Skip to contentSkip to conversation

Obtaining fan/temperature sensor status for Redfish

DISCUSSION Shawn McCarney 3 participants 3 messages Apr 25, 2019 – Jul 17, 2019

lore
Share & tools
lore
#0 lore raw

Hi,

The Redfish Thermal URI (/redfish/v1/Chassis/{ChassisId}/Thermal) returns all the fan and temperature sensors within a chassis.

Each returned fan/temperature sensor contains a Status object with the following properties:
* Health: Health state of the resource (OK, Warning, Critical)
* State: State of the resource (Absent, Disabled, Enabled, etc.)

Currently bmcweb always returns a Health value of "OK" and a State value of"Enabled".

I would like to enhance bmcweb to return more accurate values in the Status object when possible.  Below I have summarized a proposed design.  Does this sound acceptable?

== Proposed Design ==

* Initially assume Redfish Status->Health value of "OK" and Status->State value of "Enabled" as it does today.
* Use ObjectMapper associations to find the DBus inventory object path that is related to the DBus sensor (if available).
* Get Functional property of xyz.openbmc_project.State.Decorator.OperationalStatus interface for inventory object path (if available).
* Set Redfish Status->Health to "Critical" if DBus Functional property is false.
* Get Present property of xyz.openbmc_project.Inventory.Item interface for inventory object path (if available).
* Set Redfish Status->State to "Absent" if DBus Present property is false.  Also remove the Redfish Status->Health property from the response if Present is false.

== Notes ==

From the Redfish documentation I read, it was unclear whether the Redfish Status object was intended to describe the sensor itself (such as a fan controller chip) or the hardware being monitored by the sensor (such as a fan).

It was also unclear what the value of the Status->Health property should be for hardware that is not present.

I posted these questions to the Redfish forum:
* https://redfishforum.com/thread/187/intended-meaning-status-object-temps

Their responses were:
* Status is for the hardware being monitored, not the sensor itself.
* Status->Health should probably not be returned if hardware is not present.

Thanks,

Shawn McCarney

#1 lore raw
Quoted 10 lines · Shawn McCarney

Hi,

The Redfish Thermal URI (/redfish/v1/Chassis/{ChassisId}/Thermal) returns all the fan and temperature sensors within a chassis.

Each returned fan/temperature sensor contains a Status object with the following properties:
* Health: Health state of the resource (OK, Warning, Critical)
* State: State of the resource (Absent, Disabled, Enabled, etc.)

Currently bmcweb always returns a Health value of "OK" and a State value of"Enabled".

Quoted 8 lines

I would like to enhance bmcweb to return more accurate values in the Status object when possible.  Below I have summarized a proposed design.  Does this sound acceptable?

== Proposed Design ==

* Initially assume Redfish Status->Health value of "OK" and Status->State value of "Enabled" as it does today.
* Use ObjectMapper associations to find the DBus inventory object path that is related to the DBus sensor (if available).
* Get Functional property of xyz.openbmc_project.State.Decorator.OperationalStatus interface for inventory object path (if available).
* Set Redfish Status->Health to "Critical" if DBus Functional property is false.

My understanding was that 'Functional' property is for deactivation/activation of components in context of both health checking and hotplugging.

That is, I had an impression that it's the property that can be set by either the operator or the system software, so it's not necessarily a health status.

Correct me if I'm wrong, but to me it's more like it should map to State Enabled/Disabled without any reason implications. I think that health status must come from another source than the dbus 'Functional' property.

The rest sounds fine to me.

With best regards,
Alexander Amelkin,
Leading BMC Software Engineer, YADRO
https://yadro.com

#2 lore raw
Quoted 1 line · Alexander Amelkin

can be set by either the operator or the system software

Can you elaborate on what made you think the operator could set operational statuses?