Hardware: Raj's new 20-drive Tower Budget build (http://greenleaf-technology.com/blogs/prototypes/index.php?id=2601838967498401788)

 

Server built on Sunday, started preclearing 4 drives (2x4TB, 2x3TB) simultaneously Sunday morning.

 

I checked on them this morning, and everything seemed fine at 20 hours' progress.

 

An hour later, I check the system remotely, and it's unavailable.

 

I get home tonight, and the power is completely off.  I press the power button - no response.  I flip the PSU toggle off, then back on after a few minutes and try again - the system boots.

 

I tried running a few preclears again.  No dice - this time the system powers down as I'm running the fourth preclear.

 

Restarting it on a different power outlet, I tried running two preclears.  Again, system powerdown, but this time after a few minutes or so. 

 

Following this thread (http://lime-technology.com/forum/index.php?topic=21099.0), I picked up on the b14 trick, and gave it a shot.  The last two tries have been with one preclear at the eleven minute mark, and both runs ran almost immediately after 11:15 in both instances before the power died.

 

I'm leaning towards thinking this is a CPU issue in some way (planning on trying an RMA, though I was careful as could be when installing), but figured I'd post in here on the off chance that it isn't.  I'll post back the impact when I get a new chip in there, but if anyone has seen or heard anything like this before, I'd definitely appreciate any guidance at all.

 

Last syslog attached

syslog7212.txt

I wondering why you suspect the CPU.  Even with no CPU installed, I'd expect the fans to start to spin when the power button is pushed.

 

My suspicion would lie, at first, with the PSU and if that is proven to be okay, look at the motherboard.

I've had a few friends tell me of "phantom power-offs" that eventually seemed tied to CPU activity (like a preclear), saying they were most likely the result of poor seating or other damage CPU-wise (absent other signs)... is that way outside the majority?

 

(EDIT: I just realized, I'd probably written my original post poorly - I don't have any problem turning the system on (even if it shuts itself down, as long as I toggle the PSU off, then on again) or maintaining power without real activity... my problems stem from the system completely shutting itself down after brief periods of preclearing activity (a few minutes on rc5-8168, consistently 11:15 minutes on b14) but otherwise staying up and running "idle" unraid just fine.  Hope that clears my situation up a bit!)

 

 

When a system powers itself down completely when under heavy load, it's usually either heat or voltage related.

 

Which power supply are you using, and are you sure the cpu heatsink is fully flush with the cpu with just a tiny dab of heatsink compound spread evenly between them?

Had the same issue as johnathanm mentions:

http://lime-technology.com/forum/index.php?topic=18224.msg163905#msg163905

Well, this is bizarre - I checked my CPU and its fan, everything seemed fine (gel present, heatsink flush), but I re-seated it just to be safe (checked the pins, too - everything looks perfect).  Unplugged and replugged the PSU connector from the motherboard for good measure.

 

Now, the system still powers on, but it turns itself off ALMOST IMMEDIATELY after it spins everything up (talking 5 seconds or less).  Also worth noting: last night, after shutting itself off, the only trick to turning it back on via front panel was a PSU power toggle off-on... Now that it's shutting itself down super-fast, that trick doesn't work.  The power toggle is a 50-50 shot, with odds increased by more time between toggles.

 

Any ideas?  I'm leaning more towards PSU than CPU now, but the progression (20 hours fine, less than an hour fine, less than a minute fine) without any software elements outside of unraid on a stick and a preclear script has me completely confounded.

 

PSU = CORSAIR Enthusiast Series TX650 V2 650W

PSU connector to the motherboard not fully engaged?

 

I had the reverse problem when all drives were idle:

http://lime-technology.com/forum/index.php?topic=18329.msg164152#msg164152

It definitely appears to be connected - I've just unplugged and reconnected to test, and I'm getting the same results...

 

Also worth noting: I've tried the "screwdriver on motherboard pins connected to front panel" trick, and while it seems to run SLIGHTLY longer (maybe 7 seconds) using that method, it still kills itself mid-boot now...

 

I would think that a PSU would be good or bad out of the box - has anyone ever heard of a PSU "degrading" over the first couple days of operation?

I would think that a PSU would be good or bad out of the box - has anyone ever heard of a PSU "degrading" over the first couple days of operation?

 

Not over a couple of days ... but a little bit longer, yes.  And this was with a very subtle fault, probably to do with noise on the rails. The voltage on the 5 volt rail, while still well within spec, dropped about 10mV.  The symptom was simply that the on-board lan failed to respond. Swapping the PSU for an identical model resolved the problem, and the system is still running to this day (several years later).

OK, PSU has been RMA'd and hit the mail (just in time for a 4th delay...) With a little luck, I'll have my replacement in time for the weekend, and I can give it a shot.

 

Thank you for the help, everyone!

I'm going to agree that it's likely a power supply issue, but also going to add that you check 2 other things if the PSU replacement doesn't work. I've had two cases similar to this, in one case, there was a wire to the cpu fan that snagged the fan itself and prevented it from spinning. Basically resulted in an instant overheating situation. It didn't drift into the "snag" position until the unit had been running a fair while and heated up. (like your 20hr preclear scenario) The second and far more hair-pulling issue turned out to be a short of the case's power on switch. That started after the unit had been running for over a year, and drove me nuts to find. In the end, I wired the on/off lead to the reboot button and it's been good ever since.

 

Hope you get lucky though and it's the PSU!