[EMAIL PROTECTED] (Alvin Oga) writes:
> > No Linux project I know of can yet provide generic failsafe clustering.
>
> am thinking...( outloud )
>
> we can get dual hosted disk controllers...
> we can get dual CPU motherboards...
> we can get redundant power supplies...
> we can get removable disk drives...
>
> all we need is to "put all that into a generic linux box"... :-)
Fine. I don't know exactly what happens if one CPU in a multi-CPU
system completely drops dead in the middle of everything, but I don't
think Linux will like that. Not even on hardware built to be more
error-tolerant than the usual x86 junk. And in any case, a subtle
malfunction (executing corrupt code, for instance) will certainly end
up in a crash.
A dual-CPU Linux system is thus in fact twice as likely to die from a
CPU malfunction.
> problem is linux does NOT support "multi-homed" servers...
> if home1 is down...it doesn't automatically go to home2...
That's fairly simple to do for your average connectionless static web
site. What about applications that retain state, though (database
servers, e-commerce web sites, and so on). Quite a different
story. Some of the issues are relatively easy to solve, some are
difficult, and some have taken the likes of IBM and Tandem 20 years to
work through.
> am thinking...replace the "function" of what lilo does and make it
> look to one of several disks for the bootup sequences which would
> be on /dev/sda, /dev/sdb, or /dev/sdc etc...
And what about the system BIOS looking for the master boot record (ie,
LILO) on the first disk? If that disk is completely dead, fine, it'll
try the next one, but if it's just returning corrupt data? (Never mind
IDE hardware that would even handle the completely-dead case).
--
Osma Ahvenlampi
