[GTALUG] solved: mysterious restarts

[D. Hugh Redelmeier]

Spoiler: it was the power supply.  You may stop reading here.

[I'm top-posting because the earlier message was so long ago.]

More testing: the firmware's memory and disk test routines found no 
errors.

The machine would spontaneously in Windows too.  So pretty clearly a 
hardware problem (including mains power as hardware).  I tested in 
Windows so that I could more convincingly ask for warranty support from 
HP.

With all this "homework", HP support did not ask me to do any silly things 
such as re-install my OS.  They agreed that hardware was likely at fault.  
They decided a new motherboard and processor might do the trick (I said 
that was fine but I asked if it might be the power supply).

HP offered onsite service but I opted for them to send me the parts for a 
"self-install".  They sent a motherboard and a processor (but not the 
heat-sink/fan).  The instructions that they sent were not suitable for 
first-timers (HP has manuals that would help but they didn't point me at 
them).  I swapped the parts but the problem persisted.  They then sent a 
new power supply and that seems to have fixed it.

This is a business computer (i.e. more expensive and better service).
The machine seemed to be designed for easy servicing, more so than
consumer models.  HP support's shipping was quick and cost me nothing.
They did not give me a hard time.

The worst part of support was the long hold times on the phone.  I've
heard "Axel F" way too many times and it sounds better on a HiFi
system than through low bandwidth VoIP.

| From: D. Hugh Redelmeier via talk <talk at gtalug.org>
| To: Toronto Linux Users Group <talk at gtalug.org>
| Date: Tue, 14 Jun 2016 11:10:17 -0400 (EDT)
| Subject: [GTALUG] mysterious restarts
| 
| We have a computer that started doing random and frequent restarts on the 
| weekend.  We don't know why.
| 
| Any suggestions?
| 
| It looks as if the power dips momentarily and the computer reboots,
| with no message that we have observe.  But it could just as easily be
| a crash of some other kind that leaves no trace.
| 
| The system is an HP Compaq Pro 6300 Small Form Factor PC running
| Fedora 20.
| 
| The crash seems to be at different points (i.e. not one consistent
| software activity).  The crashes don't seem correlated with heavy
| workloads (eg. it crashed a couple of times while I was staring at log
| files to see if there was any hint of the problem).
| 
| Hypothesis: a Fedora 20 bug.  But the software has not been changed in
| months.  Updates have not been appled this year.  Since the behaviour
| has changed without the software changing, I don't think that Fedora
| is to blame.
| 
| Hypothesis: it might be heat-related (the room it is in gets warm).  I
| vacuumed out the interior and defuzzed the heat sinks.  This did not
| improve the uptime.
| 
| Hypothesis: it might be contact-related.  So I disconnected and 
| reconnected most internal connectors and reseated the memory.
| This did not seem to improve the uptime.
| 
| Hypothesis: it might be the power supply.  Normally, I'd swap power
| supplies to test this hypothesis but this Small Form Factor computer
| has a unique (and probably expensive) power supply.  I opted to move
| the disk to a Dell OptiPlex 990 Small Form Factor computer and use that.
| 
| The Dell, with the HP's disk, seems stable.  No rebooting.  This is in
| the same warm room, but the weather has changed.
| 
| In the original HP box, I installed a disk that I had laying around (a
| 60G drive from a discarded laptop), installed Ubuntu 16.04, and have
| been running four CPU-bound processes for 24 hours.  No crash.  I
| admit that this is in a cooler room.  The heat and power load of a
| laptop drive is less than that of a 3.5" HDD, but I would not think
| that that is significant.
| 
| The computer is a couple of years old but still has a year of
| warranty.  There are confidential files on the disk drive so I'd like
| to narrow down the problem before calling in HP support.  Asking for a
| particular replacement part is more convenient that shipping the
| computer back to HP.