Monitoring

Operators like to have dashboards showing live key metrics of the system, which show them the overall health of the application in one glance. These metrics can be non-functional metrics such as memory and CPU usage, number of crashes of a system or application component, health of a node, and so on, as well as functional and thus application-specific metrics such as the number of checkouts in an ordering system or the number of items out of stock in an inventory service.

Most often, the base data used to aggregate the numbers that are used for a dashboard are extracted from logging information. This can either be system logs, which will mostly be used for non-functional metrics, and application-level logs, for functional metrics.