DL380G4 Temperature measurement via WMI

[Igor Milavec]

Hi.

I'd like to monitor remote server's temperature using WMI. I have the latest ProLiant Support Pack installed, however, the WMI class Win32_TemperatureProbe isn't supported. I also tried using MSAcpi_ThermalZoneTemperature, however, I get very strange results:
Active: True
ActiveTripPoint: 0,0,0,0,0,0,0,0,0,0
ActiveTripPointCount: 0
CriticalTripPoint: 31,3
CurrentTemperature: 8,3
InstanceName: ACPI\ThermalZone\THM0_0
PassiveTripPoint: 9,8
Reserved: 1
SamplingPeriod: 60
ThermalConstant1: 4
ThermalConstant2: 4
ThermalStamp: 6

Does anybody know how to successfully monitor server's temperature?

Regards, Igor

[David Claypool]

Many users mistakenly feel they have to monitor temperature levels. This is bogus for a number of reasons:

- the data can NOT be used to infer ambient temperature; an external sensor is required

- the data are meaningless on their own; they are only meaningful in relation to the thresholds

- thresholds are set by the designers based on the component in questionâ s specification, the proximity of the sensor to the component and the airflow at that point.

It is only important to know what component sensors are protecting, and that information is available from the Insight agents, such as this from a DL380G3:

Sensor_Location_Temperature_Threshold_Type
1_____CPU_____ 39 C________59 C______Caution
2_____IO Board___41 C________68 C______Caution
3_____CPU______55 C________73 C______Caution

Looking at these data, why are there 2 radically different readings for CPU? One is 39C and the other is 55C. What conclusion is to be made? Only that one sensor is closer to the heat source than the other and the thresholds are adjusted accordingly. A physics algorithm is applied to the distance and airflow in order to derive the exact setting of the threshold.

ProLiant servers manage this for themselves through the Advanced Systems Management Controller. The ASM chip, working in conjunction with the serverâ s Advanced Server Recovery features set in the BIOS will implement the thermal policy (alert, continue, shut down) and will report the event to the operating system event log and send an SNMP trap through the Insight agents.

The practice required of â whiteboxâ manufacturers of having to actively query the temperature of the sensors over the network through some tool is an unecessary waste of bandwidth, useless data and just plain dumb.

[Igor Milavec]

David, temperature monitoring on servers may be "just plain dumb", but the fact is that ProLiant is providing bogus data via ACPI and does not implement any other public interface to access this data.
Are you saying it's HP's position that users must not monitor their server's temperature?

Regards, Igor

[Matthew Finbow_1]

The data can be accessed via SNMP, if the insight agents are running, and SNMP is accessible from the remote machine (access control restrictions may apply - check service configuration)

Using a free tool such as getif, you can use the Compaq mibs (cpq*) - the data should be in cpqhealth, you can use this to find which OID request will get which sensor.

On at least Windows 2003 Server, there is a WMI SNMP Provider but I don't know enough of this to comment.

As to one CPU hotter than the other it may be it is doing more work (an idle CPU should be cooler than a fully loaded CPU) - if trending you may notice system board temperature increase and decrease as system load changes (assuming a controlled external temperature)

Regards,

Matt

[David Claypool]

Igor: go back and read my response. My point is that you can't generalize from the data available. It is significant only in regards to the thresholds and no inference can be drawn. This is true for ANY system, regardless of manufacturer. The ASM chip manages this process, eliminating the need for actual temperatures to be monitored.

[Nathan Shaw]

I am trying to determine the same thing. We also need to check for under-temperature, in case of aircon malfunction. An external ambient sensor may suffice, but we haven't yet found one that suits our needs.

Nathan.

[Jeffrey Rivet]

The temperature of the servers is very relevant in terms of my developing maintenance schedules. For me, I have 50 servers located in remote sites where they do not have the luxery of clean rooms. By monitoring the system temperature over time, I can determine when they need service and cleaned. By accessing the temperatures remotely, I plan to collect this data and provide a web report for easy identification of those that need service. It is also usefull to understand what is "normal" for your system and to be able to recognise when something is different. To simply wait for a threshold to be exceeded is not very proactive and I wouldn't be working very long if I only waited for failures.

[Paolo Argentieri]

Hello David,

I like HP SIM and System Management homepage, but, as Jeffrey pointed out, sometime there is a need to monitor the temps via WMI. This is because such indicators may need to be integrated in a third party UI. This UI does not replace the HP Management homepage, yet most users focus their attention on the application that is central to their role (e.g. Operator Console).

Dell open manage makes it simple by exposing the powerEdge info via WMI (root\CIMV2\Dell namespace)

For the HP Poweredge I have compiled the MIBS with the Microsoft compiler as follows:

REM compile MIBS
%windir%\system32\wbem\snmp\smi2smir.exe /m 2 /g cpqhost.mib .\GeneralMIBS\rfc1213.mib > ./MOFS/cpqhost.mib.mof
%windir%\system32\wbem\snmp\smi2smir.exe /m 2 /g cpqhlth.mib cpqhost.mib cpqsinfo.mib .\GeneralMIBS\rfc1213.mib > ./MOFS/cpqhlth.mib.mof

REM comp MOFs
mofcomp ."\MOFS\cpqhost.mib.mof"
mofcomp ."\MOFS\cpqhlth.mib.mof"

I can see some CPQHLTH classes in "root\snmp\SMIR\CPQHLTH_MIB" namespace using MS WMI CIM Studio but no instances.

I am guessing that there is no HP SNMP provider to which the MS SNMP service can talk to.

Is there a ProLiant SNMP provider I can use?
Or a WMI to WBEM gateway (I know that HP provides a WBEM to WMI gateway)?

Please help.

Regards,
Paolo