?

Log in

No account? Create an account

Previous Entry | Next Entry

Hardware Wars

After spending a number of hours getting http compression working for the machine that runs belfry.com, The Nice Forums, and our local image server, I got handed a belated April Fools joke by fate. The new hard drive that I'd just replaced a couple months ago crashed. So I got to buy a new hard drive and spend the rest of the evening installing it and rebuilding the RAID mirror. Thank goodness for RAID. Two hints for prospective sysadmin types, though... Never install two IDE drives on the same chain if you plan to RAID them together. One failed IDE disk can take down both drives in the chain, at least temporarily, and can possibly cause more serious problems. Secondly, when you re-arrange cables in the case before closing it, make sure that you remember to plug all the drives back in!

Comments

( 9 comments — Leave a comment )
tuftears
Apr. 3rd, 2003 10:25 am (UTC)
You and I in West Corner of the Park
fluff a pack of wuffies 'til they float like larks,
set them free at the break of dawn
'til one by one they were gone...

In FurrySpace, bugs in the software
flashed the message, somethin's out there,
floatin' in orbit high,
ninety nine wufballoons go by...

(hope that helps cheer you up! Let me know if you want the rest of the lyrics.) ('gryn)
revar
Apr. 3rd, 2003 11:52 am (UTC)
*snerk*
smack_jackal
Apr. 3rd, 2003 11:00 am (UTC)
Oddly enough in the SATA standard there is only one drive per channel for exactly the same reasons you've just mentioned..
cjthomas
Apr. 3rd, 2003 11:23 am (UTC)
RAID.
I appreciate the Mighty Power of RAID (tm), as it's the only thing that let me build the fileserver without stressing about the "which IDE drive manufacturer makes lemons this week" game.

The downside:

  • I'm using an ancient version of mdtools. Drives take about a full day to check or do anything else drive-wide on.
  • I have no automatic integrity checking running beyond the RAID tools themselves. If silent corruption was happening, I'd never know it until too late.
  • I haven't tested the behavior of this system under failure conditions. Logically, I should have tried stomping a sector or three on one drive, or removing a drive (when powered off), to see what happens, before blindly trusting the system.
  • I haven't tested my ability to restore from this system.


Maybe I _should_ be stressing...
kairee
Apr. 3rd, 2003 12:53 pm (UTC)
Due to similar concerns about 'lemons this week', I recently invested in an Adaptec 2400A raid-5 card. Four channels, with one drive per channel, though I'm only using three of them. It's kind've nifty having 2+1 raid-5, and I actually feel vaguely secure about this box now. :)

The same concepts behind your warning are probabaly why Adaptec requires one drive per channel on this card.
shatterstripes
Apr. 3rd, 2003 02:32 pm (UTC)
Maybe you should try unleashing the awesome new technology of floppy-disc RAID instead.
revar
Apr. 3rd, 2003 05:11 pm (UTC)
Aiiiiiiiiiggggggghhhhhhh!

Though at least it would make floppies semi-reliable.
doodlesthegreat
Apr. 3rd, 2003 05:24 pm (UTC)
That is just evil.
cjthomas
Apr. 3rd, 2003 11:15 pm (UTC)
Delightfully evil =^.^=.
That's beautifully, beautifully wrong };>.

The irony is that you could probably tweak Linux/vanilla *BSD to do something similar, as I doubt it cares whether you're using /dev/hdx or /dev/fdx.

I also once had a copy of an old DOS program that did something similar, building a "virtual hard drive" by striping 5.25" floppies. You had to switch disks by hand every time you needed to access a different disk, and it had no error checking (meaning about one sector write in five was corrupt), but man, was it a fun toy }:>.

Probably still kicking around on that old MFM drive that I threw away the card for.
( 9 comments — Leave a comment )