CPU Temperature Reported on the Citrix NetScaler MPX Appliances

Hi all, I noticed this important, CTX article that you should be aware of, if you have Citrix NetScaler MPX appliance in your environment. Then its a good idea to read this article and change the values, so you dont end up with a dead unit because of the CPU dies of heating.

/Poppelgaard

Summary

This document contains information about the CPU temperature reporting on a NetScaler MPX 7500, 9500, 10500, 12500, or 15500 appliances.

Background

The NetScaler appliance reports that the temperature is high. However, the original temperature is found to be around 30 degree Celsius less than the reported temperature.

CPU Temperature Reporting on the NetScaler MPX Appliances

An Intel CPU does not report temperature directly, neither in Celsius nor in Fahrenheit. Instead, the CPUs have a Digital Thermal Sensor (DTS) that report the temperature less than the maximum operating temperature (PROCHOT). When the DTS reported value for a CPU approaches zero, the CPU is close to its maximum operating temperature. Before the temperature of the CPU reaches PROCHOT value, the protective measures are automatically enabled for the CPU to keep the processor from getting heated.

The NetScaler MPX appliance reports the CPU temperature in degree Celsius. The value is calculated by subtracting the reported value from the PROCHOT value. For example, if the CPU reports a value of 38 degree Celsius, and the PROCHOT specification for the CPU model is 100 degree Celsius, then the system reports a CPU temperature of 62 degree Celsius, which is 100 – 38.

Detailed issue description with CPU temperature measurements on NetScaler 7500 or 9500 platform are described in http://download.intel.com/design/intarch/papers/322683.pdf.

For normal operating temperatures CPU temperature sensor cannot provide reliable and precise data. When the CPU temperature is high that is 20 degree Celsius less than the maximum temperature, the CPU sensor becomes more reliable.

CPU temperature reported is a distance from the maximum allowed CPU temperature. If CPU reaches maximum allowed temperature, then the sensor reads 0 degree Celsius.

As long as the reported CPU temperature is below the allowed maximum 100 degree Celsius reported by health monitoring, CPU thermal throttling does not occur. Thermal throttling and eventual CPU thermal shutdown starts at temperatures past the reported temperature of 100 degree Celsius.

Note: The thermal monitoring and the protective features that are built into the processor are not affected. Consider the following points on this issue:

Thermal protection of the processor is always enabled and accurate.

System fans are always accurately controlled for proper system cooling.

The actual temperature of the CPU is lower than the reported value.

Recommendation

It is recommended to set 85 degree Celsius as an alarm threshold for CPU temperature on the NetScaler appliance to avoid invalid temperature related warnings, as shown in the following screen shot:
Note: This temperature is about 15 degree Celsius less than the maximum CPU temperature.

This document applies to:

Source

http://support.citrix.com/article/CTX133584