No communication between local server and Asterisk in container

[gtjoseph]

@gtjoseph Could the Asterisk version be a plausible cause? I looked through the changelog but couldn't find any ARI related breaking changes. In my case, everything seems to be fine but the audio server running locally (host) just doesn't receive any packages from Asterisk 20 running within Docker.

Originally posted by @jneuendorf-i4h in #11 (comment)

[gtjoseph]

Can you run Asterisk on the host just as a test and see if it works?
What kind of networking are you using with docker?
Can you connect to asterisk's built-in web server from a browser on the host?

[jneuendorf-i4h]

First of all, thanks for the quick response! 🙏

Can you run Asterisk on the host just as a test and see if it works?

My setup is on macOS and I am not sure if Asterisk can be installed on macOS as easily as on Debian/Ubuntu.
Therefore, I have not tried this approach yet.

What kind of networking are you using with docker?

So far, I stuck with the defaults. With network_mode: host (docker-compose.yml) phone calls no longer worked.

Can you connect to asterisk's built-in web server from a browser on the host?

1. Yes, on my 1st attempt, I started the ARI client on the host and the connection to the REST API worked (authentication, channel creation etc.)
2. 2nd attempt: Because I suspected networking problems due to Docker, I started the ARI client from within docker (so everything on localhost/127.0.0.1 within Docker. But the behavior stayed the same. 🤔

---

I actually simplified/adjusted this repo's code for the start, as I want to initiate a call to an SIP device (PJSIP/6001) and stream the audio from/to the audio server (without any processing for now).

Any advice or hint would be very helpful, so let me know if anything is unclear or if there is information missing.
Also, debugging tips are very welcome (so far, I've only used ari set debug all on). 😉

For better insights, here my code:

Client

src/upd.mjs

Start the audio server in a separate script to reduce complexity in the main script (for debugging).

import { RtpUdpServerSocket } from "../lib/rtp-udp-server.js"

const audioServer = new RtpUdpServerSocket(
    '0.0.0.0:9999',
    false,
    "./test.wav",
)
console.log(audioServer)

src/send.mjs

Send a test message to the audio server.

import dgram from 'node:dgram'
import { Buffer } from 'node:buffer'

const message = Buffer.from('A few bytes to send for testing purposes...')
const client = dgram.createSocket('udp4')
client.send(message, 9999, 'localhost', (err) => {
    console.log('error', err)
    client.close()
})

src/main.mjs

Start the ARI application, setup the channels and bridge, and make the call.

import ari from 'ari-client'

const options = {
    dialstring: "PJSIP/6001",
    format: "ulaw",
    listenServer: '127.0.0.1:9999',
    stasisApp: "external-media",
}

ari.connect(
    'http://127.0.0.1:8088',
    'asterisk',
    'asterisk',
    async (error, client) => {
        if (error) {
            console.error(error)
            console.warn("CLOSING")
        }

        await client.start(options.stasisApp)

        // Create a simple mixing bridge that is controlled by ARI/Stasis
        const mixingBridge = client.Bridge()
        try {
            await mixingBridge.create({type: "mixing"})
            console.log("created mixing bridge")
        } catch(error) {
            console.error(error)
            console.warn("CLOSING")
        }
        mixingBridge.on('BridgeDestroyed', (event) => {
            console.log("destroyed mixing bridge")
            console.warn("CLOSING")
        })

        const outboundChannel = client.Channel()
        outboundChannel.on('StasisStart', (event, {id}) => {
            mixingBridge.addChannel({channel: id})
            console.log("added outbound channel to mixing bridge", id)
        })
        outboundChannel.on('StasisEnd', (event, chan) => {
            console.warn("CLOSING")
        })

        // Call the phone or confbridge specified in dialstring
        try {
            await outboundChannel.originate({
                endpoint: options.dialstring,
                formats: options.format,
                app: options.stasisApp,
            })
            // localChannel.dial()
        } catch (error) {
            console.error(error)
            console.warn("CLOSING")
        }

        // Now we create the External Media channel.
        const externalChannel = client.Channel()
        externalChannel.on('StasisStart', (event, {id}) => {
            mixingBridge.addChannel({channel: id})
            console.log("added external channel to mixing bridge", id)
        })
        externalChannel.on('StasisEnd', (event, chan) => {
            console.warn("CLOSING")
        })

        /*
         * We give the external channel the address of the listener
         * we already set up and the format it should stream in.
         */
        const externalMediaChannel = await externalChannel.externalMedia({
            app: options.stasisApp,
            // external_host: options.listenServer,
            external_host: "127.0.0.1:9999",
            format: options.format,
        })
        console.log("externalMediaChannel", externalMediaChannel)

        setTimeout(async () => {
            console.warn("STOPPING")
            outboundChannel && await outboundChannel.hangup()
            externalChannel && await externalChannel.hangup()
            mixingBridge && await mixingBridge.destroy()
            client.stop()
        }, 30 * 1000);
})

Docker

docker-compose.yml

version: '3'
services:
  asterisk:
    image: asterisk
    build:
      context: .
      dockerfile: docker/asterisk/Dockerfile
    # network_mode: host
    ports:
      - "8088:8088"                   # ARI HTTP
      - "5060:5060/udp"               # SIP UDP
      - "5060:5060"                   # SIP TCP
      - "5061:5061"                   # SIP TLS
      - "10000-10099:10000-10099/udp" # RTP
      - "9999:9999/udp" # test stream from client on host to audio server in Docker

Server

ari.conf

[general]
enabled = yes
pretty = yes
allowed_origins = *

[asterisk]
type = user
read_only = no
password = asterisk

http.conf

[general]
enabled = yes
bindaddr = 0.0.0.0
bindport = 8088

[asterisk]
type = user
read_only = no
password = asterisk

rtp.conf

[general]
rtpstart = 10000
rtpend   = 10099

strictrtp = no

extensions.conf

[from-internal]
exten => 100,1,Answer()
 same => n,Stasis(external-media)
 same => n,Hangup()

pjsip.conf

[transport-udp]
type=transport
protocol=udp
bind=0.0.0.0

[6001]
type=endpoint
context=from-internal
disallow=all
allow=ulaw
auth=6001
aors=6001

[6001]
type=auth
auth_type=userpass
password=unsecurepassword
username=6001

[6001]
type=aor
max_contacts=1

[mattbonnell]

seeing the same thing here; listening on port 9999 with netcat but doesn't seem anything is coming through.

[vlxdisluv]

have you solved you problem? @jneuendorf-i4h

[jneuendorf-i4h]

Thanks for catching up @vlxdisluv! The currently working solution is running Asterisk not in a Docker container but in a (Proxmox) VM (LXC untested).

A reasonable cause is network_mode: host on macOS:

The host networking driver only works on Linux hosts, and is not supported on Docker Desktop for Mac
https://docs.docker.com/network/drivers/host/

So I guess, the issue can be closed, even though it would be helpful to have documentation for running Asterisk in Docker (on specific machines/operating systems).

[vlxdisluv]

@jneuendorf-i4h I have successfully configured Asterisk in Docker container (macOS m1 arm) because it can establish a peer-to-peer connection (between two softphones), but I don't have enough knowledge to write a proper Asterisk configuration for a workflow with external media. I need to initiate an outgoing ARI call to a softphone and receive real-time RTP packets and then send them to some STT provider. In this case, I also established a connection, but I cannot receive RTP packets. Could you share your working Asterisk configs?

[jneuendorf-i4h]

@jneuendorf-i4h I have successfully configured Asterisk in Docker container (macOS m1 arm) because it can establish a peer-to-peer connection (between two softphones), but I don't have enough knowledge to write a proper Asterisk configuration for a workflow with external media. I need to initiate an outgoing ARI call to a softphone and receive real-time RTP packets and then send them to some STT provider. In this case, I also established a connection, but I cannot receive RTP packets. Could you share your working Asterisk configs?

Have you checked the audio signal of your p2p/softphone connection? I would not be surprised if the connection can be established while not receiving RTP packets. RTP debugging can be done via several different approaches, e.g.

• pjsip set logger on
• rtp set debug on
• tcpdump -i <INTERFACE> -w dump.pcap + Wireshark (open file, telephony > VOIP calls: shows all call legs > flow)

On the JS client side, you could try (using the nomenclature from above)

setInterval(async () => {
    const rtpStats = await outboundChannel.rtpstatistics()
    console.log("RTP", rtpStats)
}, 5000)

[phamkhactu]

Hi @jneuendorf-i4h,

I am newbie in this field, I work as AI engineer. Now I have task to get audio from asterisk to control speech to text, and then generative answer and generate audio from text to speech and then send back to asterisk through anyway. After researching, I think that there are many methods to approach to send audio from calling to external, but I focus on 3 way:

1. AGI
2. ARI
3. Softphone
   I have read, but I am confuse, I have some question. I am very happy if you can support me. In my case, I have asterisk container in VMware and local machine which contains VMware. I want to send audio from VMware to local machine.

First question: I know AGI, need to write script in asterisk container, and then using socket, ..etc to send audio to external API(local machine). What is the difference between ARI? ARI need write script or not? I asked in GPT, it gave me a script need to write in container VMware, and then using ws to send it, I saw the other thing difference to AGI is that it have a file audio 1s about calling sip directly writes streaming rtp byte(write header only first in it and then rewrite byte audio in this audio) and then send to external. Does it right? If not, it means directly to external? If can, Would you can mind giving me document or youtobe to setup it if you know about it?

Second question: Could you give me some advance for my case? which method should I use?

Third question: I want to send audio back to play for calling, how I config it.?

Thank you very much!!

[jneuendorf-i4h]

Remark: Since I work for a company I can't share the entire setup or code.

---

@phamkhactu I haven't continued trying to solve the container problem. Instead, I run the Asterisk outside of Docker as a workaround. This seems to work from within VMs because it is working on VPS or Cloud instances (without NAT).

It's still ARI based and I don't know anything about AGI 😅 The script is very much based on this repository but I am sending the data to an own media server (written in JavaScript, based on node:dgram's Socket). I think each sent message contains header which I wrote an RtpPacket class for based on

• https://github.com/1995parham/krtp and
• https://datatracker.ietf.org/doc/html/rfc3550#section-5.1

I use the socket to send audio back:

const socket = dgram.createSocket("udp4")
const packet = new RtpPacket(
    payload,
    sequenceNumber,
    ssrc,
    timestamp,
    payloadType,
)
const chunk = packet.serialize()
socket.send(chunk, port, address, callback)

When sending multiple packets, I found it necessary to use a throttling of about 20ms.

[phamkhactu]

@jneuendorf-i4h, with your sharing, I believe that in my case the same with you (I set Vmware asterisk DHCP not NAT, have conntect between asterisk and local machine, I telnet successfully from local to asterisk)

And your sharing above, it means that I don't need to write script in asterisk, only setup configure as you share above, and then you WS get audio(rtp). Does it right? I will try it. I am very happy If you give me code example to get audio from asterisk using ARI? I'm using python, but having conflict version, I am stuck in it, I don't have much experience in js.

Many thanks.