Power Usage Metrics in Monitoring Show Zero for All ESXi Hosts After Upgrade to 8.0 U3

Products

• VMware vSphere ESXi
• VMware Aria Operations (formerly vRealize Operations) 8.x

Issue/Introduction

1. ESXi hosts show zero values for power usage under Monitoring. The historical data in VMware Aria Operations indicates that the readings were initially accurate, but the power metrics later consistently showed zero.

ESXi host under Monitor > Advanced > Chart Options and selected Power Usage. The value shows "0" (refer screenshot)

Environment

8.0 U3

8.18.0

Cause

1. When the sensord service starts, it checks whether the power readings can be fetched through IPMI or through RAPL. If the readings are available via IPMI, it enables the ipmiEnabled variable and periodically collects data using the powersvc plugin. If a few polling attempts fail, the plugin retries until it reaches the maximum retry limit.
   
   
2. In ESXi 8.0 U3, a regression introduced with the RAPL changes causes ipmiEnabled to be set to false immediately after a single polling failure. Once this happens, the plugin stops retrying and exits, leading to continuous zero readings for power metrics.

The following log entries confirm the failure pattern: 

# cd /var/run/log/sensord.log

 ####-##-##T##:##:##.###Z Er(163) sensord[########]: [Originator@#### sub=Powersvc] Failed to read from power provider
 ####-##-##T##:##:##.###Z Er(163) sensord[########]: [Originator@#### sub=Powersvc] Failed to read from power provider
 ####-##-##T##:##:##.###Z Er(163) sensord[########]: [Originator@#### sub=Powersvc] Failed to read from power provider
 ####-##-##T##:##:##.###Z Er(163) sensord[########]: [Originator@#### sub=Powersvc] Failed to read from power provider
 ####-##-##T##:##:##.###Z Er(163) sensord[########]: [Originator@#### sub=Powersvc] Failed to read from power provider
 ####-##-##T##:##:##.###Z Er(163) sensord[########]: [Originator@#### sub=Powersvc] Failed to read from power provider
 ####-##-##T##:##:##.###Z Er(163) sensord[########]: [Originator@#### sub=Powersvc] Too many failures, powersvc offline

# cd /commands/vmware_-vl.txt

 VMware ESXi 8.0.3 build-######
 VMware ESXi 8.0 Update 3

Resolution

1. This issue occurs due to a known regression in ESXi 8.0 U3. The only way to restore normal power metric collection is to upgrade the hosts to ESXi 8.0 P05 or any later build that includes the fix.
   
   
2. The temporary readings visible during support bundle collection appear because the command.
   
   2.1. Runs sensord in one-time mode, which captures data momentarily but doesn’t sustain ongoing power monitoring.
   
   

   # /usr/lib64/vmware/sensord/bin/sensord -v -l -D

   
   
   
3. Upgrading to the latest build resolves the RAPL regression and ensures that ipmiEnabled remains active even if a single polling attempt fails. This allows the powersvc plugin to continue collecting accurate power metrics, restoring normal monitoring behavior in VMware Aria Operations.