I am trying to drive out the reason I am having some hard drive "hard resetting link" error problems on a Linux Debian box.

I am trying to eliminate the power supply as a problem in trouble shooting, as my problem is probably either software, power supply, or controller as the errors occur on all my SATA ports even after swapping cables and moving around drives.

My question is can a power supply put out correct voltages (tested with a multimeter on a molex connector inline with the hard drives) and still be bad (assume so)? My voltage on 12V was 12.02-12.00 and 5V was 5.04-5.01 while invoking writing operations in all the drives. Supply voltage to the power supply is 122V.

The power supply is a Antec basiq 350W powering a file/webserver/desktop (utilizing VMs) and during periodic checking I have never seen a problem with voltages via the BIOS in the past.

How would I check if the PS is good without purchasing a new PS? (I have no know-good PS to swap in)

Edit: The ports fail at different times usually, some times 2 at a time, but never 4 at one time (based on logs). If I invoke heavy write operations, I can usually get the error to occur, but not always.

Update: Using information from answers below, I tried to mess with the balance of the drives and the accessories on the two available 12V rails. A long shot but what the heck.

There are two legs of cables with molex connectors from the PS. Assuming (big assumption here) one leg is one rail, and the other on the second rail, I tried isolating the drives to one leg knowing they are pretty lower power devices, and all the accessories and a video card to another not knowing how much they were using together. This configuration left me errors on only one drive, port 1. Then, I focused on that drive. That particular drive was using a Molex-SATA power adapter, AND it was located at the very end of all the connections; also, the drive had both SATA and molex power input options. I switched from the SATA to the molex power input farther up the connection chain and have yet to be able to reproduce the problem on any ports, even with the CPU under 100% load during write operations.

Granted I have seen it takes many hours to get an error, but I was able to reproduce the error on port 1 fairly consistently via heavy writes w/o any CPU load and now can't. Sooo.. I'm crossing my fingers that the problem is found. Buying a PS is much easier than elusive software/firmware/bug issues! Hard to find a solution when you do not know problem!

Thanks to those who gave some insight!

UPDATE: The configuration has stopped the errors on all ports EXCEPT port 2. Not sure weather the drive is failing or the port because it is intermittent and it seems to work fine; until it is not mounted because of the hard reset link. But I think the PS was adding to the confusion and needs replaced with a higher output supply. Thanks again to those that helped!

UPDATE: Bought a quality 550 Watt +/- power supply and have not had issues since. Still kind of annoyed that 350 watts is not enough. :)

  • 1
    I'm not familiar with that error, but I can help you test the PSU. Install a monitoring tool like MotherboardMonitor or Sandra Pro and activate a log of the voltage on all rails. The data isn't entirely accurate, but the deviations from baseline will be precise. If you get fluctuations greater than spec the PSU is defective. – OCDtech Jan 15 '13 at 19:17
  • @OCDtech Will probably not help. HD's crash, but MB is stable. That usually means very short spikes. Those will typically not show in MB monitoring software. – Tonny Jan 15 '13 at 19:29

It is perfectly possible to measure good voltages and still have a bad powersupply.
You suspected that already...

The thing is that a multi-meter is a relatively slow device that won't pick up on very short power-spikes or dips, but these can easily cause a reset of the harddrives.
(There are some multi-meters that can do this, but they are very high-end expensive ones.)

The motherboard is generally not affected as it has it's own voltage regulators that provide some buffering and filtering, thereby isolating the circuits on the motherboard from the worst.

Without a second power-supply there is really no way to tell, but a cheap power-supply, just for testing, won't cost much. $25-$40 will probably get you one. If the system is stable with that you know for sure it was the power-supply. If it is still unstable you have a spare for future use.

But something else: Are you sure you are not overloading the power-supply or running it very close to it's maximum capacity ? 350W could be a little low for a server that also does VM's.
Especially with USA low input voltages (122 V) some power-supplies have serious issues if any of it's output rails needs to deliver more than 85-90% of it maximum rating. Check the label on the power-supply for the max ratings per voltage and do a little calculation what you are powering from it. Just to be sure.

  • I guess I thought if I was overloading the PS, the voltages would be low? Are you suggesting it might not exhibit low voltages and still be overloaded? The server is only used my 2 users right now, so server load (CPU usage, HDD writes, etc) is almost nill. – Damon Jan 15 '13 at 19:50
  • The point about the multimeter I did not think of. I think I just need to break down and get a secondary PS. Maybe someone else will have an idea on how to avoid that, but I really wanted a second opinion from a community way smarter the me. :) I also suspect it is NOT the PS, but I really need to rule it out before chasing a ghost. And truthfully, regardless of what I "think", it really might be the PS, hence the question and effort. Thanks! – Damon Jan 15 '13 at 19:52
  • @Damon If you are close to overloading the voltage might still look stable, but isn't really. Besides: There is a difference in voltage and current. Even with stable voltage the current may fluctuate which isn't good either. – Tonny Jan 15 '13 at 19:53
  • Valid point. Especially with electronics like hard drives which are probably sensitive to both. So many variables, ugh. – Damon Jan 15 '13 at 19:55

If you want a proper answer Electrical Engineering may be a better place to ask but a very simple version is that it is very possible that a voltage tested 'open circuit' (not under load) will be at the correct level but under the correct level when under load (with some drop to be expected). You could bridge the outputs with a resistor of suitable value (and wattage rating!) to simulate your system's peak demand before doing a load test but I doubt that would be worth it. If you are getting more than nothing it is probably OK.

  • The resistor is a good idea to simulate load directly instead through software. Thanks! – Damon Jan 15 '13 at 19:46
  • @Damon Do you have access to an oscilloscope? That may be the only practical way to check whether the output voltage is inconstant at a very short timespan. Peaks and troughs of course may not be its fault but that of a dirty feed (local interference - any heavy electrical machinery nearby?). – pnuts Jan 17 '13 at 12:54
  • err i've tested power supply voltage under load i'm sure.. by backprobing. As least I think I have. and I had no issue doing that. (dunno if it was safe - pnuts?) But indeed as pnuts has said, an oscilloscope would be more accurate. A multimeter is slow and can show good when a psu is bad. Though i've picked out many a bad psu with a multimeter. – barlop Nov 29 '13 at 20:47
  • @Sorry, not a general answer, just that I would not attempt it with my PC. – pnuts Nov 29 '13 at 21:01

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.