Skip to content

Latest commit

 

History

History
280 lines (204 loc) · 8.55 KB

README.md

File metadata and controls

280 lines (204 loc) · 8.55 KB
Raw

backup

(c) 2016-2018 UCL, Dr. Andrew C.R. Martin

A flexible backup script for Unix/Linux systems driven by a configuration file. Allows backups from a directory to another directory (typically remotely mounted), via SSH or via the rsync protocol. It makes use of rsync to perform the actual backups. backup can also perform backups of PostgreSQL databases.

backup is driven by a configuration file which specifies the directories to be backed up and the location(s) to which each directory should be backed up. By default, files in the backup are not deleted unless it is a Sunday when the program is run.

Obviously backup needs permissions to read the directories being backed up, so normally it would be run as root. To backup databases, the code must be run as root, or alternatively the PostgreSQL superuser (normally 'postgres') must not have a password for read access to all databases.

Program Installation

To install the program in your local directory, simply type:

   ./install.sh

This will install the Perl script in $HOME/scripts (as backup.pl) and a link will be provided in $HOME/bin (as backup).

This can be overridden by specifying the script directory followed (optionally) by the binary directory on the command line.

For example, to install the script in /usr/local/apps/backup and a link in /usr/local/bin, you would do:

   ./install.sh /usr/local/apps/backup  /usr/local/bin

If you want to keep the script in the directory where you unpacked the program, then just do:

   ./install.sh .  /usr/local/bin

A utility script has also been provided if you have sudo permissions to install in your home scripts and bin directories and to install system wide in /usr/local/sbin:

   ./home_and_system_install.sh

Configuration File

The configuration file may contain comments introduced by a # sign at any point.

The configuration file has three sections:

  • An OPTIONS section to define global options
  • A DISK section to define disks to be backed up
  • A DATABASE section to define databases to be backed up

OPTIONS

The OPTIONS section defines global settings. Currently the following settings are supported:

  1. EXCLUDE is used to define patterns for files that should be excluded from all backups. Typically this might be used to exclude editor backup files, compiler object files, etc. Multiple EXCLUDE records may be specified. A typical setting would be:

     EXCLUDE **/*~
 EXCLUDE *.o

The pattern syntax is as used by rsync.

  1. PGDUMP is used to specify the location of the PostgreSQL pg_dumpall program. e.g.

     PGDUMP /usr/local/bin/pg_dumpall

If not specified, it is assumed that pg_dumpall is available in the standard path.

  1. PGSUPERUSER is used to specify the PostgreSQL superuser. This defaults to postgres. e.g.

     PGSUPERUSER johnsmith

  2. PGNOSU is used to change the way that the pg_dumpall program is run. e.g.

     PGNOSU

By default, the script uses the su command to become the PostgreSQL superuser. This means that the script must be run as root (since otherwise a password would be needed to change user). It also means that the directory where the database backup is placed must be writable by the PostgreSQL superuser.
If PGNOSU is specified, then the script doesn't change to run as the PostgreSQL superuser, but instead simply runs the pg_dumpall program specifying that the dump should be done as the PostgreSQL superuser (with the --user option). This does not require that backup is run by root or that the destination directory is writable by the PostgreSQL superuser. However it does require that the PostgreSQL superuser is not password protected in the database.

  1. RSYNCPW is used to set a password for a remote rsync daemon. This is only used if you are doing backups to a remote rsync daemon rather than local disks or to a remote server over ssh. Currently only one password may be specified and consequently you would normally be backing up to a single remote rsync daemon server. e.g.

     RSYNCPW rsync

DISK

The DISK section is the main section specifying directories to be backed up and the locations to which they should be sent. There may be multiple DISK sections.

Each section begins with a DISK command followed by the directory to be backed up. This is then followed by one or more BACKUP commands followed by the destination to which the files should be sent.

   DISK /data
   BACKUP /nas/backup/data NODELETE
   BACKUP user@remotehost1:/backup/data
   BACKUP user@remotehost2::backup/data

This specifies that the directory /data should be backed up to three locations:

  1. the locally mounted folder /nas/backup/data. Backed up files are never deleted even if they have gone away from the source directory.

  2. via SSH to /backup/data as the specified user on the specified remote machine. Note the single colon followed by a slash (:/).

  3. via the RSYNC DAEMON to backup/data as the specified user on the specified remote machine. Note the double colon followed by NO slash (::). This is a 'short directory' name as specified in the rsync daemon configuration file on the remote host.

Optionally, one or more EXCLUDE commands may be used to exclude patterns from the backup. This works in the same way as the EXCLUDE command in the OPTIONS section, but is restricted to this particular disk's backups.

Note The DISK command is used to specify a directory whose contents are to be backed up. With rsync, you need to add a / to the end of the directory for this to happen (otherwise with rsync the specified directory and its contents will be backed up). Adding a / to the end of the directory name will not change the behaviour of backup.

DATABASE

The database section works in the same way as the DISK section.

The DATABASE command introduces a DATABASE section and is followed by the port number to be backed up optionally preceded by the host followed by a colon. For example:

   DATABASE 5432

--or--

   DATABASE localhost:5432

This is followed by one or more BACKUP commands which specify a backup file. e.g.

   BACKUP /nas/backup/pg/myhost_5432.sql

Note 1 The destination backup directory (in this case /nas/backup/pg), must be writable by the PostgreSQL superuser.

Note 2 Remote backups over SSH are not currently supported. However, you can always run a standard file-type backup of an SQL backup to a remote host.

Example configuration file

   # Set global excludes
   OPTIONS
   EXCLUDE **/*~                         # Exclude anything that ends in a ~
   PGDUMP  /usr/local/bin/pg_dumpall     # Full specification for pg_dumpall
   
   # Backup /home/
   DISK   /home
   BACKUP /localbackup/home
   BACKUP /nas/backup/home NODELETE      # Never delete files in this backup
   
   # Backup /data/
   DISK    /data
   BACKUP  /nas/backup/data              # Backup locally
   BACKUP  user@remotehost:/backup/data  # Backup over ssh
   EXCLUDE tmp/                          # Exclude any tmp directories
   
   # Backup PostgreSQL database on the local host on port 5432
   DATABASE localhost:5432
   BACKUP   /nas/backup/pg/myhost_5432.sql

Initialization

backup takes a cautious approach to performing backups. To do this, it keeps a file called .runbackup in each source directory and in each destination directory. This ensures that no "backup" is performed if either the source or destination directory has been unmounted.

You can create these files manually - you can simply type, for example

   touch /data/.runbackup
   touch /nas/backup/data/.runbackup

(Note that no such file is needed for backups over SSH.)

Having created your configuration file, you can simply type:

   backup -init

to create all the destination directories and .runbackup files automatically.

Invoking the program

By default, backup will look for a file called backup.conf in the current directory. Failing that it will look for backup.conf in the directory where the program itself lives. Normally, however, you run backup by specifying the configuration file on the command line:

   backup mybackup.conf

Typically you will want to run nightly backups from a cron job, so on a Linux machine, you would create a script in your /etc/cron.daily directory containing something like:

   DATE=`date +%F`
   LOG=/var/tmp/backup.$DATE
   /path/to/backup /path/to/backup.conf > $LOG

Further help

Further help may be obtained by typing

   backup -h