Post by sgaree on Sept 11, 2018 17:02:33 GMT
Is it expected that HealthRollup of a property is always the most severe value of the Health and HealthRollup values of all subordinate resources and the Health of a resource always only reflects the resource itself with no impact of subordinate resources (e.g. HealthRollup never affects superior Health properties?)
Or is it left to the implementer to determine how the HealthRollup values are transitioned up the containment hierarchy?
Take the following examples of a server consisting of a Chassis, System, and Power Supply.
Are the two corresponding tables of property values correct for the expected implementation of health rollup?
Is this just one possible implementation?
Example 1. A power supply has failed, so it has a Health of Critical. It has no subordinate resources, so its HealthRollup is OK.
The System itself has no detected faults, so its Health is OK.
The Chassis as well has no health issues, so Health is OK.
Example 2. A power supply has a warning.
The System has a BIOS internal error, resulting in a Critical Health.
The Chassis again has no health issues.
1.
2.
Let me expand on the first example to clarify why I ask about implementation flexibility.
Assume the server has N+2 power supply redundancy. I won't bring a Redundancy resource into the discussion, though it might be the mechanism through which this example would be logically implemented.
While the PSU health is certainly Critical (it's dead Jim), the admin might well consider the System to only warrant a Warning state.
This generally comes up when discussing what happens when health statuses go Warning or Critical as it relates to waking someone up in the middle of the night, sounding alarm claxons, etc. Could the following be a valid implementation of Example 1?
TIA,
-Scott
Or is it left to the implementer to determine how the HealthRollup values are transitioned up the containment hierarchy?
Take the following examples of a server consisting of a Chassis, System, and Power Supply.
Are the two corresponding tables of property values correct for the expected implementation of health rollup?
Is this just one possible implementation?
Example 1. A power supply has failed, so it has a Health of Critical. It has no subordinate resources, so its HealthRollup is OK.
The System itself has no detected faults, so its Health is OK.
The Chassis as well has no health issues, so Health is OK.
Example 2. A power supply has a warning.
The System has a BIOS internal error, resulting in a Critical Health.
The Chassis again has no health issues.
1.
| Resource | Health | HealthRollup |
| PSU | Critical | OK |
| System | OK | Critical |
| Chassis | OK | Critical |
2.
| Resource | Health | HealthRollup |
| PSU | Warning | OK |
| System | Critical | Warning |
| Chassis | OK | Critical |
Let me expand on the first example to clarify why I ask about implementation flexibility.
Assume the server has N+2 power supply redundancy. I won't bring a Redundancy resource into the discussion, though it might be the mechanism through which this example would be logically implemented.
While the PSU health is certainly Critical (it's dead Jim), the admin might well consider the System to only warrant a Warning state.
This generally comes up when discussing what happens when health statuses go Warning or Critical as it relates to waking someone up in the middle of the night, sounding alarm claxons, etc. Could the following be a valid implementation of Example 1?
| Resource | Health | HealthRollup |
| PSU | Critical | OK |
| System | OK | Warning |
| Chassis | OK | Warning |
TIA,
-Scott
