I have an intermittent problem using WBEM to monitor health on two identical ESXi 5.0.0 hosts (Intel S3420TH). We have two more identical hosts running ESXi 4.1 which do not exhibit the fault. The server is returning invalid characters within the XML stream in varying elements of the CIM_NumericSensor results. Maybe 2 consecutive hours of the day it will fail, and the other 22 it is fine.
Snippet of failing XML output:
<PROPERTY NAME="NormalMax" TYPE="sint32">
<VALUE>148</VALUE>
</PROPERTY>
<PROPERTY NAME="NominalReading" TYPE="sint32">
<VALUE>109</VALUE>
</PROPERTY>
<PROPERTY NAME="Name" TYPE="string">
<VALUE>????????????????(18.0.32.99)</VALUE>
</PROPERTY>
<PROPERTY NAME="MonitoredDeviceId" TYPE="string">
<VALUE>7.2</VALUE>
Hex identifies the question-marks above as a series of "EF EF EF EF" characters, which are not valid in UTF-8 without encoding. When the problem clears itself, this particular Property name value is returned as "BB +1.1V P1 Vccp(18.0.32.99)". The problem also affects Property "BB +1.5V P1 DDR3(20.0.32.99)" with the same EF's injected intermittently.
I've verified the problem with 3 different tools: pywbem, wbemcli, and Fiddler. Is this likely an ESXi issue, or hardware? We intend to upgrade the two ESXi 4.1 hosts to 5.x, and wonder if the fault will appear. Thanks for any guidance.
Todd