654499 - minis with gray screens

Status: RESOLVED FIXED

(Infrastructure & Operations :: RelOps: General, task)

Description

[Dustin J. Mitchell [:dustin] (he/him)]

In reboots bugs like bug 653574, we often see talos minis hanging at startup with a gray screen.  There seem to be a number of "frequent fliers" in this category.

Options:
 * run diagnostics on a frequent flier
 * try to replicate while showing console output (hold 'c' on startup, IIRC)

Comment 1

[Zandr Milewski [:zandr]]

(In reply to comment #0)

>  * run diagnostics on a frequent flier

Grabbed a gray screen in the last batch of reboots and ran Apple Hardware Test. This machine (talos-r3-fed64-013) wasn't particularly a frequent flier, but it was there, and I was bored.

The quick test didn't report a problem, but the extended test reported 

4HDD/11/40000004: SATA(0,0)

Google says that's a dying drive. This makes me rather nervous, as we've lost the drives in a few machines now.

Comment 2

[Zandr Milewski [:zandr]]

w7-032 did not show any problems in an AHT extended test, and that's certainly a frequent flier.

Comment 3

[Dustin J. Mitchell [:dustin] (he/him)]

There's an nvram hack to always be verbose on startup -- it might help:

http://osxdaily.com/2007/03/25/always-boot-mac-os-x-in-verbose-mode/

Only problem is, I don't know how to do that from anything but OS X.  That's a start, I guess.

Comment 4

[Dustin J. Mitchell [:dustin] (he/him)]

Attachment: m654499-puppet-manifests-p1-r1.patch

Comment 5

[Dustin J. Mitchell [:dustin] (he/him)]

landed (whoops, wrong credit for the review, sorry!) and deployed.

I can look into how to do this from Fedora, but on the assumption this is a hardware problem, hopefully we'll get lucky and see it on a snow or leopard box.  I don't have much hope of pulling this off on windows XP systems, much less unmanaged W7.

Comment 6

[Dustin J. Mitchell [:dustin] (he/him)]

Incidentally, I have verbose mode enabled in the nvram on the mini I have here, and Fedora is now installed - there's nothing verbose to see.  So it's only relevant with Mac OS X.

Comment 7

[Amy Rich [:arr] [:arich]]

The gray screen problem has not gone away, but closing RE Dustin since the work in this bug is complete.