Penguin Computing Relion 1900e, Technical Manual

Page: 87 / 174

Relion 1900e/2900e Manual
Revision 1.0 72
the BMC FW to halt updates to the value of the associated fan tach sensor and set that sensor’s IPMI sensor
state to “reading-state-unavailable” when this mode is active. Management software must comprehend this
state for fan tach sensors and not report these as failure conditions.
The scenario for which this occurs is that the BMC Fan Speed Control (FSC) code turns off the fans by setting
the PWM for the domain to 0. This is done when based on one or more global aggregate thermal margin
sensor readings dropping below a specified threshold.
By default the fans-off feature will be disabled. There is a BMC command and BIOS setup option to
enable/disable this feature.
The SmaRT/CLST system feature will also momentarily gate power to all the system fans to reduce overall
system power consumption in response to a power supply event (for example, to ride out an AC power
glitch). However, for this scenario, the fan power is gated by HW for only 100ms, which should not be long
enough to result in triggering a fan fault SEL event.
7.3.14 Standard Fan Management
The BMC controls and monitors the system fans. Each fan is associated with a fan speed sensor that detects
fan failure and may also be associated with a fan presence sensor for hot-swap support. For redundant fan
configurations, the fan failure and presence status determines the fan redundancy sensor state.
The system fans are divided into fan domains, each of which has a separate fan speed control signal and a
separate configurable fan control policy. A fan domain can have a set of temperature and fan sensors
associated with it. These are used to determine the current fan domain state.
A fan domain has three states:
• The sleep and boost states have fixed (but configurable through OEM SDRs) fan speeds associated
with them.
•
The nominal state has a variable speed determined by the fan domain policy. An OEM SDR record is
used to configure the fan domain policy.
The fan domain state is controlled by several factors. They are listed below in order of precedence, high to
low:

Boost
o
Associated fan is in a critical state or missing. The SDR describes which fan domains are boosted
in response to a fan failure or removal in each domain. If a fan is removed when the system is in
‘Fans-off’ mode it will not be detected and there will not be any fan boost till system comes out of
‘Fans-off; mode.
o
Any associated temperature sensor is in a critical state. The SDR describes which temperature
threshold violations cause fan boost for each fan domain.
o
The BMC is in firmware update mode, or the operational firmware is corrupted.
o
If any of the above conditions apply, the fans are set to a fixed boost state speed.

Nominal
o
A fan domain’s nominal fan speed can be configured as static (fixed value) or controlled by the
state of one or more associated temperature sensors.
7.3.14.1 Hot-Swap Fans
Hot-swap fans are supported. These fans can be removed and replaced while the system is powered on and
operating. The BMC implements fan presence sensors for each hot-swappable fan.