Backup and Restore

1
Backup and Restore
2
References

• Textbook reference
• Nemeth et al 2001, chapter 10
• Limoncelli et al 2002, Chapter 21
• Burgess 2000, ch 10
• Frisch 1995, ch 10.
• Lecture Objectives
• Lecture Questions
• FreeBSD reference Backups

3
Theory

• Everybody loses files from time to time.
  Accidental deletion is one cause, while erroneous
  programs or hardware failures are more serious.
• Backups are a mechanism to reduce the
  inconvenience of small problems and to provide
  survivability when something more serious
  happens, like a disk crash, fire in the machine
  room, flood, sabotage,terrorism, or warfare.

4

• Usually, backup strategy (where one exists at
  all) is based on what used to be done years ago.
• It really is time to consider everything now from
  first principles.
• Backups are like insurance - they represent an
  amount of time and effort expended now to reduce
  time and effort required in the future.
• Future costs might be decreased productivity,
  schedule slippage, or even a total inability to
  continue business.

5
Questions

• Things to consider include
• What files need to be backed up?
• Files which hold only ephemeral, scratch data
  don't need backup.
• Anything in the operating system which is
  identical to the original CD-ROM release doesn't
  need backup.
• However, it might be more convenient to back
  everything up than to exclude certain files or
  directories.

6

• Although the kinds of things to be done to
  increase convenience when a draft letter needs to
  be recreated may differ from those needed when
  reestablishing a business on a new computer after
  a natural disaster, a proper backup strategy will
  handle both.

7

• Where are they?
• Data which is gathered into a single partition or
  directory will be much easier to backup in one
  piece.
• Who will do it?
• Files on an individual's workstation may never be
  backed up if the workstation user has to do the
  job.
• Where, when, and under what conditions will
  backups be done?
• Ideally, backups are done after hours on
  unmounted filesystems. Often this is not
  possible.
• How often do these files change?
• Something changing slowly may only need weekly
  backup, while a busy database may need to be
  backed up several times a day.

8

• How quickly must a file be restored?
• Something on a large medium may take a while to
  find. If it's a licence file controlling your
  everyday use of software, you want to get it very
  quickly.
• It may be possible to order material on the
  medium so that key data is most quickly found.
• Alternatively, you may put such files onto a
  small medium like a floppy, and leave everything
  else on a multi-GB tape.
• Where must the data be restored?
• If you might need to restore onto another
  machine, this may change what software, hardware,
  or recording format you use.

9

• Another question which may control some of your
  decisions is the storage capacity of the media
  used.
• If you are using tapes with a capacity much
  bigger than your disks, you may choose to back
  everything up, even if you don't otherwise need
  to.
• If your medium is smaller than a particular
  partition or file (such as a very large database
  occupying several disks), you may need to do
  attended backups so that somebody can change the
  medium when required.

10
Strategies

• A full backup means that all data on a computer
  (or a disk partition) are backed up.
• This can be time-consuming, and restoring a file
  from such a backup may take as long (attended) as
  the original medium took to write (unattended).

11

• If a full backup takes several tapes (or whatever
  media), a restore can take ages, even if you do
  know which tape the needed file is on.
• Some system will require you mount all tapes in
  order, not just the needed one.

12

• An incremental backup means that files are backed
  up only if they have changed since a previous
  backup.
• Finding the appropriate tape (or whatever) in a
  sequence of incremental backups can also be
  time-consuming.

13
Backup Levels

• There is the concept of a backup level.
• A full backup is level 0, while an incremental
  backup has a higher level, say 1 or 2.
• A level 2 backup contains all files changed since
  the last backup at a lower level (like 0 or 1).

14

• One typical scheme has a full (level 0) backup
  done each Monday, and a level 1 (incremental)
  backup done each other day.
• An alternative has a level 0 backup on the first
  Monday of each month, a level 1 incremental
  backup on each other Monday, and a level 2
  incremental backup every other day.
• This could be extended to include quarterly or
  annual backups.

15

• Whatever plan is chosen, think about how the
  system is used.
• You don't need frequent backup of the parts of
  your system which change infrequently (such as
  the root filesystem and commands).
• You do, however, need current backup of essential
  configuration files.
• Some areas, like /tmp do not need to be backed up
  at all.
• If you provide large scratch areas to users, you
  may decide to make them responsible for backup.

16

• Whether backups must be attended or whether they
  may be unattended depends on a number of things.
• If your tape drive is shared with users, or is in
  heavy use, an unattended backup hogging the
  device all night may be impractical, impolitic,
  or just a plain security hole.

17
Storage of Backups

• It's no use making backups if you can't find the
  media when you need to restore from them.
• Know where things belong, and always keep them
  there. This also applies to non-backup media like
  software distributions, etc.
• Know what is on each medium. You can't restore a
  file if you don't know which tape it's on.

18

• Make routine restoration easy.
• Keep recent backup media available, although not
  necessarily too near the computer room.
• Ideally, the same medium is used consistently for
  the same kind of backup.
• Some places use colour coding, some print
  detailed sticky labels at the end of each backup.
  
• Write-protect media after backup (but not
  during...). You don't want an attempted restore
  to destroy your backup.
• Consider environmental aspects.
• Avoid excess humidity or temperature.
• Avoid dust and similar air polutants.

19

• Keep media some distance from the computers or
  you might lose your backups at the same time as
  the disks.
• Handle media properly.
• Protect media as any other secure item - consider
  theft and vandalism.

20

• For more significant threats (in terms of damage,
  even if not risk) consider long-term and offsite
  storage.
• If your organisation would be unable to function
  without your data, keep such copies.
• These should be full backups, and may be produced
  monthly, or at whatever frequency is appropriate
  to your level of risk and discomfort.
• If you could restart with data a week old, but no
  older, then make copies weekly for offsite
  storage, even if you don't actually move them
  offsite until you make the next one.

21

• These backups should be verified to ensure they
  are good copies, and must be stored far enough
  away from your originals not to be subject to the
  same problems (whether fire, bombs, earthquake,
  etc.).
• Keep them in a locked, fireproof, environmentally
  controlled location.
• Another consideration for such long-term backups
  is the lifetime of your media - you may need to
  copy or replace tapes as often as every year.
• Another time to make full, verified, backups is
  when you're rebuilding a filesystem.

22
Media

• Many media exist suitable for backup
• magnetic tape
• floppy disks
• removable disks
• floptical and magneto-optical disks
• recordable CD-ROMs
• hard disks

23
UNIX Backup Tools

• find
• cpio
• tar
• dd
• dump/restore ufsdump/ufsrestore
• Commercial Products