Macbook Air intermittently can't find startup disk

About six months ago, my Macbook Air started having a strange issue where it would often not be able to find the startup disk, giving me the flashing question mark. Initially, I was convinced the hard drive (ssd) was dead. If I booted from a recovery USB, disk utility would not even find the drive. However, for whatever reason and luck, restarting the computer eventually fixed it; the startup disk would be found and the computer would boot up with no problem. I then checked the disk with a few utilities, none of which found any problems. But, if I turned off the computer, again it would most likely not find the startup drive. I would say it finds the startup drive about ten percent of the time. So if the computer restarts, then I might be rebooting it a dozen times before it boots up. Another thing is that if the computer has booted up, if I put it to sleep, it will not wake up; or it tries to but then it freezes. So that necessitates another round of restarts hoping it will boot. So I have backed up the hard drive contents fairly well, and I've since never turned the computer off or let the computer go to sleep unless it freezes. Fortunately, OS X is stable and I have something like 95 day uptime going, however, the computer is essentially a desktop now with no portability. Does anyone know what might be wrong with the computer? I'm pretty sure I've tried resetting the PRAM and SMC, though this was a long time ago... Anyway, does anyone know anything about a problem like this? I am running 10.7.5 on a Mid-2011 Macbook Air 11-inch, 1.6 GHZ i5.

Solution 1:

Hard disk drive failure

A few weeks ago I found similar symptoms with a colleague's iMac:

sometimes it behaved as if no drive was installed
sometimes during normal use, the Mac stopped working (not a total freeze; the reported symptoms were as if there was an I/O issue) – this was less likely to occur if the Mac was cool
in FireWire target disk mode, every time I tried to fully optimise its disk, the Mac behaved as if the drive had disappeared.

A replacement hard disk solved the problems for the user.

Testing your own hard disk drive

If you have an external optical drive, you could perform a thorough backup then try the most powerful test of HDAT2.

Caution

A routine that is intended to reallocate bad blocks may lead to dataloss.

In response to commentary

A clean installation might complete without error – and you might precede that with a single write of zeros throughout a volume – but still (I could/should have mentioned this in my first edition): I'd be wary.

If an entire drive has failed to present itself to software, at any time in the past, then to me that sounds a quiet but unforgettable alarm … expect the unexpected, and so on.

I should either:

a. no longer trust that hardware; or

b. perform extraordinarily time-consuming stress tests, which might reveal the nature of the failure and/or partially rebuild trust in the hardware.

In a geeky way I'm sometimes happy to spend/waste that time, but option (b) is not for everyone.

Related, in Ask Different

For bad and/or marginal blocks on hard disks, can anything be more effective than a single pass of zeros?

– all three answers were good. The accepted answer there links to a Tom Nelson article that includes:

use of Drive Genius (page 3) – in my experience not extending the scan significantly reduces the likelihood of detection of bad blocks
a Disk Utility approach to stress testing (page 4) – and for that stress test to be most effective, you should aim to repeatedly write over as much as possible of the disk before your clean installation of the OS.

(I wonder … can a Disk Utility approach be thorough enough to repeatedly write over the areas of disk that would be used for the GPT, and so on?)

Personally: if presented with a disk that might benefit from a multi-pass stress test, I would prefer HDAT2 to Disk Utility. Set aside the machine for a few days, allow HDAT2 to write repeatedly to the entire device …