[Tfug] OT: Disk testing

[Bexley Hall]

Hi, Adrian,

--- Adrian <choprboy at dakotacom.net> wrote:

> OK... so I;ve been lurking in this for a while and
> figure it;s time to chime in with my own ideas... 
> I think you are making this way to hard. The
problem, 
> as you seem to say, is that you want to test IDE
> (not SATA) drives in a quick and efficent manner.
> The problem, of course, being that IDE is not 
> hotswap-able.
> 
> The answer is to use something that is
> hotswap-able...

Yes -- hence my original USB external drive
suggestion.

> Get an IDE->Firewire enclosure and be done with it.
> Either a bunch of single internal swappable 
> bays, such as the following:
>
http://granitedigital.com/catalog/pg31_firewiresmarthotswapbay.htm
> 
> Or a JBOD array like the following:
> 
http://granitedigital.com/catalog/pg22a_firewireidehotswapraid_1.htm

This ignores the "on-the-cheap" requirement -- and,
requires the host to be firewire enabled.

> The Firewire connection will be fast enough for
> several simultanious drives, is hotswap-able, and
> appears as a standard SCSI-like disk under Linux.

I think USB 2.0 can address the bandwidth requirements
easily.  Note that you still have disk access times to
consider so the disk isn't streaming data continuously
(especially for random seek tests).  I am pretty
sure the PCI bus or the CPU would crap out long
before the USB would saturate.  *Especially* if
using some OTS approach (as below).

> Hack together a quick BASH or PERL script that loops
> forever, periodically checks for new devices, and
> then runs a child "badblocks -n" on found drives.
> If the child doesn;t return any errors (or perhaps a
> minimum number of errors), then 
> the drive is fine and report a message to such
> effect to the user.

There's more involved than just "badblocks".
The drive has to be labeled, *then* scanned/tested,
formatted, file system built, OS installed, etc.
The heuristics involved in deciding when to
"give up" on the disk would also be variable.
E.g., you might accept *no* errors on a 10G
disk but would be willing to accept "a few" on
a 100G disk.  In some cases, you might even want
to low-level format a "defective" drive to see
if it can be salvaged -- though expecting the
same box to be able to do all of these things may
be far too ambitious a goal  :<  Having to move
a drive from one machine to another just to
accomplish all of these steps would be wasteful
of time (and invites drives to be dropped/mishandled)

--don


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com