using multiple dump files in export while compressing

[jt]

We compress(while running the export, not compressing afterwards) our export dump file during the export run so it could stay below the 2G limit. Now even the compressed file exceeds 2G. I know in 8i you can use more than one dump file by specifying in file=exp.dmp1,exp.dmp2... But how do I do this while compressing at the same time?

The part of script we use now is:

mknod $HOME/$DAY$1 p
compress -f <$HOME/$DAY$1> $HOME/$DAY$1.Z&
(
exp parfile=$HOME/genexp$1.par file=$HOME/$DAY$1 log=$HOME/exp$1.log$XDAY
)

So basically

mknod file.dmp p
compress -f file.dmp.Z&
(
exp parfile=parfile file=file.dmp
log=logfile
)

So how do I add a second dmp file to this script so I can have a second compressed dmp file when the first one exceeds 2G?

[marist89]

Ah, time for some smoke and mirrors.

We got around this problem by starting several named pipes and then accessing those named_pips in the files= clause.

For example:

#setup the pipes and file list
NUM_NP="1 2 3 4 5"
for i in $NUM_NP
   rm -f /tmp/exp_${i}
   /etc/mknod /tmp/exp_${i} p
   DMP_FILE=${i}.Z

   compress < /tmp/exp_${i} > $DMP_FILE &

   if (( $i == 1 )); then
      FILE_LIST=/tmp/exp_${i}
   else
      FILE_LIST=$FILE_LIST","/tmp/exp_${i}
   fi
done

#do the export
exp / file=$FILE_LIST filesize=8000M  ...

# kill any pipes that didn't get used
for i in $NUM_NP; do
   kill %$i
done

We had to futz with the 8000M to get out .Z files somewhere around 2G, but we got pretty close.

[Edited by marist89 on 07-10-2001 at 04:44 PM]

[Dave_A]

Why not use the split command to break the compressed dmp file into manageable chunks?

for example (requires lots of tidying, path names etc.)...

mknod exp_pipe p
mknod compress_pipe p

# start a compress process, listening on the named pipe
# that the export process will write to, output the compressed
# export to another pipe for 'split' to pick up. Note, start
# process in background, so it can sit and wait for the export
compress < exp_pipe > compress_pipe &

# start a split process, listening on the named pipe that the
# compress process writes to. This process will split the
# incoming data into 2047MB chunks, called xaa, xab, xac etc.
# see the man page on split for how to change the output file
# names.
split -b2047m < compress_pipe &

# perform the export, writing to the first named pipe
exp / file = exp_pipe ...


HTH

David.