did but I thought what you meant there was that if I didn't use the docker ip I used that my bad
So for a docker compose do you put that under the node ip as another environment variable?
And the external ip would be the same I use to ssh? Like I was using before 

so i u[dated docker compose to be
```version: '3'
networks:
   caddy:
      external: true
services:
   livekit:
      image: "livekit/livekit-server"
      restart: always
      environment:
         LIVEKIT_KEYS: "APImzH8pdMsUvj2: hMy34yUgIFRyDQhwQfUprjdDH5pvKzFV5jeZ9YXv5Il"
         dev: null
         node-ip: 23.95.166.25
         use_external_ip: "true"
      expose:
         - 7880
         - 8000-10000
      ports:
         - "7880:7880"
         - "7881:7881"
         - "7882:7882/udp"
      networks:
         - caddy
   caddy:
      image: caddy:2-alpine
      restart: unless-stopped
      ports:
         - "80:80"
         - "443:443"
      volumes:
         - /data/caddy/Caddyfile:/etc/caddy/Caddyfile
         - /data/caddy/data:/data
         - /data/caddy/config:/config
      networks:
         - caddy```

and i confirmed thats the external ip of the docker

with this `docker exec -it 35193857f81f wget <http://ipecho.net/plain> -O - -q ; echo`

I don't think your compose file is setting the config correctly. "node-ip" isn't an environment variable. it's meant to be passed in as a command line arg

if you are using env, then set the following config as var `LIVEKIT_CONFIG`. see <https://docs.livekit.io/deploy#configuration>

<@U02KJGV339S> are you using firefox? Did you configure your server with SSL? If you don't have SSL certificate then from Firefox you'll face this type problem.

I’m not using docker at all :see_no_evil:

<@U01KZ61GV5Z> just getting back to this. So for the config file I would add LIVEKIt_CONFIG: "path/to/file"
and if I am not using turn I would just remove that correct? and put it below LIVeKIt_KEYS?

<@U02SQNHPNHG> LIVEKIT_CONFIG will need to be the full body of the config. Is there a reason why you are using docker-compose? is this for local testing or for a server deploy?

It's for a server and I'm trying to make it easily redeploy able. In case we ever need to move it or setup multiple servers etc

I see, you can try embedding the config file like:
```environment:
  LIVEKIT_CONFIG: |
    &lt;content of your config.yaml&gt;```

hi, are you using <https://docs.livekit.io/deploy/vm|this process> to start it? if so, it'd be straight forward to install redis within that template

So far I only tried the livekit image from dockerhub. I didn't have a look at this guide as I have never used cloud-init before. I will sure have a look at it

are you trying to run it directly? or via Docker?

if docker you can do `docker run -e LIVEKIT_CONFIG=$(cat config_file.yaml)`

thanks so much <@U01KZ61GV5Z> I will try that :grinning:

if I run with docker compose how would I inject the config file? thanks.

it should be the same.. refer to the env var $LIVEKIT_CONFIG in the compose file.. then `
```docker-compose run -e LIVEKIT_CONFIG=$(cat config_file.yaml)```


This bug with url recordings should be fixed in livekit-server v0.15.2, have you updated your server?

Now recording is starting successfully. But stopRecording is failing.

The API fails with `{"code":"internal","msg":"request timeout","meta":{"cause":"*errors.errorString"}}`

docker logs from recorder is as below.

```{"level":"debug","ts":1642244174.7573707,"logger":"livekit-recorder","caller":"logger/logger.go:22","msg":"[x264enc0] LATENCY - Element's latency has changed"}
{"level":"debug","ts":1642244176.6807399,"logger":"livekit-recorder","caller":"logger/logger.go:22","msg":"[pipeline] STREAM-START - Pipeline stream is starting"}
{"level":"debug","ts":1642244176.6816459,"logger":"livekit-recorder","caller":"logger/logger.go:22","msg":"[filesink0] STATE-CHANGED - State changed from READY to PAUSED"}
{"level":"debug","ts":1642244176.682031,"logger":"livekit-recorder","caller":"logger/logger.go:22","msg":"[output] STATE-CHANGED - State changed from READY to PAUSED"}
{"level":"debug","ts":1642244176.682392,"logger":"livekit-recorder","caller":"logger/logger.go:22","msg":"[pipeline] ASYNC-DONE - Async task completed"}
{"level":"debug","ts":1642244176.6834185,"logger":"livekit-recorder","caller":"logger/logger.go:22","msg":"[filesink0] STATE-CHANGED - State changed from PAUSED to PLAYING"}
{"level":"debug","ts":1642244176.6837373,"logger":"livekit-recorder","caller":"logger/logger.go:22","msg":"[output] STATE-CHANGED - State changed from PAUSED to PLAYING"}
{"level":"debug","ts":1642244176.6837795,"logger":"livekit-recorder","caller":"logger/logger.go:22","msg":"[pipeline] STATE-CHANGED - State changed from PAUSED to PLAYING"}
{"level":"debug","ts":1642244197.7077339,"logger":"livekit-recorder","caller":"logger/logger.go:22","msg":"handling request","recordingId":"RR_z3PJoKe9ECcG","requestId":"RPC36MaJ3WWdYK6"}
{"level":"debug","ts":1642244197.7079377,"logger":"livekit-recorder","caller":"logger/logger.go:22","msg":"sending EOS to pipeline"}```


when an EOS is sent but there’s no `EOS received` message, that usually means the recorder is stuck because it’s not receiving any video data (which is an open bug that we plan on fixing).
This might be because it failed to connect to the room, or because the video tracks are all muted, or because everyone has left the room already

if the issue is the third case for you, you can try setting `template_address: <https://recording1.web.app/#>` and using the template input with `layout: my-layout`.
Then your layout can <https://github.com/livekit/livekit-recorder/blob/a0d7606449e4e671264c3a83d38afe533af495d1/web/src/common.ts#L29|log> `END_RECORDING` to the chrome console, and it should stop the recording before getting stuck

With template_address, layout it is working now. Thanks! I think it was not working because of connectivity issue. In URL mode, how are we supposed to pass the roomName?

Url mode doesn't take any extra input and won't generate a token - if you're recording a single room, it's recommended to use the template input

currently, livekit-server itself doesn't terminate the certificates. you'd have to use a load balancer, or reverse proxy that can terminate TLS

I'm playing around with automatic SSL with caddy at the moment. that's a pretty decent option to put in front of.

identity is the same value you've assigned to the participant in the access token