A Channels 3/4 compatible MQTT worker
This worker is a standard Channels Consumer which contains a fully async MQTT server allowing channels messages to be used to publish, subscribe to and receive MQTT messages
The proxy understands mqtt.subscribe and mqtt.publish so you can change the topics dynamically.
When an MQTT subscribe is performed it is done on behalf of a channels group and all Channels Consumers in that group will receive mqtt messages as mqtt.message
The overview is :
MQTT <> Channels-MQTT-Proxy (in runworker) <> Redis/Channels-layer <> ASGI applications (in Daphne/runserver) <> Websockets/HTTP <> Browser
pip install chanmqttproxy
In Channels the asgi application handles all types of connection routing. Websocket and http connections are listened for by a suitable server (eg daphne run by manage.py runserver) which will instantiate classes and run objects to handle them. The 'mqtt' Consumer does not accept incoming connections, just channel messages. So it must be started as a worker which handles the Channel messages; the MQTT client connection is then created inside the MqttConsumer worker when the first channel message is received.
The code may look familiar if you've used the Channels Chat tutorial :)
In fact it will add the ability to monitor the chat on the chat/ MQTT channel and messages sent to that channel will appear on all clients.
In site/asgi.py
:
from chanmqttproxy import MqttConsumer
from channels.routing import ChannelNameRouter, ProtocolTypeRouter
application = ProtocolTypeRouter({
"channel": ChannelNameRouter({
"mqtt": MqttConsumer.as_asgi()
}),
... # rest of http/websocket routes
})
To define the MQTT broker, in site/settings.py
:
# Local mqtt settings
MQTT_HOST = "mqtt.example.com"
MQTT_USER = "mqtt-test"
MQTT_PASSWORD = "mqtt-test"
MQTT_VERSION = 311 # defaults to 50
At this point you have a working async Channels/MQTT bridge
To subscribe to a topic in your AsyncConsumer
async def connect(self):
... # existing group_add() calls
# Join mqtt group
await self.channel_layer.group_add(
"mqttgroup",
self.channel_name
)
# Ensure MQTT messages come to the room
# This simplistic approach subscribes the room every
# time a websocket connects but that's OK
await self.channel_layer.send(
"mqtt",
{
"type": "mqtt_subscribe",
"topic": f"chat/{self.room_name}",
"group": "mqttgroup",
})
await self.accept() # existing accept() call
To handle messages from a topic in your AsyncConsumer
# Receive message from mqtt group and send to websocket
async def mqtt_message(self, event):
message = event['message']
topic = messagep["topic"]
payload = messagep["payload"]
# Send message to WebSocket
await self.send(text_data=json.dumps({
'message': payload
}))
To publish messages to a topic in your AsyncConsumer
# Receive message from WebSocket
async def receive(self, text_data):
... # existing group_send etc
# Publish on mqtt too
await self.channel_layer.send(
"mqtt",
{
"type": "mqtt_publish",
"publish": { # These form the kwargs for mqtt.publish
"topic": f"chat/{self.room_name}_out",
"payload": message,
"qos": 2,
"retain": False,
}
})
For debug logging I use this at the end of settings.py:
import logging.config
LOGGING_CONFIG = None
logging.config.dictConfig({
'version': 1,
'disable_existing_loggers': False,
'formatters': {
'console': {
# exact format is not important, this is the minimum information
'format': '%(asctime)s %(name)-12s %(levelname)-8s %(message)s',
},
},
'handlers': {
'console': {
'class': 'logging.StreamHandler',
'formatter': 'console',
},
},
'loggers': {
# root logger
'chanmqttproxy': {
'level': 'DEBUG',
'handlers': ['console'],
},
'mysite': {
'level': 'DEBUG',
'handlers': ['console'],
},
'chat': {
'level': 'DEBUG',
'handlers': ['console'],
},
},
})
The complete code for the Channels Chat tutorial application (up to part 3) with the channels-mqtt-proxy additions is here: https://github.com/lbt/channels-mqtt-proxy/tree/main/examples
Create and activate a suitable venv.
Enter the examples/
directory.
You'll need a Redis instance for Channels to work sensibly and an MQTT broker like mosquitto of course.
Setup Django/channels etc in your venv by running:
pip install ..[examples]
Then edit mysite/mysite/settings.py
to point to your mqtt server and run;
python3 mysite/manage.py migrate
python3 mysite/manage.py runworker mqtt &
python3 mysite/manage.py runserver
Navigate to http://127.0.0.1:8000/chat/ and use "lobby" for the room.
Use your mqtt listener to listen to the topic chat/lobby_out
and
publish to the topic chat/lobby
Notice that if we had just used chat/ for both topics then when the chat room sends a message and the proxy client publishes it to the MQTT topic the message would appear twice in the room. This is because even if you're the one that publishes a message, if you're subscribed to the topic, you will receive it too.
If you make changes to the code note that the Channels runworker does not auto-reload and will still hold old subscribe/group information.
The MqttConsumer is only instantiated after the first message arrives rather than when the worker starts. This means it may not be connected so the await self.mqttproxy.connected.wait() is required on every subscribe/publish (which is not a lot of overhead but...
It's not clear how to issue an unsubscribe or deal with all clients disconnecting. If this is done in the AsyncConsumer disconnect() then it needs a client-count which is probably unreliable. Currently MQTT messages will always be sent to the Channels group and it handles member timeout as per https://channels.readthedocs.io/en/stable/channel_layer_spec.html#capacity and https://channels.readthedocs.io/en/stable/channel_layer_spec.html#persistence
On initial subscribtion all retained messages are dropped. This is not ideal when retained messages are used to indicate 'last known state' for clients.
However this is an MQTT concept and doesn't carry over to Channels unless we handle retention and continue to store messages for each Channel client that subscribes - and then somehow only send retained messages to that new client and not the existing clients who've seen them once..
Ideally there would be a mechanism for the app to pre-subscribe and send retained (and new) messages to code that could update the 'current state' model which would then be maintained and used to initialise new browser clients.
There doesn't seem to be a way to tell the worker to exit on Ctrl-C if we trap it to clean up the MQTT connections. Also note that in some situations the Ctrl-C fails. Eg if the broker doesn't support V5.0 and fallback to V311 is underway.
Thanks to Gurtam for https://github.com/wialon/gmqtt which is a great asyncio MQTT client that I use extensively in my HA systems and integrate with PyQt too.
The Channels tutorial was really helpful in understanding the concepts.
Also to Xavier Lesa for https://github.com/xavierlesa/channels-asgi-mqtt which is based on the paho-mqtt synchronous library and inspired me to write this.