Librenix  
(Show all Scripting articles . . .
)  
Headlines | Linux | Apps | Coding | BSD | Admin | News
Information for Linux System Administration 

Scripting: A parallel Linux backup script

Up
vote
Down

This example bash shell script demonstrates a simple method of creating backups of multiple filesystems to multiple tape devices simultaneously. While the script presented writes to four tape drives in parallel, it can easily be modified to write to other device types and to create a different number of backup streams. The script is set up for the bash shell under Linux, but modifying it for another variety of Unix should simply be a matter of changing the locations of utility files such as tar, echo, cp, and sleep.

The script can be downloaded from http://librenix.com/scripts/par.tar.sh. Download the file now and load it into an editor as this article will refer to it frequently. Also, you may want to modify bits of it to match your filesystem names and your devices.

The first line of the script looks like this:
 #!/bin/bash
If the bash shell isn’t in the /bin directory on your system, you’ll need to modify this line. Enter the command which bash now to verify the location of bash. My Fedora Linux system and my Mac OS X system both have bash in /bin, but my FreeBSD system does not. If you have a non-Linux flavor of Unix, you’ll probably need to use the ‘which’ command to verify the locations of each command used in the script. The commands used are:
 bash
cd
sleep
echo
date
tar
wait
ls
wc
Note that ‘wait’ and ‘cd’ are usually implemented as internal shell commands and may not have external commands associated with them. If that is true for your system, leave ‘cd’ and ‘wait’ with no directory prefix just as they are in the original script.

Now, the first command in the script resets the current working directory to ‘/’:
 cd /
Since the script precedes each directory to be backed up with a ‘.’ to represent the current working directory, starting out at ‘/’ is necessary. The reason for this precaution is that some implementations of the tar command will only load files from a tar archive into the exact directory that was specified when the file was backed up. By prefixing the names with a ‘.’ we preserve the ability to recover the files into any subdirectory we want, without overwriting the original files.

Immediately after the ‘cd /’ command is where you would put any commands to shut down all services that must be quieted prior to a backup. The example script has a (commented out) command to initiate an Oracle database shutdown followed by a ‘sleep’ command to allow time for the shutdown to complete. The example database shutdown and the following delay probably don’t apply to your system. Obviously, you’ll have to add commands yourself to stop any applications that might interfere with the backup.

Next, we use the ‘date’ command to create two sets of four tiny files to stick at the start and end of each tape. Note that the presence of a ‘date.#’ file at the beginning of each tape lets you quickly find out when a tape was created and on which drive. The ‘zzzz.#’ files, appended to the end of each tape, only serve to let you easily verify that a backup completed without overrunning the end of the tape.

Next, we start the four actual ‘tar’ backup commands, each with sample directories named ‘./dir1’, ‘./dir2’, etc. Of course, you’ll need to modify the list of directories to match the actual directories you wish to back up. Note that you’ll probably want to balance the directory sizes so that all of the largest directores aren’t on the same tape. Also, note that each ‘tar’ command is run in the background and logs to a tar.#.log file in the /tmp directory. Obviously, you might want to put the logfiles somewhere else.

After each ‘tar’ command there is an entry like this: ‘TASK=$0’, or ‘TASK=$1’. These arbitrarily-named ‘TASK’ variables are used to store the process ID of each ‘tar’ command so that the script can wait for them with the four ‘wait’ commands that follow in the next block of code. There, we have the four ‘wait’ commands waiting on the $TASK0, etc, variables. (The addition of the ‘$’ to each TASK# shell variable is not a typo -- it’s necessary to read back the contents of the variable.)

Next, after the script has waited for the completion of each of the four ‘tar’ commands, it appends some information to a history file for later reference. It stores the date of the backup, the filesize of the logfile, and the number of files backed up on each tape to each of four history files. While the script will overwrite the logfiles (tar.#.log) each time it is run, it will append these three lines to each of the four history files (tar.#.history).

The final steps in the script are commented out. Those are the commands necessary to restart any applications that were brought down for the backup. Again, in the example we assume an Oracle database needs to be restarted. You’ll need to add the commands necessary to start any applications that were stopped at the beginning of the script.
mail this link | score:9228 | -Ray, April 10, 2005
More Sysadmin articles...

Buy Large Wall Art on Canvas

admin headlines

Testing the Digital Ocean $5 Cloud Servers with an MMORPG

USB Redirection hack on Two Node Controller+Compute Neutron GRE+OVS Fedora 20 Cluster

Gluster 3.4.2 on Two Node Controller+Compute Neutron GRE+OVS Fedora 20 Cluster

Installing Debian testing On GPT HDDs From A Grml Live Linux

Running ownCloud 5.0 On Nginx (LEMP) On Debian Wheezy

Running SilverStripe On Nginx (LEMP) On Debian Wheezy/Ubuntu 13.04

Step By Step Ubuntu 13.04 (Raring Ringtail) LAMP Server Setup

 

Firefox sidebar

Site map

Site info

News feed

Features

Login
(to post)

Search

 
Articles are owned by their authors.   © 2000-2012 Ray Yeargin