Monitoring MQTT Brokers

An MQTT broker severs as a hub for all MQTT messages and so it is important that it is always online and if not that you are ware of this.

There are several ways of monitoring the broker.

  • Using the Ping Command
  • Using the $SYS topic
  • Sending messages through it.
  • Monitoring Connection Status

Using the Ping Command

This very easy to implement as all machines support the ping command.

The ping command can be run manually or automatically.

The problem with it is that it doesn’t really monitor the MQTT broker but the machine that it is running on.

So if you were relying on this method if someone stopped the broker the pin command will not detect it as the machine is still up.

broker-monitor-ping

Using the $SYS Topic

Almost all broker publish broker statistics on the $SYS topic.

These are usually published at 5 second intervals but with the retain flag set.

If the broker is goes down then no statistics are published and so you will need to detect this.

Because only changes are published you need to monitor a topic that changes frequently.

However the $SYS topic can be disabled and so will not be available.

broker-monitor-sys

Flow Download

download

Broker Dashboard

This is a variation of the broker  monitor and has a chart embedded. It is meant as a starter template for your own broker dashboard.

download

Sending and Receiving Messages

This IMO is the better method as you can monitor multiple instances separately.

Many MQTT brokers run MQTT over multiple ports to support SSL and Websockets.

In addition MQTT hosting providers run multiple instances of MQTT on a single machine.

A broker can be running

  • Plain MQTT
  • MQTT over SSL
  • MQTT over websockets
  • MQTT over websockets with SSL
  • MQTT on Other Ports

The technique is to send a message to each instance and to monitor the received message.

broker-monitor-send-receive

Monitoring Connection Status

I added this method after posting a video on monitoring.

All MQTT client nodess will report a connection failure which you can detect using the status node, and so you can use this to trigger an alert.

However personally I haven’t found this method fool proof which is why I didn’t include it initially. It is however probably the easiest to implement.

Videos

Node-Red MQTT Broker Monitor Using Ping

download

 

Monitor MQTT Broker Status and Delay Using Node-Red

Flow Download

download

 

The message latency is the time taken for a message to be received from when it was published.

To measure the latency we simply publish a message and embed the publish time stamp and on receiving the message we can calculate the round trip time (message latency).

mqtt-broker-latency

Factors Affecting Latency

There are several factors that affect the message latency the main ones are:

  • Broker Load
  • Network load and type
  • Message size
  • QOS level

Node-Red Test Tool

The node-red test tool  allows you to compare the latency for different QOS levels and message sizes. The results are displayed in a graph and a table which is shown below:

mqtt-latency-chart
They are also logged to a database. To log to the database you will need to set the database path in the sqlite node and also create the database table using the inject node as shown below.

Logging to a database should be disabled by default.You can enable logging using the switch. A database error will be caught in the catch node and will disable logging.

Flow Download

download

Related Tutorials and Resources:

 

Please rate? And use Comments to let me know more

Leave a Reply

Your email address will not be published. Required fields are marked *