Collect metrics on device performance
Utilities like systeminformation can collect device performance metrics like CPU load, network throughput, and available storage. The System Metrics block uses systeminformation to forward metrics to the balenaBlock ecosystem via MQTT. From there you can graph the data or forward it to the cloud with one of the other available blocks.
Alternatively, when available resources are constrained, the block may be run as a standalone service. See the MQTT_ADDRESS variable in the Configuration section below.
We will use a docker-compose example script that uses the system metrics block to collect CPU load and temperature and available memory, and then forwards it to a separate MQTT broker container. First create a multi-container fleet in balenaCloud and provision a device with balenaOS. See the online docs for details.
Next push the docker-compose script to the balena builders, substituting your fleet's name for <myFleet> in the commands below.
git clone https://github.com/balena-io-examples/system-metrics.git
cd system-metrics/doc/simple-example
balena push <myFleet>
You should see data flowing to the MQTT broker like the log output below.
Published msg: {'short_uuid': 04166f8, "currentLoad":0.8995528642244212,"cpuTemperature":32.9,"mem":4762161152}
Published msg: {'short_uuid': 04166f8, "currentLoad":0.5756115873115185,"cpuTemperature":32.9,"mem":4762664960}
Environment variables you may configure are listed in the sections below. Variables may be defined as balena Fleet variables or Device variables.
List of metrics to collect, comma separated. A metric is a function name from the systeminformation project, for example cpuTemperature. Defaults to currentLoad, cpuTemperature, mem. Each metric must be specified in one of these forms:
| Form | Example | Notes |
|---|---|---|
| metric/aspect | cpuTemperature/max |
An aspect is a particular result object for the metric as shown in the documentation. |
| metric | cpuTemperature |
Uses the default aspect; for cpuTemperature this aspect is main. See list of metrics with a default aspect. |
| metric/(function parameter) | networkStats/(wlp0s20f3) |
Some systeminformation metric functions accept parameters to qualify the source of reported data. In the example, wlp0s20f3 is the WiFi interface. The function parameter applies to all requested aspects for the metric. |
Interval between metrics readings, in seconds. Defaults to 10. The container also initially waits 10 seconds before the first reading.
Alternatively, you may define the reading interval in milliseconds with the variable READING_INTERVAL_MS.
Runtime logging level, one of error, warn, info, debug, trace; defaults to info.
This variable is required for services to find the MQTT broker. Usually it is defined in the docker-compose.yml file, like in the example, so you don't need to explicitly define it for a fleet. However, you also may run System Metrics as a standalone block that outputs to the system log. In this case either don't define MQTT_ADDRESS in your docker-compose.yml, or override it to be blank for the fleet or device.
We also have created a docker-compose script that integrates with the balenaSense application. You should be able to simply push that script like the example above to see the data graphically.