backup daily to remote host, and delete old backups smarter

backup.sh

#!/bin/bash
date=`date "+%Y-%m-%dT%H_%M_%S"`
HOME=/YOUR/LOCALHOME
SERVER=ID@HOST
DIR=backup

rsync -azP \
  --bwlimit=500 \
  --delete \
  --delete-excluded \
  --exclude-from=$HOME/.rsync/exclude \
  --link-dest=../current \
  $HOME $SERVER:$DIR/incomplete_back-$date \
  && ssh $SERVER \
  "mv $DIR/incomplete_back-$date $DIR/back-$date \
  && rm -f $DIR/current \
  && ln -s back-$date $DIR/current"

timemachine.py

#!/usr/bin/env python
#
# This script cleans up your rsync backup storage like how TimeMachine does.
#
# It is distributed as a public domain, you can use it at any condition you
# need with your own risks.  For comments or patches, write to
# Hyeshik Chang <[email protected]>.
# 

import os
import glob
import shutil
from datetime import datetime, timedelta

BACKUPDIR = '/PATH/TO/BACKUP'
CURRENTLINK = 'current'
DIRPATTERN = '*/*' 
TIMEFORMAT = '%Y-%m/%d-%H:%M:%S'
DIREXCEPTIONS = []
KEEPINGRULE = [
    # (time from now,       interval)
    (timedelta(days=1),     None),
    (timedelta(days=7),     timedelta(days=1)),
    (timedelta(days=30),    timedelta(days=7)),
    (timedelta(days=365),   timedelta(days=30)),
    (timedelta(days=1100),  timedelta(days=365)),
]

def dirname2datetime(dirname):
    return datetime.strptime(dirname, TIMEFORMAT)

def get_snapshots():
    r = []

    for dirname in glob.glob(DIRPATTERN):
        if dirname.startswith(CURRENTLINK):
            continue

        if not os.path.isdir(dirname) or dirname in DIREXCEPTIONS:
            continue

        r.append((dirname, dirname2datetime(dirname)))

    return sorted(r, reverse=True)

def main():
    os.chdir(BACKUPDIR)

    snaps = get_snapshots()
    curtime = datetime.now()

    revrules = KEEPINGRULE[::-1]
    rules = zip(revrules, revrules[1:] + [(timedelta(0), None)])
    for (period, interval), (nextperiod, _) in rules:
        if interval is None:
            continue

        tmlimit = curtime - nextperiod
        newest = curtime - period - interval # accept the first in every period
        while snaps and tmlimit > snaps[-1][1]:
            dirname, dirtm = snaps.pop()
            if dirtm - newest < interval:
                shutil.rmtree(dirname)
            else:
                newest = dirtm

if __name__ == '__main__':
    main()