Monitor

The Mender Monitor add-on gives you instant visibility into issues with your device fleet. In addition, it allows you to configure alerts when certain parts of your infrastructure malfunction or need your attention.

To save bandwidth and support edge processing, Monitor is a fully distributed solution, meaning that we send no data from the device to the server unless triggered by an alert. Compared to classic approaches of sending all the data about all the nodes to the server all the time, this leads to massive bandwidth savings and allows instant local processing.

Alerts

Each time a service is not running, or a log file contains a given pattern, or a signal is received on a given D-Bus bus, all users who have access to the given device on the Mender server receive an email notification. You can mute the email notifications in the Mender Server settings.

There is one exception to the above rule: if a service goes up and down often enough we consider it to be flapping. When we detect this, service subsystem sends a flapping alert and does not send anything else until the number of service state changes per one FLAPPING_INTERVAL goes below FLAPPING_COUNT_THRESHOLD. See the advanced configuration chapter for more information about these settings.

Architecture

The Monitor add-on consists of two parts: the API to the backend that allows you to integrate with existing solutions, and a mender-monitor systemd service which runs on the devices. When certain events occur, the service sends alerts to the Mender Server. If you need more control, or you want to customize your monitoring solution, you can use the set of bash functions that compose the service directly from your scripts.

Running and configuring

To start monitoring daemon, run the following command:

systemctl start mender-monitor

You can view the service's logs running the following command:

journalctl -u mender-monitor