I've got a whitebox server running CentOS 5.7. It's a Dual Xeon 5620, 24GB of RAM. The mainboard is a SuperMicro X8DT6-F and the chassis is a SC825TQ-R720LPB. Dual 720W Power supplies.

We had a big power outage a couple weeks back that took down everything, I don't have any pre-power outage figures for this server, and the only reason I noticed these is because when I was bringing up the servers I was checking them out with more scrutiny than usual.

http://i.imgur.com/rSjiw.png (Image of voltage readings)

As you can see, CPU1 DIMM is low, +3.3V is high, 3.3VSB is high, +5v is high, +12v is REAL LOW (out of normal 5% (plus/minus))... and VBAT is off the charts.

With my whitebox VAR we've tried the following:

  1. Swap out PSU with another server I have with the same PSUs.
  2. Try different power cord
  3. Update BMC/IPMI firmware in case readings were wrong (They aren't)
  4. Update BIOS
  5. Try different PDU
  6. Try a different outlet and/or circuit
  7. Replaced Voltage Regulator Unit

At this point, the only thing we haven't done, seemingly is replace the mainboard.. which is what the next step will be unless something else shines some light on the situation.

I should mention the system is rock solid otherwise which is a surprise given the 12v voltage is that far off.

  • 4
    Have you tried measuring the voltages with an actual volt meter? Could be the sensors are malfunctioning. Vbat would be the easiest to get at to test. – Grant Aug 23 '12 at 3:17
  • Not yet. Its on the list. – Mindflux Aug 23 '12 at 3:21
  • Can you tell us more about the hardware? How many drives/RAID cards do you have hanging off that 12 volt rail? Are you using the stock fans (5-7K RPM IIRC), do you have frequency regulation on, etc. What's the load avg? Linux will downclock the cores when they're not being utilized. – Gazzonyx Aug 23 '12 at 3:33
  • 5 drives in the system on the SAS2008 chip on board. I'm not sure what you mean by frequency regulation? Fans are stock. Linux is set up ondemand for CPU throttling but I upped the min scaling to 2Ghz from 1.6 and dropped the percentage to throttle at down from 80% to 20% (CPU is used in short bursts). Power issues were noticed before those items were changed. – Mindflux Aug 23 '12 at 3:36
  • BIOS seems to suggest everything is in spec which makes me curious how IPMI can be so far off.. presuming they read off the same sensors. Very very curious. – Mindflux Aug 27 '12 at 3:20

Given that you say the server is rock solid and working properly, and that those voltages are significantly off, I'm betting that either the monitoring software is misinterpreting the voltages, or the sensors are malfunctioning/defective. If those were the actual voltages, the system would have plenty of issues, and the battery would likely be damaged from overcharging.

I'm assuming you are using IPMI to get the voltage readings - since that's controlled by the hardware manufacturer it's unlikely to be wrong. So that leaves faulty voltage sensors, which would be built into the motherboard on most systems.

To test this theory, use a volt meter to measure the battery and the 12v power rails. Since they are the furthest off it should be easy to see if they are anything close to what the sensors report.

Solution is to either replace the motherboard, or ignore the sensors and hope for the best.

  • Yes, I'm using IPMI. Both the IPMI client on my desktop as well as the IPMI "web interface" show the same voltages... that's not to say the IPMI BMC is completely bonkers.. but I am going to have to agree the motherboard is the likely culprit here now. – Mindflux Aug 24 '12 at 13:39

Resetting the BMC config cleared out the errors on the sensors and everything reads fine now.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.