Speaker
Description
The availability of historical data on a complex control system
is essential to understanding it and aids greatly in being able
to resolve issues. The process of collecting and storing
these data as well as visualization is collectively
known as Monitoring. This paradigm is often extended with
event conditions and autonomous or manual actions based on these
events in a process called Alarming. Both are often realized
by operating on time-series data retrieved using time range
queries. These data can be served by specialized databases such
as Prometheus, providing efficient and high throughput data access.
But how could such a database be integrated into the Tango controls
ecosystem?
In this presentation we demonstrate the device_metrics
decorator. A single Python class decorator to automatically export Tango
device attributes to a Prometheus database. The decorator
automatically handles the conversion between data dimensionality
and types as well as offering control over included attributes.
To illustrate we will show some of the inner workings of the decorator
as well as the technical challenges that we had to address.
