archiveteam-megawarc-factory/chunker

#!/bin/bash
# Move uploaded .warc.gz files to an archive directory.
# When the archive is large enough, make a tar and start with a
# new archive.
#
# Be careful: this script assumes that any file in the upload directory
# that has a name that ends with *.warc.gz is a fully uploaded file and
# can be moved somewhere else. Remember this when running Rsync.
#

PATH_TO_UPLOADS=$1 # /home/archiveteam/uploads
PATH_TO_TARGET=$2  # /home/archiveteam/processed
MAX_MEGABYTES=$((1024*25))

# find every .warc.gz in the upload directory
find "$PATH_TO_UPLOADS" -type f -name "*.warc.gz" \
| while read filename
do
  # skip partial uploads
  if [[ $filename =~ rsync-tmp ]]
  then
    continue
  fi

  # move to the current/ directory
  echo "Moving ${filename}"
  mkdir -p "${PATH_TO_TARGET}/current"
  mv "${filename}" "${PATH_TO_TARGET}/current/"

  # if the current/ directory is large enough,
  # rename it to archive-XXXXX and start a new current/
  cur_size=$( du -BM -s "${PATH_TO_TARGET}/current" | grep -oE "^[0-9]+" )
  if [[ $cur_size -gt $MAX_MEGABYTES ]]
  then
    timestamp=$( date +'%Y%m%d%H%M%S' )
    echo "Current archive is full, moving to ${timestamp}."
    mkdir -p "${PATH_TO_TARGET}/archive"
    mv "${PATH_TO_TARGET}/current" "${PATH_TO_TARGET}/archive/${timestamp}"
  fi
done