Penguin Computing Relion 1900e, Техническое руководство

Страница: 86 / 174

71 Revision 1.3
Relion 1900e/2900e Manual
As a result, fan tach sensors do not auto-rearm when the fault condition goes away but rather are rearmed
for either of the following occurrences:
a. The system is reset or power-cycled.
b. The fan is removed and either replaced with another fan or re-inserted. This applies to hot-
swappable fans only. This re-arm is triggered by change in the state of the associated fan
presence sensor.
After the sensor is rearmed, if the fan speed is detected to be in a normal range, the failure conditions shall
be cleared and a de-assertion event shall be logged.
7.3.13.2 Fan Presence Sensors
Some chassis and server boards provide support for hot-swap fans. These fans can be removed and
replaced while the system is powered on and operating normally. The BMC implements fan presence
sensors for each hot swappable fan. These are instantiated as IPMI discrete sensors.
Events are only logged for fan presence upon changes in the presence state after AC power is applied (no
events logged for initial state).
7.3.13.3 Fan Redundancy Sensor
The BMC supports redundant fan monitoring and implements fan redundancy sensors for products that
have redundant fans. Support for redundant fans is chassis-specific.
A fan redundancy sensor generates events when its associated set of fans transition between redundant and
non-redundant states, as determined by the number and health of the component fans. The definition of fan
redundancy is configuration dependent. The BMC allows redundancy to be configured on a per fan-
redundancy sensor basis through OEM SDR records.
There is a fan redundancy sensor implemented for each redundant group of fans in the system.
Assertion and de-assertion event generation is enabled for each redundancy state.
7.3.13.4 Power Supply Fan Sensors
Monitoring is implemented through IPMI discrete sensors, one for each power supply fan. The BMC polls
each installed power supply using the PMBus* fan status commands to check for failure conditions for the
power supply fans. The BMC asserts the “performance lags” offset of the IPMI sensor if a fan failure is
detected.
Power supply fan sensors are implemented as manual re-arm sensors because a failure condition can result
in boosting of the fans. This in turn may cause a failing fan’s speed to rise above the “fault” threshold and can
result in fan oscillations. As a result, these sensors do not auto-rearm when the fault condition goes away but
rather are rearmed only when the system is reset or power-cycled, or the PSU is removed and replaced with
the same or another PSU.
After the sensor is rearmed, if the fan is no longer showing a failed state, the failure condition in the IPMI
sensor shall be cleared and a de-assertion event shall be logged.
7.3.13.5 Monitoring for “Fans Off” Scenario
On Intel ® Server Systems supporting the Intel ® Xeon ® processor E5-2600 v3, v4 product family, it is likely that
there will be situations where specific fans are turned off based on current system conditions. BMC Fan
monitoring will comprehend this scenario and not log false failure events. The recommended method is for