Linux administration

• Disks fail
• Bugs in software can cause corruption
• Configuration mistakes by administrator
• Accidental deletion or overwriting (e.g., with rm, mv or cp)
• Malicious deletion or virus attack
• Theft of machines with hard drives in
• Fire, or other disasters which can destroy hardware

• Traditionally, backups have been made onto tapes
  • Can store lots of data on reasonably cheap tapes
• Copying to a different hard disk
  • There is a risk of losing the backup along with the original
  • Better if on a remote computer
• CD writers can be used to store backups on CDs
  • Convenient for long-term storage
  • Handy to remove to remote locations

• Full backup — includes everything of importance
  • Might not include system files which are from the install CD
  • Can include a lot of files, many of which hardly ever change
• Differential backup — only includes changes since last full backup
  • Nightly backup only needs to include files changed since the last full backup
  • Recovery requires the full backup on which it was based
• Incremental backup — only includes changes since last backup
  • Nightly backup only includes files changed in the last 24 hours
  • Recovery requires the last full backup and a complete sequence of incremental backups after that

• The backup schedule should be regular and well known by those who rely on it
  • It must be decided what to backup and what can be left out
• Typically a full backup is done once a week or once a month
  • Daily changes are recorded in a differential or incremental backup each night
• Large sites might have more than these two levels to their strategy
• Monthly tapes might be kept for a long time, in case a really old file becomes important

• tar can package up files for distribution or backup
  • Originally for "tape archive" manipulation
  • Files can be stored anywhere
• Encapsulates many files in a single file
  • Known as a tar archive or a tarball
• Has unusual command-line option syntax
  • Common options are given as single letters
  • No hyphen is needed
• tar must be given exactly one action option
  • Indicates which operation to perform
  • Must be the first option

• Use the c option to create an archive
• For example, to create an archive called docs.tar.gz containing everything in the documents directory:

$ tar czf docs.tar.gz documents

• f specifies the archive's filename
  • Must be followed directly by the filename
  • Common to use .tar extension
  • Any subsequent options require a hyphen
• The z option compresses the archive with gzip
  • .tar.gz extension used to indicate compression
  • .tgz extension also popular
• The list of files and directories to archive follows the options.

• To check that a tar file has been made correctly, use the t operation (for 'list'):

$ tar tzf docs.tar.gz

• The z and f options work as for the c operation
• To show more information about files, add the v (for 'verbose') option
  • Shows information similar to ls -l
  • Can also be specified with c to list filenames as they are added

• Use the x operation to extract files from an archive:

$ tar xzvf docs.tar.gz

• The v option lists the files as they are extracted
• To extract individual files, list them on the command line:

$ tar xzvf docs.tar.gz documents/phone-numbers.txt

• Other useful options:
  • k (--keep-old-files) will not overwrite any existing files, only extracting missing ones
  • p (--preserve-permissions) will set extracted files to have the permissions they had when archived

• Under Linux, tape drives are accessed through several groups of device files
• Each device group has number, with the first drive numbered 0
• These are the most commonly used devices:
  • /dev/st0 — SCSI tape drive, which will be automatically rewound after each operation
  • /dev/nst0 — the same drive, but with no automatic rewinding
  • /dev/nft0 — the same without rewinding
  • /dev/ht0 — ATAPI tape drive
  • /dev/nht0 — the same without rewinding

• Tape drive devices can be read and written directly by tar
• To write a backup of /home to the first SCSI tape drive:

# tar cvf /dev/st0 /home

• We haven't used compression (the z option)
  • This might make the backup slower, at least on less powerful machines
  • Compressing the whole archive would make it much less resilient against corruption
• In the example the auto-rewinding device is used, so the tape will be rewound after tar is finished, and the archive can be extracted:

# tar xvf /dev/st0 /tmp/restored-home

• mt can move tapes backwards and forwards, and perform other operations on them
• Usage: mt [-f device] command [count]

• The -f option sets the tape device to use (e.g., /dev/st0)

  • The default is usually /dev/tape, which should be a symlink to a non-rewinding device, like /dev/nst0

• These are some of the more common commands:

  • fsf, bsfm — move forwards and backwards one (or count) files
  • eod — go to the end of the valid data
  • rewind — go to the start of the tape
  • offline — eject the tape

• Being selective about what is included in the backups can drastically reduce the time and space taken

• For example, /bin, /sbin, /lib and /usr could be restored from an installation CD

  • But it might still be worth backing them up to make restoration simpler

• The things which are most likely to be important in backups are:

  • /home
  • The CVS repository, or other places where project work is stored
  • Some directories under /var (particularly email)

• Some other areas which shouldn't be backed up are:
  • /tmp — usually doesn't contain anything of lasting value
  • /proc — automatically generated by the kernel
  • /dev — if using devfs this is also generated automatically
  • /mnt — some media mounted here, like CD ROMS, typically aren't backed up
  • Filesystems mounted remotely whose backup is taken care of elsewhere

• It is common to have a script to perform backups each night

  • Might perform different types of backup, e.g., a full backup on Saturday night and a differential one on other nights

• Such a script can be run with cron, making backup automatic

• Example scripts are readily available on the WWW

• cpio — alternative archiving program
• afio — similar to cpio, but allows files to be compressed individually
  • Compressed archives can be made which are resilient to corruption
• amanda: Amanda is an open source software that works on Unix/GNU Linux and Windows. It supports native backup utilities and formats such as GNU tar for backups on Unix/Linux.

• a. Create a single file in your home directory containing a backup of all the contents of /etc/.
• b. Create another archive with the same contents, but compressed to save disk space. Compare the sizes of the two archives.
• c. List the contents of each of your archives.
• d. Extract the contents of one of the archives into a directory under your home directory.

• e. Create a new subdirectory, and extract a single file from the archive into it.

  • Your current directory must be the one into which to extract.
  • You will need specfiy the path of the file to be extracted, but without a leading slash.

With an archive of /etc/, extracted under your home directory:

• a. Modify at least two files in your extracted copy. Re-extract one of them from the archive, losing your changes in that one file but preserving your changes elsewhere.
• b. Delete some files in your extracted copy. Make tar discover which these are and re-extract them from the archive, without clobbering changes made to other files.

• a. Produce a list of the names of all files under /home/ which have been modified in the past day. Only include regular files in this list.
• b. Create a tarball containing all files under /home/ changed modified in the past day. Why is including directories in this list not sensible?
• c. Create a tarball containing all files on the system that have changed in the past day, and which are in directories you deem worthy of being backed up.
• d. Set up a cron job to make a daily incremental backup of the system.
  • It should run at 18:00 every day.
  • The files created should be stored under /var/tmp/backup/.
  • Each day's backup should be in a file named with that day's date, such as /var/tmp/backup/2017/0111.tgz