[wplug] Since we're off topic

[Drew from Zhrodague]

> The problem with this problem is that it's so intermittent.  For example, 
> my regression testing this evening went like this:
> 
> 1) unplug CD-ROM IDE cable - try to boot.  Blank screen, no POST.
> 2) unplug HDD IDE cable - try to boot.  BIOS POST, no IDE devices, "Insert 
> system disk and press enter"
> 3) reattach HDD IDE calbe - try to boot.  Successful.
> 
> I then began composing this message and the computer utterly froze.  No 
> warnings or anything, no bluescreening (this is a Windows box) - just a 
> halt.  
> 
> So I powered down, re-unplugged the HDDs and tried again.  I repeated
> maybe half-a-dozen times.  To date (since this evening, when I started) I
> have not seen the computer fail to POST when the hard drives are
> unplugged.  But holy mother of crap is it bad news if my HDDs are failing.  
> First of all, I am paralyzed at work and school without software I have
> installed on this machine, and second of all, etc., etc.  You know, data
> loss and stuff.
> 
> Since the problem seems to be correlated to the *first time* I try to 
> power up the machine after n hours, I will wait for awhile and try again 
> with the HDDs not plugged in.  If I see it fail even once without the 
> HDDs, then it's probably the motherboard (there isn't anything else here 
> except for a wireless NIC).  If not, then I'm in data recovery hell and 
> will probably prevail upon the list again soon.
> 
> Hell, why wait?  Is there anyone who would be able to provide a host 
> machine that I could use to ghost my old drive? (assuming I can get it to 
> power up for awhile).

	Sounds like a power-supply, dewd. Don't skimp on those. Bad power 
supplies are the leading cause of machine failure (minus user-error). What 
kid of power supply do you have in that thing? Bad RAM may also be the 
cause, but I've seen few cases of that. Of course, I only get monster 
power-supplies anymore, so...

	This is prolly a great time to point out the virtues of redundant 
disk space. For those of you who setup Linux boxes, would it really kill 
you to buy an extra identical disk, and setup a raid-1? Linux has this 
kernel-raid feature, where you can effectively do whatever kind of RAID 
you'd like (mirroring, speed/space, parity, concatination). Linux Kernel 
RAID is the only way to go, if you absolutely hate losing data.

	I usually my partitions like this. If I have only two disks, I do 
a RAID-1 (mirroring). With more than two, it becomes more fun:

	/boot	raid-1		250-500m
	swap	raid-1/5	1-4g
	/	raid-1/5	rest

	My GIS host at home has 4 of those WD 120g disks, with 8m of cache
each. I use the two onboard IDEs, plus a PCI ATA100 card. Seems to work
okay, but I'd rather have two of the ATA100 cards, and not use the
onboards. The arm-swing is still the slowest part, so it's useful to save
an interrupt. Might not be the fastest thing in the world, but disks DO 
DIE, and I've lost enough data to never want to lose a single bit ever 
again.

	Redhat's install package lets you set your partitions this way. It 
used to be that you would mount the space of another disk somewhere else, 
but disks are HUGE these days. Buy two, and set it up right. You'll never 
worry about data loss again.


-- 

Drew from Zhrodague		http://www.WiFiMaps.com
drew at zhrodague.net		Location Based WiFi