🚀 go-pugleaf

In article <11...@ucsd.Edu> br...@ucsd.edu (Brian Kantor) writes:
>I once typed rm -rf / & (God, it hurts just to type that in a
>news article!) and realized the error about 10 uSec after hitting the
>return key.  Yanking the power plug out of the wall stopped it before it
>had a chance to write the updated buffers to disk and it suffered no
>loss at all.

I once worked on a bid for a Government contract which involved fault-tolerant
processing.  Everything was handled by redundant hardware (CPUs, memories,
disk controllers and mirrored volumes) and there was, of course, integral
battery backup.  If a board went south, it lit an error LED (Good Scout! :-),
and you simply yanked it out while the system was still running (hey, that's
why there's *two* of everything) and shoved in a new one.  The system said,
"Ooh, a new component.  Better test it ... looks okay, I'll put it online".

It seems that one of our number had been working a few too many hours and
accidentally cleaned out his working directory rather than the old directory.
He immediately told the local system rep (Thanks, Chris!) whose response was:
"Pull *both* CPU cards.  NOW."

Still, I think the thing that saved us was the filesystem organization, which
allowed the "fsdb"-like maintenance program to go in and "undelete" the files.
-- 
cha...@C3.COM	{decuac.dec.com,cucstud}!c3pe!charles	ex::!echo Boo: