hashed_update

Copy changed blocks from one file to another.

The program reads input file and hash file and copies only blocks where the hash is mismatched.

Example script:

# initial full copy
hashed_update fast_source NULL      slow_sink hashfile    65536

# periodically update
while true; do
    sleep 120
    hashed_update fast_source hashfile     slow_sink hashfile.new 65536
    mv hashfile.new hashfile
done

In case of using this for keeping in sync some mounted filesystem, usage of LVM snapshots both for source and for destination is recommended (Example).

hashed_update: Copies changed blocks from source to destination, according to stored separately block md5 hashes.
Usage: hashed_update input input_hashes output output_hashes blocksize
       Use NULL instead of arguments for various modes
       Output file must exist and be seekable
  Create hashes:   hashed_update smth NULL NULL   smth.hashes   262144
  Measure changes: hashed_update smth smth.hashes NULL NULL     262144
  Just copy:       hashed_update smth NULL        outfile NULL  262144
  Hashed update:   hashed_update smth smth.hashes outfile NULL  262144
  Hashed update plus save new hashfile: hashed_update smth smth.hashes outfile outfile.hashes 262144