[comp.sys.sgi] Backup Strangeness Exabyte

olson@anchor.esd.sgi.com (Dave Olson) (09/05/90)

In <1990Sep5.144841.9059@lavaca.uh.edu> mitchell@turing.chem.uh.edu (Mike Mitchell) writes:

|     I'm having a similar problem with Tar on the 8mm Exabyte drive.
|     I use the verbose mode of Tar after using mt to get to the end of a
|     tape, every thing looks good on the screen i.e. files are added to
|     the tape, etc.  But when I'm done, often nothing has been written to
|     the tape.  I've been told that there is a screw-up in the time-outs
|     for different operating system processes or at least a mis-match in
|     time-out lengths.  It seems that when the load average is greater
|     than the number of processors (roughly) that the system starts
|     screwing up.  Supposedly 3.3.1 to be released in mid September
|     (I'm told) will fix this problem.  Till then I keep duplicate
|     tapes and cuss a lot.
| 

Are you absolutely sure that you haven't got a typo somewhere, and
are not just creating a file in /dev?  I know it sounds silly, but
it happens all the time.  Do you see the tape drive light 'blink'
when you are writing data?

There is no timeout problem that I am aware of (and I would be the person
to be aware of it!) with any of the tape drives, on MP or single processor
systems.

Once again, the problem with 8mm in 3.3 is that the default block size
is 512 bytes, rather than 1024.  This results in incompatiblity with
3.2, and other vendors systems, and a loss of 1/2 of the tape capacity.
This is fixed in 3.3.1.  This is the only known SCSI tape problem in 3.3.

If that is your problem, you might want to try using the variable
block size device, with some multiple of 1024 bytes for the block
size (i.e, an even count for the blocksize to tar).


--

	Dave Olson

Life would be so much easier if we could just look at the source code.