41e927622e
This borrows heavily from the compute notifications code. The basic idea is that there are few classes of samples that can be had (Fan, Temperature, Voltage, Current). Within these are potentially multiple individual sensors and values. In all cases the volume and unit of a sample can always be extracted from the value of a 'Sensor Reading' key. The volume is cast to a float. If the value is 'Disabled' the data is dropped. Each subclass exists to allow flexible handling via stevedore, but all activity is performed in the superclass. The supplied good test data is the currently blessed sample from the Ironic team. The bad test data tickles either skipping missing data or producing warnings in the log. Identifier and name construction is worth a close look because this is an area of frequent discussion: * resource_id is composed from node_uuid and sensor id * meter name is based on the class name and is generic, only paired with a resource_id does it identify a specific sensor * unit is parsed out of sensor data and mapped to SI units Change-Id: Ie7ac92aebb45fc23d6134b4d6693f8a1e4d4ea40 Implements blueprint ironic-notifications |
||
|---|---|---|
| .. | ||
| __init__.py | ||
| ipmi.py | ||