Re: [bu-users] trivial restore test failing

[Vincent S. <bu...@hi...>]

On Tue, Jan 20, 2009 at 02:19:27PM -0600, jack byers wrote:
> so, I now  cant duplicate the problem!
> 
> just to confirm, i also then rm'd the files just bu'd to /restorebu
> [root@bootp restorebu]# pwd
> /restorebu
> [root@bootp restorebu]# ls
> home
> [root@bootp restorebu]# rm -rf home
> [root@bootp restorebu]#
> 
> and then redid the orginal (if not verbatim)
> [root@bootp byers]# bu -d /restorebu drhmail09
> 
> Incremental:      on
> Backup directory: /restorebu
> 
> /home/byers/drhmail09 -> /restorebu/home/byers/drhmail09
> /home/byers/drhmail09/.#rhjan09 -> /restorebu/home/byers/drhmail09/.#rhjan09
> /home/byers/drhmail09/#rhjan09# -> /restorebu/home/byers/drhmail09/#rhjan09#
> /home/byers/drhmail09/rhjan09 -> /restorebu/home/byers/drhmail09/rhjan09
> 
> [root@bootp byers]#
> 
> so now it is working as it should, no explanation as to
> original failure.

I guess we will just have to chalk it up to another one of those random
unexplained glitches in Linux.  Linux can really have you chasing
shadows in your applications when there is really no bug in the app.  

For example, back when we ran Linux, for years we had an unexplained bug
in our in-house cad system we use for circuit board layout, where when
you would hit a certain zoom level it would just wig out and stop
redrawing the screen and you could no longer zoom in or out.  We could
find no bug in the code to explain it and gave up on finding it for
a long time and just worked around it.  This bug went on for years
through major revisions of Linux.  At the time, we never considered it
might be caused by an OS bug.

Then after major show stopper bugs, such as severe memory management
problems, non-working NFS, multiple system freezes per day, etc., etc.,
finally forced us to change to BSD, much to our surprise, the bug in our
CAD system went away, and never came back.  Not only with a native
compilation, but even when running the Linux binary on BSD under Linux
emulation, the bug was gone.  We tested under Linux emulation just to
see if it might have been from a library it was linked to.  If the same
bug existed on BSD with the Linux binary under emulation, it would point
to a library rather than a core OS problem such as memory management or
something.

I still from time to time get email about a problem with bu that turns
out to be more core Linux OS bugs.  The last one I remember was where
a user was getting his process table filled with zombie processes when
running bu.  It turned out to be yet another ancient bug that resurfaced
in Linux that has not existed in other UNIX's for many years.