found drama

get oblique

debugging Malkovich

by Rob Friesel

While I wouldn’t say that I spoke to soon (re: “Poor Malkovich“), perhaps the situation bears some additional prudence to fully explore the options for remedy and repair.  After all, it should not be necessary to replace that iMac G5.  So let’s put all the cards out there.  Firstly — what the hell is going on?

Malkovich1 has been shutting down at random intervals over the past week or so.  It’s a troubling trend because (a) there is no warning when the shutdown is about to happen and (b) it then takes several minutes before the iMac will turn on again (i.e., respond to someone pressing the power button) during which time (c) there is a whole lot of anxiety about whether it will turn on ever again2.  Thus far, no pattern has emerged.  Some days pass with no shutdowns whatsoever, others have one or two; we’ve been lucky in as much as no single 24-hour period has been marked by greater than two of these random shutdowns3.  So we’ve been digging through system logs and trying to cross-reference the symptoms with those described in forum posts and the Apple support site’s knowledge base articles.  From this research, three theories4 have emerged:

  1. Over-heating; and/or
  2. a failing power supply; and/or
  3. a failing logic board.

Over-heating. If you’ve worked with an iMac before, you can see how this might easily be the case.  All the components are crammed into a thin, tight chassis that has little in the way of ventilation and cooling systems.  Over-heating seems to be a pretty common malady for iMacs (see also: here, here, and here).  From what I’ve found so far, unless you want to take drastic measures, there isn’t much you can do about the heating problem except periodically clean the dust out, keep the vents clear, and try to keep the room A/C-ified if you can.

If the problem turns out to be over-heating, I imagine I could try to just suck it up for a year or so.  But aside from my suspicions, there isn’t much to support the over-heating theory.  According to the discussion threads on the Apple site, if the shutdowns were due to over-heating then I should see something like this in system.log:

localhost kernel[0]: AppleSMU -- shutdown cause = -110

Instead, there are no lines matching “thermal” or “temperature” or “heat” and the only line matching AppleSMU or shutdown is:

localhost kernel[0]: AppleSMU -- shutdown cause = 1

…and seriously, WTF is that?

Failing power supply. I suspect that this is our most likely culprit, if only because Malkovich is already on its third power supply5.  In my research to date, I have been unable to identify specific steps to help diagnose a failing power supply.  As a Mac “civilian”, I’m left to some guess work.

That being said, I turn (naïvely or otherwise) to Apple’s support site to peruse the knowledge base articles looking for clues.  The most applicable kb article I’ve been able to find is this one:  iMac G5: Troubleshooting when your computer won’t turn on

If you care about semantics (like I do) then you’ve instantly picked up on the same quirk that I have:  my issue is not that the iMac has trouble turning on but that it randomly turns off.  Just the same, I did get a little value out of the article, namely (a) that I’m resetting the SMU every time that I unplug the iMac and that this should offer some relief from some of the symptoms and/but also (b) that I can in fact pull the cover off and do some diagnostic work, watching the flickering of some LEDs to get a sense of whether the issue lies with the power supply or (gasp) the logic board.

Failing logic board. I’m sad to say that with this scenario my optimism is reflected in the fact that I’m only throwing out the faulty logic board theory as a worst case scenario.  Based on the Apple kb article (re: reading LEDs and validating power supply vs. logic board issues; vide supra), it’s beginning to sound like the issue is that this third power supply is on its way out.  If the issue were with the logic board, the situation should be more dire.  Granted, I have yet to pull the cover off and examine the flicking of those little LEDs.  But at this stage, I think it’s still safe to make some guesses.

Not that I’m saying that a faulty logic board wouldn’t explain these random shutdowns or that they don’t explain them — I’m merely saying that there is not a stronger case to be made for the logic board explanation than the power supply explanation.  The system.log file really doesn’t show any messages that indicate a problem. What I’ve concluded from this (so far) is that whatever the issue is, it’s occurring in the hardware layer somewhere and that when it fails, it fails suddenly and totally and gives the software no opportunity to leave any trace evidence. Perhaps that sudden and total failure is happening in the labyrinthine circuitry of the logic board; or perhaps the power supply simply chokes on its own volts.

Again, as a Mac “civilian”, I’m left only with guesswork for now.

So what’s next? I’d really like to get to the bottom of this if I can. While I’d love a bright new shiny toy, I’d also love to solve this problem and stick to my original plans. I’m going to pop the hood and clean out Malkovich as a starter; it’s probably due to have the dust blown out anyway. That said, as I’ve already expressed my doubts about the over-heating theory, I don’t expect that this will get us terribly far.

That leaves the failing power supply vs. logic board theories to duke it out. I don’t expect that the diagnostic LEDs are really going to tell me much. I keep looking at “Step 12” in that Apple kb article and shaking my head: the path to “logic board failure” is predicated on a computer that gets power but doesn’t turn on. Again, “turning on” isn’t my problem — randomly turning itself off is my problem. If it is in fact a faulty power supply…

Well, I admit that I’ll be disappointed. Even the PCs I had before this didn’t burn through power supplies this quickly. (This would make 4 power supplies in 4 years!)

Comparatively speaking, replacing the power supply is a cheap fix ($110 vs. $860 for the logic board6) that could buy another year or two of useful life for this machine.  But then again…  “Another year or two”?  Sounds like we’re expecting that power supply to fail after another 10-14 months just like its predecessors.  O! the conundrum…  The questions then become:  are we really being that rough on these power supplies?  (And if so, what’s the baseline level of activity or “abuse” that’s reasonable to expect of them?)  And is there anyone out there that has had a nice “good long run” with their iMac?  Or any other theories?

UPDATE: Since writing this there has not been another random shutdown.  So that’s good, though not exactly encouraging or mind-changing.  I popped the cover on Malkovich and went to town on the dust with a Giotto’s Rocket Blaster; the LED diagnostics were (predictably) inconclusive which lends evidence to the bogus power supply theory, if you ask me.  But at least it’s clean in there now.  —F_D (5/2/2009)

  1. Malkovich is the hostname for our 20″ iMac (2.0 GHz G5 with ambient light sensor). []
  2. You can imagine how frustrating this is, especially when a copy of Barry Lyndon is trapped inside with 10 minutes left to go in the film. []
  3. This is not particularly comforting; poor ol’ BenNevis (my previous MacBook Pro) started off the same with its random shutdowns (a.k.a., b0rk!) — one or two at a time, one or two per day.  But over the course of a month or so that grew to approximately one per hour until it seemingly spiked at one per CPU clock cycle and stayed there. []
  4. A fourth theory developed when I saw this in system.log:

    Malkovich [0x0-0xf00f].com.apple.iTunes[121]: 2009-04-27 19:42:14.644 osascript[255:613] *** _NSAutoreleaseNoPool(): Object 0x130d30 of class NSCFData autoreleased with no pool in place - just leaking

    …but wouldn’t a memory leak just cause the system to get sluggish? and/or maybe just lock up?  I haven’t found any literature to support my theory that the memory leak is causing the shutdowns, so I’ve tossed it.  For now. []

  5. The first failed in April 2006 and was replaced by the good folks at Ormsby’s; that power supply then failed on July 4th 2008, two months after the AppleCare ran out (and just days after H. was born), and was replaced by the kindly folks at Small Dog Electronics.  This is not even to mention the hard drive issues we had in 2005. []
  6. For $860 you’re halfway to a 24″ iMac, or else maxing out a Mac mini. []

About Rob Friesel

Software engineer by day, science fiction writer by night. Author of The PhantomJS Cookbook and a short story in Please Do Not Remove. View all posts by Rob Friesel →

2 Responses to debugging Malkovich

justin says:

Yuck. FWIW, my MBP was experiencing increasingly frequent shutdown problems, but that turned out to be just a bad battery (it only occurred while not plugged in). I took it into the apple store out here, where they booted it to a nano, looked at the diagnostics (displayed in Classic-style chrome – w00t?), and swapped me a new battery. I happened to have still been under applecare, though, and the battery had less than 100 cycles on it.

If you end up debating an imac vs a maxed-out mini, I’d lean towards the mini. The form factor makes for a great HTPC…

Good luck!

found_drama says:

@Justin– Thanks. The more I research the problem here, the more I think it’s the power supply. Since it’s a user-serviceable part and can be had at a not-crippling cost, it might just might make sense to go ahead and fix it. The paranoia there is that I don’t want to pay again for diagnostics and labor AND parts when I could just pay for parts — only to find out that my guess work was bad and the issue continues.

RE: iMac vs. mini… jury is still out on that too!

Leave a Reply

Your email address will not be published. Required fields are marked *

*

*