Clarification and use of Utilization

[gsindigi]

Utilization is defined both at /system/openconfig-system-utilization level and /platform/openconfig-platform-common. Few questions while looking at these

1. Are there any guidelines on usage front of these?
   a. Is it required to configure both?
   b. Is it that the system level utilization takes precedence over /components/component/*/utilization/resources/resource/name?
2. The way it is defined today helps to track over-utilization by having used-threshold-upper/-clear. Are there anything similar to track under utilization as well? E.g., decreasing fan speed beyond a lower threshold may need an attention and a corresponding lower-clear value to clear the state.
3. Few components e.g., chassis, linecard, integrated-circuit have this utilization/resources and in /system/utilization/resources/resource has active-component-list. How to relate these?

[dplore]

The /system/utilization tree is used to configure system-wide resource thresholds. There's a short discussion with links to references to example implemenation documentation of this functionality at #817

The resources which support system wide threshold configuration are system defined. We expect to see thresholds across all INTEGRATED_CIRCUIT, LINECARD and CHASSIS components in a system (as defined in the description for the container)

The /components/component/*/utilization/resources/resource/ tree allows configuring thresholds for individual components.

We didn't specify inheritance rules in these leaves in the documentation, but I would expect the per component threshold to override system level thresholds.

We have this OpenConfig featureprofile test which covers utilization for INTEGRATED_CIRCUIT components. It only covers configuration of the /system thresholds and state at the /components level.

Regarding low thresholds, these could be added via new pull requests.

Hope this helps clarify! Feel free to reopen if you have more questions.