LiveKit Community

welcome Jonathan! yeah, the server sdk should run fine in a lambda. Let us know if you see any issues with it!

You need to make changes and add the same redis address used by your livekit server

the Livekit Server is using the default localhost:7880 address. If I set the REDIS_HOST to localhost:6379 the livekit server doesnt want to connect to redis.  I get the error: "unable to connect to redis: dial tcp 127.0.0.1:6379: connect: connection refused."

Copying from our thread for anyone else who needs it:

IIRC you need to:

Replace `bind 127.0.0.1` with `bind 0.0.0.0` in the /etc/redis/redis.conf file
Replace `protected-mode yes` with `protected-mode no` in this same file
Add `--network host` to your docker run command

Alright its working now - I just needed the flag --network=host

That means your livekit server is sending a request to redis but it isn't being picked up by the recorder - each recorder instance can only record one room at a time, so if you're already recording you would need more recorders, otherwise make sure the recorder deployed successfully and that you're using the same redis `db` for the server and recorder configs

I think I can figure it out by checking for `track.sender` and then using `createLocalVideoTrack` to re-establish the video.

hi! when you unplugged the camera, I'm guessing the track became unpublished?

that error message concerns me a bit.. I'm wondering why it's trying to restart the track. can you share the SDK calls that you are making and their sequence?

I have a button in my UI to turn the camera back on and I used that, it called unmute

``` if (!isVideoEnabled) {
      this.localVideoTrack?.mute();
    } else {
      this.localVideoTrack?.unmute();
    }```

sequence of events:
1. unplugged USB (accidentally)
2. plugged it back in
3. hit the "enable camera" button which called unmute() on the localVideoTrack which was not gone. 


ah ok, got it. maybe we could handle this bug differently then

the unmute event tries to automatically "restart" the track.. but in this case it was unpublished when the camera was unplugged.. perhaps it should automatically republish the track

that's basically what I did.. I made a little function
```  // Returns true if we had to start a new track
  private ensureLocalVideo = async ():Promise&lt;boolean&gt; =&gt; {
    if(this.localVideoTrack?.sender) {
      return false;
    }
    // if the track has been unpublished, start a new one
    // TODO, use options and add catch
    const videoTrack = await createLocalVideoTrack();
    this.localVideoTrack = videoTrack;
    this.room?.localParticipant.publishTrack(this.localVideoTrack);
    return true;
  }```

this looks good. let me think if there's a way for the SDK to do this automatically.

i made it return a boolean in the case of calling restartDevice when changing deviceId so we don't restart the camera twice... if that makes sense. I dunno, sometimes my code is a mess :joy:

yeah, you can do so with <https://docs.livekit.io/guides/server-api#updateparticipant|updateParticipant>. every participant (including the local participant) should receive a <https://docs.livekit.io/client-sdk-js/enums/roomevent.html#metadatachanged|metadataChanged> event

Ahh, it's a server api call. I was looking under the client apis. Yeah, I guess that works.

Hrmm, am I correct that you have to be a room admin for this? So an end user can't update their own metadata without being an admin?

yup, server APIs are signed with api key/secret

the server sdk will automatically attach the required perms

Ahh, yeah. I guess that makes sense since metadata is protected by the JWT during a connect. I guess I shouldn't use metadata for what I was thinking. I need something that the user's client can update itself.

ah i see.. would your backend be able to validate permissions?

i.e. instead of calling livekit client sdk, talk to your own backend which could validate perms and then update metadata using server sdks

Not currently. At the moment I'm running this in a pretty light setup. The clients actually have access to an API key to generate their own auth tokens, so I _could_ do this with those credentials. But my plan is to rip that out once I built a proper service to handle the authentication.

I have other ways I can pass &amp; store the info. Metadata just looked convenient at first, but I was misinterpreting what its use was.

yeah cool, it's more for security that we disable changing it by the client

Makes sense. It'd be nice to have a similar insecure datastore for tracking user-specific data that would be synced between the clients. Something that a user could update their own record, and maybe room admins could also change for a user. However, not critical since there are other ways it can be done.

we don't currently expose this. what are you trying to do when they are finished?

append them to the SDP response and send them back to the client (I'm working on a kind of signalling proxy for some devices that can't use any of the supported client libraries)

ah I see. I'm don't think there is a way to determine this with the protocol today.

okay, well it seems like the livekit-server always returns exactly 4 ICE candidates (at least the way i have it set up), so I guess for now i'll just have it finish gathering after 4 candidates, but I guess i have to persuade the downstream team we need to do real trickling.

Screenshot 2021-08-23 at 10.16.31 PM.png

can you share the command line of how you are running the server?

Screenshot 2021-08-23 at 10.20.27 PM.png

Screenshot 2021-08-23 at 10.21.27 PM.png

after press the connect button i’m getting this error

not that I know of, but if you have docker its really just one command

```docker run --rm \           
  -p 7880:7880 \
  -p 7881:7881 \
  -p 7882:7882/udp \
  -e LIVEKIT_KEYS="somekeysyougenerated" \
  livekit/livekit-server \
  --dev \
  --node-ip=YOUR_LOCAL_IP```

already able to run through docker but i need run it through source according to documentation i can’t able run the server from local machine i mean

% ./bin/livekit-server generate-keys
these command actually not running

Screenshot 2021-08-23 at 11.13.22 PM.png

Screenshot 2021-08-23 at 11.15.05 PM.png

is there anything wrong during setup the server … …?

Screenshot 2021-08-23 at 11.22.03 PM.png

doesn't look like it's run successfully...

and can you `ls` the livekit-server dir for me?

Screenshot 2021-08-23 at 11.25.36 PM.png

so, doesn't look like `/Users/mir00r/go/bin` is in your $PATH

and you'll only end up with a bin dir in the livekit-server dir after a successful build with mage

image.png

You should be able to generate keys using docker too: <https://docs.livekit.io/guides/getting-started/>

Screenshot 2021-08-23 at 11.46.31 PM.png

<@U01KZ0SCDKL> now ./bootstrap.sh says nothing

Screenshot 2021-08-23 at 11.47.12 PM.png

Could be cool to make a screen cast though

Screenshot 2021-08-23 at 11.49.08 PM.png

wow it’s working thanks man <@U01KZ0SCDKL>

btw when we get flutter documentation ..?

Screenshot 2021-08-24 at 7.53.38 PM.png

are you using different tokens for the participants?

actually bit confuse on this section i create one token using this command

`docker run --rm -e LIVEKIT_KEYS="APIMr2NJJy7VRHV: tZWn7MLOUiDuABbCHAeHDOXpIjRfmCMSFED7a0zWJrfB" livekit/livekit-server create-join-token --room myroom --identity ND_6vftug79`

the identity key get after running this command

`LIVEKIT_KEYS="APIMr2NJJy7VRHV: tZWn7MLOUiDuABbCHAeHDOXpIjRfmCMSFED7a0zWJrfB" ./bin/livekit-server --dev`

 and successfully created localParticipant and login successfully

but for remoteParticipant where i can found the identity key to generate the token .. for testing i use random identity like below

`docker run --rm -e LIVEKIT_KEYS="APIMr2NJJy7VRHV: tZWn7MLOUiDuABbCHAeHDOXpIjRfmCMSFED7a0zWJrfB" livekit/livekit-server create-join-token --room myroom --identity PA_ZJxrMzArg4NM`

and generate the token and try to login from another browser but can’t able login ..

identity is provided by you to identify the user.. i.e. a username.

from the browser, which app are you using to log in?

are you trying to log in with two instances of the same flutter app, each with a different token?

Screenshot 2021-08-24 at 11.10.20 PM.png

Screenshot 2021-08-24 at 11.09.07 PM.png

the first participant login to room successfully

but second participant can’t able to login into room

can you confirm if I'm understanding what you are doing correctly?

<@U01KZ61GV5Z> from livekit github repository

•  i cloned livekit server and setup it perfectly my local machine 
• running livekit server from my local machine according to livekit documentation 
• Then i clone livekit flutter client sdk repo and setup in my local machine 
• After running the flutter app i enter the room using URL, and generated token 
• Then i try to generate another token from my local machine livekit server to join the first participant 

my point is

how can I create conference calls using your sample flutter application and livekit server in my local machine ..?

to connect any remote percipient what is the information i need to share with remote candidate …?
<@U01KZ61GV5Z> <@U01KZ0SCDKL>

to connect any remote percipient what is the information i need to share with remote candidate …?

yeah, this is possible with reliable delivery. we'll give it a test and update the docs accordingly

data channel chunks messages that exceed MTU, but doing so on unreliable channel would cause a message spanning multiple packets to be dropped entirely

do you already have a redis instance set up? if so you can specify it <https://github.com/livekit/livekit-helm/blob/master/server-sample.yaml#L19|here>

Greetings! I’m trying to deploy livekit and have a doubt about the ICE ports. Is it possible to deploy on a platform where direct exposure of each instance to the internet is not possible?

Answering my own question here. It looks like this is not possible.

I’m deploying on <http://Fly.io|Fly.io> which does not support direct VM access.

Hey Josh, that's correct. it needs to expose these UDP ports so clients that are behind NAT could connect to it

unfortunately you can't load balance UDP very well given its session-less nature

Thanks. Another thing then. For environments that do not directly expose ports, but can do port forwarding at the load balancer, it might be possible to use hostnames or another approach like HTTP headers to direct the path of a request. Would it be difficult to adapt Livekit to use another method besides knowing its own public IP address?

<@U02CEMTDG2E> You may have already figured this out, but you can do a single node on Fly easily by using your Anycast IP as the node IP.

You can then scale that to multiple nodes by having a separate fly app for each node so you get a separate IP for each one. If you want the signaling servers load balanced, it does mean that you then have to run another load balancer like HAProxy in front of it and do some config magic if you want to still have the benefits of the Anycast routing. It's not ideal, but it works.

I have some notes written up about using livekit on fly and will be publishing those soon.

Yeah, I figured it out, and it’s not ideal :smile:

Fly also has the ability to add regional IPs, so you could have one node per region in a single app assigned to the regional IP. However, it's isn't guaranteed that that it will always route to that region - if the node there is offline or overloaded, it'll reroute, which will cause hard-to-track-down bugs, so I was hesitant to try that way.

Another idea I had was using TURN to handle the load balancing. This setup would only expose the node on the ipv6 addresses, and TURN would take the ipv4 address and relay to ipv6. But it looks like the livekit turn server isn't quite up to that task yet (ipv6 isn't fully implemented).

I agree with you that being able to route via some other header info. I think that could be done with TCP, but I don't think fly has a way to select a specific node with UDP at all at the moment.

<@U02C10ZEP8R> have you managed to get it to run on <http://fly.io|fly.io>? I'm not familiar with the platform, but curious on the NAT capabilities there

<@U01KZ61GV5Z> Yep, I've got it running there now.

I've got my quickstart guide on how to do a single node instance on Fly pretty close to ready, I'll publish it soon. The multi-node nodes need more cleanup before I can make those public, though.

`Would it be difficult to adapt Livekit to use another method besides knowing its own public IP address?`
<@U02CEMTDG2E> LiveKit discovers the node's public address and it doesn't have to be explicitly configured. For best availability, ideally the ports are opened on the firewall. However, WebRTC can work behind NAT too. (tho there are some situations where that wouldn't work)

<@U02C10ZEP8R> that's awesome! I can link to your guide from our docs.

The auto-discovery doesn't work on Fly. It ends up grabbing some edge router IP address which won't actually route to a Fly app. You have to specify NodeIP in that case.

I am trying to convince fly to enhance its router for these kinds of things, but it’s low on their list of priorities.

<@U02CEMTDG2E> Have a post on the community for that or anything? I'd be happy to add my voice to that in hopes of slightly bumping the priority.

Nah, I have been using an internal Slack as I am consulting for them, but I should post on community anyway

Pushing on multiprocess support worked at least!

Ah, I see I was already commenting on your thread there today

Oh, thanks for the reminder about that thread. I'll update that as soon as I publish the guide.

When I try to open the video after the signal lost I cant play the recorded video

a `SIGKILL` probably means you’re running out of resources

Yeah, when running locally via docker (with --dev) flag, it sets it up to use single port mode (where it handles all UDP traffic on a single port). This isn't as performant in production, but is much easier to work with when mapping docker ports.

the mapping itself is required due to Mac/Windows Docker not supporting host networking

so if I wanted to start with single node docker deploy for a sorta-production but low volume deployment and  enabled host networking, would I keep it in single port mode?  or should I change that, is it configurable?

yup you can. see: <https://github.com/livekit/livekit-server/blob/master/config-sample.yaml#L35>

Screenshot 2021-08-30 at 2.47.37 PM.png

Screenshot 2021-08-30 at 2.49.32 PM.png

Screenshot 2021-08-30 at 2.57.55 PM.png

Screenshot 2021-08-30 at 2.50.47 PM.png

Screenshot 2021-08-30 at 2.48.35 PM.png

In my experience, "failed to fetch" is a general error for not being able to reach the server for one reason or another. What's the status say in that second screenshot, or is there anything else in the console log? It will sometimes give additional info like if it was unable to resolve the domain or if the connection was refused.

Do you have a load balancer in front of your server terminating SSL? If so, is port 7880 correct for that? I don't believe LiveKit itself provides TLS on the signaling port, only for TURN if you configure it to do so. For the signaling port, if you want SSL, you need to put it behind some kind of reverse proxy to handle that.

nope don’t have load balancer <@U02C10ZEP8R> and there is nothing left to show you for better understanding the problem

If you don't have a load balancer, then port 7880 is not `wss://` . LiveKit doesn't provide a secure websocket connection. Try just `ws://`. You'll only be able to do this from localhost, though. Chrome and most other browsers don't let you do media over insecure connections. Outside of testing, you'll want to put this behind some sort of reverse proxy to terminate SSL for you.

Screenshot 2021-08-30 at 2.50.13 PM.png

when i enter connect button to login the room there is no logger actually printed on my livekit server logger … anyone facing this type of issue ever  …?

<@U02C10ZEP8R> I've discussed <https://livekit-users.slack.com/archives/C025KM0S1CK/p1627883801026400?thread_ts=1627870972.020200&amp;cid=C025KM0S1CK|here> why I don't think that would be a good change. Do you agree with my arguments there?

I understand your point, but believe the opposite of your conclusion. I think that if a remote unmute operation is needed, then that should be responsibility of the client side code to handle. You could send a signal the the client code to have it initiate the unmute operation from there end.

And in this case, the code in question is used for when the client &amp; server get out of sync with each other (for instance, when there is latency between the client and server). In that case, I think the client should always win on what their state should be.

<@U028SSAKLQ1> how do you currently issue a remote umute? I'm not sure this change would affect you.

<@U02C10ZEP8R> thanks for your explanation. I think this is a discussion worth having, but unfortunately I have to leave right now. So far I'm still convinced that the server should be able to control this without any "rce hacks" but I'll think about this some more (especially considering the sync issue) and will continue the discussion here ASAP.

Sounds good. I get your point about needing a way to do it and, but I think it's possible that this change wouldn't even affect your use case. I can test it out with the change in place once I know how you currently issue the remote unmute commands.

all good points. I'll review and follow up

looking into this, <@U02C10ZEP8R> is right. With the delays in signaling, there isn't a way we could distinguish a signaling layer delay from a change made by a server API.

we still need some sort of synchronization here to determine which is the "latest" change. Without it, we won't be able to mute a clien'ts tracks remotely

<@U01KZ61GV5Z> just for clarification: From a technical standpoint it shouldn't really matter whether the server or the client control the "single truth" about a clients state and therefore have precedence in a conflict situation (as long as it's either or), right?

In this case I see two options:
1) Decide about who should be the deciding party (A) and implement a way to trigger a state change in (A) by the other party (B).
2) Apply a sync mechanism that can handle conflict resolution.

yeah, so the tricky thing is both client and server could initiate a change. For example, an admin should be able to remotely mute a participant

currently there isn't a way to distinguish that event from an acknowledgement of the participant muting a track.

Is that what `LocalParticipant.updateInfo` does right now? Would it be the method that gets called if you were to remotely mute a track?

I will have a PR for this shortly. will tag you on it to take a look.:grinning:

<https://github.com/livekit/client-sdk-js/pull/40>

hey <@U02AB4XB03A>, any redis instance would work. It doesn't need a ton of resources for this.

Keep in mind memory store can only be accessed from same zone if using internal ip

<@U01TC9Z6ELD> our goal is global footprint. what is better/best  way to connect  Livekit GKE cluster to memorystore?

Our team never quite solved this :) global redis is hard.

We currently have a multi cluster svc set up but we are only using US due to other services slowing down the eu cluster (firebase was very slow to connect from eu for some reason). Given that, we have redis in same us zone as our cluster.

<https://webrtcforthecurious.com/> is a great starting point

Hi SIZeroth! Thank you for all your PRs! :pray::skin-tone-4: 

<@U01KZ61GV5Z> currently, one user only can login in to one session , but it is a common scenario that a user login from diffence devices.

for exmaple ,one user loign from pc to share screen and login from mobile to talk

hey Jason, would you consider using a different identity for the user?

within the system, identity is treated as a unique identifier.

This is what we did, we just add a device uuid to the identity

Id:uuid, then you can split(':')[0] to get the original id

thanks ,david ,rowny ,I will add deviceid as identity

This makes sense to add, we’ll look into it

Yup, some webhooks are available today. We’ll be adding more over time.

Since that first section is just the username/password, I think you can do something like this:

```redis:
  address: <http://ec2-14-200-128-5.compute-1.amazonaws.com:13810|ec2-14-200-128-5.compute-1.amazonaws.com:13810>
  db: 0
  password: p8b998511b4ad083b6```


Or set the hostname with the `REDIS_HOST` environment variable and the password with `REDIS_PASSWORD`