Took a while but...
A while back I sent a message to the list about a strange DD error
message. The problem was that DD was detecting an end-of-tape prematurely.
It turns out that it was a bug (again!) in the Computer Asscociates
Unicenter product that handle our tapes. When rewinding after a backup
and doing a DD on 14 filesets to check the tape, the Unicenter
software would skip every other file causing DD never to read the
intended 14 files on the tape.
But in the process of debugging this problem I received a lot of
usefull suggestions from the list:
Claude SOMA [soma_c_at_decus.fr] --- Pointed out that the blocksize
of my 'vdump' was 60K and to change the bs= parameter on DD to match.
Lucio Chiappetti [lucio_at_ifctr.mi.cnr.it] --- Gave me tips on running
a "tape analysis" using /usr/field/tapex (see man tapex) which gave more
details on the content of a tape.
John Hascall [john_at_iastate.edu]
alan_at_nabeth.cxo.dec.com
Dr. Tom Blinn [tpb_at_zk3.dec.com] --- They all clarified that ENOSPC, the
errno
code for the message "No space left on device" is the code for when you run
off the end of the device or that the last backup used all the tape.
Lew Randerson [lranderson_at_Pppl.GOV] --- Suggested that I check that
/dev/null
was actually a null device. If it would have been accidentally removed, DD
could have ran out of space in the /dev directory (root partition) while
writing to a 'real' /dev/null file.
Thanks again for all the help, this list is great!
Yvon Lauriault
National Library of Canada
P.S. My original message follows,
-----------------------------------------------------------------
Hi Gurus,
Last friday we ran our backups as usual but had a strange DD
error message. Because of unreliable tape drives we have taken
the habit of running DD against the tapes after each backup to
make sure they are fine. At the end of the (16) vdump's we use
the following commands to rewind and check the tape:
mt -f /dev/rmt1h rewind
dd if=/dev/rmt1h of=/dev/null ibs=1024 obs=1024 files=16
Last time we got the following error message:
dd read error: No space left on device
134978+5 records in
134978+0 records out
Since we are just reading the data from tape and writing to
the bit bucket ( /dev/null ) where is it running out of space?
Thanks in advance!
Yvon Lauriault
National Library of Canada
Received on Wed May 06 1998 - 19:10:57 NZST