Improved PR

This commit is contained in:
Tigerblue77 2023-02-04 11:51:18 +00:00
parent 78c0e59562
commit aa954d4ace

View File

@ -2,28 +2,28 @@
# Define global functions
# This function applies Dell's default dynamic fan control profile
function apply_Dell_profile () {
function apply_Dell_fan_control_profile () {
# Use ipmitool to send the raw command to set fan control to Dell default
ipmitool -I $LOGIN_STRING raw 0x30 0x30 0x01 0x01 > /dev/null
ipmitool -I $IDRAC_LOGIN_STRING raw 0x30 0x30 0x01 0x01 > /dev/null
CURRENT_FAN_CONTROL_PROFILE="Dell default dynamic fan control profile"
}
# This function applies a user-specified static fan control profile
function apply_user_profile () {
function apply_user_fan_control_profile () {
# Use ipmitool to send the raw command to set fan control to user-specified value
ipmitool -I $LOGIN_STRING raw 0x30 0x30 0x01 0x00 > /dev/null
ipmitool -I $LOGIN_STRING raw 0x30 0x30 0x02 0xff $HEXADECIMAL_FAN_SPEED > /dev/null
ipmitool -I $IDRAC_LOGIN_STRING raw 0x30 0x30 0x01 0x00 > /dev/null
ipmitool -I $IDRAC_LOGIN_STRING raw 0x30 0x30 0x02 0xff $HEXADECIMAL_FAN_SPEED > /dev/null
CURRENT_FAN_CONTROL_PROFILE="User static fan control profile ($DECIMAL_FAN_SPEED%)"
}
function enable_third_party_PCIe_card_Dell_default_cooling_response () {
# We could check the current cooling response before applying but it's not very useful so let's skip the test and apply directly
ipmitool -I $LOGIN_STRING raw 0x30 0xce 0x00 0x16 0x05 0x00 0x00 0x00 0x05 0x00 0x00 0x00 0x00 > /dev/null
ipmitool -I $IDRAC_LOGIN_STRING raw 0x30 0xce 0x00 0x16 0x05 0x00 0x00 0x00 0x05 0x00 0x00 0x00 0x00 > /dev/null
}
function disable_third_party_PCIe_card_Dell_default_cooling_response () {
# We could check the current cooling response before applying but it's not very useful so let's skip the test and apply directly
ipmitool -I $LOGIN_STRING raw 0x30 0xce 0x00 0x16 0x05 0x00 0x00 0x00 0x05 0x00 0x01 0x00 0x00 > /dev/null
ipmitool -I $IDRAC_LOGIN_STRING raw 0x30 0xce 0x00 0x16 0x05 0x00 0x00 0x00 0x05 0x00 0x01 0x00 0x00 > /dev/null
}
# Returns :
@ -31,7 +31,7 @@ function disable_third_party_PCIe_card_Dell_default_cooling_response () {
# - 1 if third-party PCIe card Dell default cooling response is currently ENABLED
# - 2 if the current status returned by ipmitool command output is unexpected
# function is_third_party_PCIe_card_Dell_default_cooling_response_disabled() {
# THIRD_PARTY_PCIE_CARD_COOLING_RESPONSE=$(ipmitool -I $LOGIN_STRING raw 0x30 0xce 0x01 0x16 0x05 0x00 0x00 0x00)
# THIRD_PARTY_PCIE_CARD_COOLING_RESPONSE=$(ipmitool -I $IDRAC_LOGIN_STRING raw 0x30 0xce 0x01 0x16 0x05 0x00 0x00 0x00)
# if [ "$THIRD_PARTY_PCIE_CARD_COOLING_RESPONSE" == "16 05 00 00 00 05 00 01 00 00" ]; then
# return 0
@ -45,7 +45,7 @@ function disable_third_party_PCIe_card_Dell_default_cooling_response () {
# Prepare traps in case of container exit
function gracefull_exit () {
apply_Dell_profile
apply_Dell_fan_control_profile
enable_third_party_PCIe_card_Dell_default_cooling_response
echo "/!\ WARNING /!\ Container stopped, Dell default dynamic fan control profile applied for safety."
exit 0
@ -56,7 +56,9 @@ trap 'gracefull_exit' SIGQUIT SIGKILL SIGTERM
# Prepare, format and define initial variables
# Check if FAN_SPEED variable is in hexadecimal format, if not convert it to hexadecimal
# readonly DELL_FRESH_AIR_COMPLIANCE=45
# Check if FAN_SPEED variable is in hexadecimal format. If not, convert it to hexadecimal
if [[ $FAN_SPEED == 0x* ]]
then
DECIMAL_FAN_SPEED=$(printf '%d' $FAN_SPEED)
@ -67,19 +69,19 @@ else
fi
# Log main informations given to the container
echo "Idrac/IPMI host: $IDRAC_HOST"
echo "iDRAC/IPMI host: $IDRAC_HOST"
# Check if the Idrac host is set to 'local', and set the LOGIN_STRING accordingly
# Check if the iDRAC host is set to 'local' or not then set the IDRAC_LOGIN_STRING accordingly
if [[ $IDRAC_HOST == "local" ]]
then
LOGIN_STRING='open'
IDRAC_LOGIN_STRING='open'
else
echo "Idrac/IPMI username: $IDRAC_USERNAME"
echo "Idrac/IPMI password: $IDRAC_PASSWORD"
LOGIN_STRING="lanplus -H $IDRAC_HOST -U $IDRAC_USERNAME -P $IDRAC_PASSWORD"
echo "iDRAC/IPMI username: $IDRAC_USERNAME"
echo "iDRAC/IPMI password: $IDRAC_PASSWORD"
IDRAC_LOGIN_STRING="lanplus -H $IDRAC_HOST -U $IDRAC_USERNAME -P $IDRAC_PASSWORD"
fi
# Log the fan speed objective, CPU temperature threshold, and check interval
# Log the fan speed objective, CPU temperature threshold and check interval
echo "Fan speed objective: $DECIMAL_FAN_SPEED%"
echo "CPU temperature treshold: $CPU_TEMPERATURE_TRESHOLD°C"
echo "Check interval: ${CHECK_INTERVAL}s"
@ -88,7 +90,8 @@ echo ""
# Define the interval for printing
readonly TABLE_HEADER_PRINT_INTERVAL=10
i=$TABLE_HEADER_PRINT_INTERVAL
IS_DELL_PROFILE_APPLIED=true
# Set the flag used to check if the active fan control profile has changed
IS_DELL_FAN_CONTROL_PROFILE_APPLIED=true
# Start monitoring
while true; do
@ -97,53 +100,56 @@ while true; do
SLEEP_PROCESS_PID=$!
# Retrieve sensor data using ipmitool
DATA=$(ipmitool -I $LOGIN_STRING sdr type temperature | grep degrees)
DATA=$(ipmitool -I $IDRAC_LOGIN_STRING sdr type temperature | grep degrees)
INLET_TEMPERATURE=$(echo "$DATA" | grep Inlet | grep -Po '\d{2}' | tail -1)
EXHAUST_TEMPERATURE=$(echo "$DATA" | grep Exhaust | grep -Po '\d{2}' | tail -1)
CPU_DATA=$(echo "$DATA" | grep "3\." | grep -Po '\d{2}')
CPU1_TEMPERATURE=$(echo $CPU_DATA | awk '{print $1;}')
CPU2_TEMPERATURE=$(echo $CPU_DATA | awk '{print $2;}')
# Define functions to check if CPU1 and CPU2 temperatures are above the threshold
CPU1_OVERHEAT () { [ $CPU1_TEMPERATURE -gt $CPU_TEMPERATURE_TRESHOLD ]; }
CPU2_OVERHEAT () { [ $CPU2_TEMPERATURE -gt $CPU_TEMPERATURE_TRESHOLD ]; }
# Define functions to check if CPU 1 and CPU 2 temperatures are above the threshold
function CPU1_OVERHEAT () { [ $CPU1_TEMPERATURE -gt $CPU_TEMPERATURE_TRESHOLD ]; }
function CPU2_OVERHEAT () { [ $CPU2_TEMPERATURE -gt $CPU_TEMPERATURE_TRESHOLD ]; }
# Initialize a variable to store comments
# Initialize a variable to store the comments displayed when the fan control profile changed
COMMENT=" -"
# Check if CPU1 is overheating and apply Dell profile if true
# Check if CPU 1 is overheating then apply Dell default dynamic fan control profile if true
if CPU1_OVERHEAT
then
apply_Dell_profile
# Set the flag to indicate that Dell profile is applied
if ! $IS_DELL_PROFILE_APPLIED
then
IS_DELL_PROFILE_APPLIED=true
fi
apply_Dell_fan_control_profile
if CPU2_OVERHEAT
if ! $IS_DELL_FAN_CONTROL_PROFILE_APPLIED
then
COMMENT="CPU 1 and CPU 2 temperatures are too high, Dell default dynamic fan control profile applied for safety"
else
COMMENT="CPU 1 temperature is too high, Dell default dynamic fan control profile applied for safety"
IS_DELL_FAN_CONTROL_PROFILE_APPLIED=true
# Check if CPU 2 is overheating too, Dell default dynamic fan control profile already applied before
if CPU2_OVERHEAT
then
COMMENT="CPU 1 and CPU 2 temperatures are too high, Dell default dynamic fan control profile applied for safety"
else
COMMENT="CPU 1 temperature is too high, Dell default dynamic fan control profile applied for safety"
fi
fi
# Check if CPU 2 is overheating then apply Dell default dynamic fan control profile if true
elif CPU2_OVERHEAT
then
apply_Dell_fan_control_profile
if ! $IS_DELL_FAN_CONTROL_PROFILE_APPLIED
then
IS_DELL_FAN_CONTROL_PROFILE_APPLIED=true
COMMENT="CPU 2 temperature is too high, Dell default dynamic fan control profile applied for safety"
fi
else
# Check if CPU2 is overheating and apply Dell profile if true
if CPU2_OVERHEAT
apply_user_fan_control_profile
# Check if user fan control profile is applied then apply it if not
if $IS_DELL_FAN_CONTROL_PROFILE_APPLIED
then
apply_Dell_profile
if ! $IS_DELL_PROFILE_APPLIED
then
IS_DELL_PROFILE_APPLIED=true
fi
COMMENT="CPU 2 temperature is too high, Dell default dynamic fan control profile applied for safety"
else
# Check if user profile is applied and apply it if not
if $IS_DELL_PROFILE_APPLIED
then
apply_user_profile
IS_DELL_PROFILE_APPLIED=false
fi
IS_DELL_FAN_CONTROL_PROFILE_APPLIED=false
COMMENT="CPU temperature decreased and is now OK (<= $CPU_TEMPERATURE_TRESHOLD°C), user's fan control profile applied."
fi
fi
# Enable or disable, depending on the user's choice, third-party PCIe card Dell default cooling response
# No comment will be displayed on the change of this parameter since it is not related to the temperature of any device (CPU, GPU, etc...) but only to the settings made by the user when launching this Docker container
@ -155,15 +161,15 @@ while true; do
enable_third_party_PCIe_card_Dell_default_cooling_response
THIRD_PARTY_PCIE_CARD_DELL_DEFAULT_COOLING_RESPONSE_STATUS="Enabled"
fi
fi
# Print the results, including the current fan control profile and comment
# Print temperatures, active fan control profile and comment if any change happened during last time interval
if [ $i -eq $TABLE_HEADER_PRINT_INTERVAL ]
then
echo "Time CPU1 CPU2 Inlet Exhaust Fan Control Profile Third-party PCIe card Dell default cooling response Comment"
echo " ------- Temperatures -------"
echo " Date & time Inlet CPU 1 CPU 2 Exhaust Active fan speed profile Third-party PCIe card Dell default cooling response Comment"
i=0
fi
echo "$(date +%T) $CPU1_TEMPERATURE°C $CPU2_TEMPERATURE°C $INLET_TEMPERATURE°C $EXHAUST_TEMPERATURE°C $CURRENT_FAN_CONTROL_PROFILE $THIRD_PARTY_PCIE_CARD_DELL_DEFAULT_COOLING_RESPONSE_STATUS $COMMENT"
i=$(($i+1))
printf "%19s %3d°C %3d°C %3d°C %5d°C %40s %51s %s\n" "$(date +"%d-%m-%Y %T")" $INLET_TEMPERATURE $CPU1_TEMPERATURE $CPU2_TEMPERATURE $EXHAUST_TEMPERATURE "$CURRENT_FAN_CONTROL_PROFILE" "$THIRD_PARTY_PCIE_CARD_DELL_DEFAULT_COOLING_RESPONSE_STATUS" "$COMMENT"
((i++))
wait $SLEEP_PROCESS_PID
done