SCSI tape data loss

[Kern Sibbald]

Sorry, I didn't mean to create a ruckus -- I just
don't like data loss -- especially in "Bacula". :-)

What I wrote in the other mail was more or less a
core dump. When things calm down (and hopefully I get
to the bottom of this), I'm going to try to point out
the things that could *possibly* be changed without
creating chaos with existing programs.  Some things
like not specifying a large enough buffer and getting
an error (on FreeBSD) are preferable to Linux just
returning part of the buffer -- but it is a difference
worth knowing about.



On Wed, 2003-06-04 at 16:51, Matthew Jacob wrote:
> Yes, well, I have not been able to myself reproduce it.
> 
> Some of the issues you brought up in the other mail have been worth all
> this ruckus in any case.
> 
> I'll be OOT from thu-wed so I may not respond quickly or get to looking
> at bacula myself. I did leave a couple tape drives hooked up to one of
> my 4.8 boxes so I can fool around with stuff remotely if I get a chance.
> 
> > Hello,
> >
> > The latest tests indicate that the sequence that I
> > gave you below does work. I'm going to look more
> > carefully at the Bacula code (the code path is
> > fairly complicated) because perhaps it simply
> > releases the drive without doing a rewind.
> >
> > I'm going to also work on simplifying the case
> > that we know fails (and it seems to be perfectly
> > reproducible).
> >