Hi..Masters,
I realy hats off to linuxquestions.org site. This is very very useful site to everyone who all are on Linux platform. I am daily once or twice visiting this site so I could get some idea about shell scripting and at last I could made a shell script to do an incremental backup. All of the commands are not my own, I selected some commands from linuxquestions.org itself and also from here and there. I checked it contineously more than 6 hours by changing date and running script. It is doing well as my intention and the incremental backups also will happen only then if any changes are made since last backup file and if there is no changes then it will refuse by creating an empty archive. That is what I needed. Althogh I request you all to suggest me if still any modifications have to be done in the shell script. Or if this is good for an incremental backup with tar then I will go with this for backup.

Here is my shell script
#!/bin/bash
week=`date +%w`
date=`date +%d`
weekly_dir="/backup/weekly"
monthly_dir="/backup/monthly"
owb="/backup/old_week"
omb="/backup/old_month"
backup_dir1="/root /home /etc /var/www /var/spool/cron /data"
backup_dir2="/data/"
f=data
###do monthly full backup on every 1st of the month###
if [ $date -eq "01" ]; then
rm -rf $omb/*; mv $monthly_dir/* $omb
echo archiving $f.`date '+%Y-%m-%d'`.tar.bz2 ...
tar -cjf $monthly_dir/$f.`date '+%Y-%m-%d'`.tar.bz2 $backup_dir1
ls -ltrh $monthly_dir
fi
###do weekly full backup on every sunday###
if [ $week -eq "0" ]; then
rm -rf $owb/*; mv $weekly_dir/* $owb
echo archiving $f.`date '+%Y-%m-%d'`.tar.bz2 ...
tar -cjf $weekly_dir/$f.`date '+%Y-%m-%d'`.tar.bz2 $backup_dir2
else
###do incremental backup###
###prepare backup file name and compute the new number###
cd $weekly_dir
MAX=6
n=`ls -rt1 [0-5]* | tail -1 | cut -d. -f1`
new_n=`expr \( $n + 1 \) % $MAX`
filename=$new_n.$f.`date '+%Y-%m-%d'`
###remove the oldest one###
if [ -f $new_n.$f.* ] ; then rm -f "$new_n.$f"*; fi
###read the time stamp of recent file in $weekly_dir###
stat $( ls -rt $weekly_dir | tail -1 ) | grep Change | cut -d: -f 2,3 | while read time;
do
###create a file in /tmp so find command can find only new files in backup_dir2 ###since timestamp of /tmp/timetamp
touch -d "$time" /tmp/timestamp;
done
echo archiving $filename.tar.bz2 ...
###search only new files in $backup_dir2 and tar it###
tar -cjf $weekly_dir/$filename.tar.bz2 $(find $backup_dir2 -type f -newer /tmp/timestamp)
ls -ltrh $weekly_dir
rm -f /tmp/timestamp
cd
fi

I don't really have the attention span to read your script at this time (haven't slept in a couple of days), however, for incremental backups, why not use rsync instead of tar?

For example:

Using that command, you won't re-backup every single file and directory. You'll only make up the difference.

In my experience, rsync works great for everything except for virtual machines and other [extreme] large files.

2 members found this post helpful.

Yes master, you are right. I have used rsync many times. It works great. But I just wanted to check an incremental backup with only tar. So I posted this shell script here to finalize by some experts like you. Although rsync is good one, user friendly and can be done with only below command to sync with permission, time stamp, archiving while transfer and also with secure shell etc. But I request you all to suggest me if still any modification has to be done in this shell script.