Initial revision

[bacula/docs] / docs / manual / strategies.tex

%%
%%

\section*{Backup Strategies}
\label{_ChapterStart3}
\index[general]{Strategies!Backup }
\index[general]{Backup Strategies }
\addcontentsline{toc}{section}{Backup Strategies}

Although Recycling and Backing Up to Disk Volume have been discussed in
previous chapters, this chapter is meant to give you an overall view of
possible backup strategies and to explain their advantages and disadvantages. 
\label{Simple}

\subsection*{Simple One Tape Backup}
\index[general]{Backup!Simple One Tape }
\index[general]{Simple One Tape Backup }
\addcontentsline{toc}{subsection}{Simple One Tape Backup}

Probably the simplest strategy is to back everything up to a single tape and
insert a new (or recycled) tape when it fills and Bacula requests a new one. 

\subsubsection*{Advantages}
\index[general]{Advantages }
\addcontentsline{toc}{subsubsection}{Advantages}

\begin{itemize}
\item The operator intervenes only when a tape change is needed.  (once a
   month at my site).  
\item There is little chance of operator error because the tape  is not
   changed daily.  
\item A minimum number of tapes will be needed for a full restore.  Typically
   the best case will be one tape and worst two.  
\item You can easily arrange for the Full backup to occur a different  night
   of the month for each system, thus load balancing and  shortening the backup
   time. 
\end{itemize}

\subsubsection*{Disadvantages}
\index[general]{Disadvantages }
\addcontentsline{toc}{subsubsection}{Disadvantages}

\begin{itemize}
\item If your site burns down, you will lose your current backups,  and in my
   case about a month of data.  
\item After a tape fills and you have put in a blank tape, the  backup will
   continue, and this will generally happen during  working hours. 
   \end{itemize}

\subsubsection*{Practical Details}
\index[general]{Details!Practical }
\index[general]{Practical Details }
\addcontentsline{toc}{subsubsection}{Practical Details}

This system is very simple. When the tape fills and Bacula requests a new
tape, you {\bf unmount} the tape from the Console program, insert a new tape
and {\bf label} it. In most cases after the label, Bacula will automatically
mount the tape and resume the backup. Otherwise, you simply {\bf mount} the
tape. 

Using this strategy, one typically does a Full backup once a week following by
daily Incremental backups. To minimize the amount of data written to the tape,
once can do (as I do) a Full backup once a month on the first Sunday of the
month, a Differential backup on the 2nd-5th Sunday of the month, and
incremental backups the rest of the week. 
\label{Manual}

\subsection*{Manually Changing Tapes}
\index[general]{Tapes!Manually Changing }
\index[general]{Manually Changing Tapes }
\addcontentsline{toc}{subsection}{Manually Changing Tapes}

If you use the strategy presented above, Bacula will ask you to change the
tape, and you will {\bf unmount} it and the remount it when you have inserted
the new tape. 

If you do not wish to interact with Bacula to change each tape, there are
several ways to get Bacula to release the tape: 

\begin{itemize}
\item In your Storage daemon's Device resource, set
   {\bf AlwaysOpen = no}
   In this case, Bacula will release the tape after  every job. If you run
several jobs, the tape will be  rewound and repositioned to the end at the
beginning  of every job. This is not very efficient, but does let  you change
the tape whenever you want.  
\item Use a {\bf RunAfterJob} statement to run a script after  your last job.
   This could also be an {\bf Admin}  job that runs after all your backup jobs.
   The script could be  something like:  

\footnotesize
\begin{verbatim}
      #!/bin/sh
      /full-path/console -c /full-path/console.conf <<END_OF_DATA
      release storage=your-storage-name
      END_OF_DATA
      
\end{verbatim}
\normalsize

In this example, you would have {\bf AlwaysOpen=yes},  but the {\bf release}
command would tell Bacula to  rewind the tape and on the next job assume the
tape  has changed. This strategy may not work on some systems,  or on
autochangers because Bacula will still keep the  drive open.  
\item The final strategy is the similar to the previous  case except that you
   would use the unmount command  to force Bacula to release the drive. Then you
   would  eject the tape, and remount it as follows:  

\footnotesize
\begin{verbatim}
      #!/bin/sh
      /full-path/console -c /full-path/console.conf <\&ltEND_OF_DATA
      unmount storage=your-storage-name
      END_OF_DATA
      # the following is a shell command
      mt eject
      /full-path/console -c /full-path/console.conf <<END_OF_DATA
      mount storage=your-storage-name
      END_OF_DATA
      
\end{verbatim}
\normalsize

\end{itemize}

\label{Daily}

\subsection*{Daily Tape Rotation}
\index[general]{Rotation!Daily Tape }
\index[general]{Daily Tape Rotation }
\addcontentsline{toc}{subsection}{Daily Tape Rotation}

This scheme is quite different from the one mentioned above in that a Full
backup is done to a different tape every day of the week. Generally, the
backup will cycle continuously through 5 or 6 tapes each week. Variations are
to use a different tape each Friday, and possibly at the beginning of the
month. Thus if backups are done Monday through Friday only, you need only 5
tapes, and by having two Friday tapes, you need a total of 6 tapes. Many sites
run this way, or using modifications of it based on two week cycles or longer.


\subsubsection*{Advantages}
\index[general]{Advantages }
\addcontentsline{toc}{subsubsection}{Advantages}

\begin{itemize}
\item All the data is stored on a single tape, so recoveries  are simple and
   faster.  
\item Assuming the previous day's tape is taken offsite each  day, a maximum
   of one days data will be lost if the  site burns down. 
   \end{itemize}

\subsubsection*{Disadvantages}
\index[general]{Disadvantages }
\addcontentsline{toc}{subsubsection}{Disadvantages}

\begin{itemize}
\item The tape must be changed every day requiring a lot of  operator
   intervention.  
\item More errors will occur because of human mistakes.  
\item If the wrong tape is inadvertently mounted, the Backup for  that day
   will not occur exposing the system to data loss.  
\item There is much more movement of the tape each day (rewinds)  leading to
   shorter tape drive life time.  
\item Initial setup of Bacula to run in this mode is more complicated  than
   the Single tape system described above.  
\item Depending on the number of systems you have and their data  capacity, it
   may not be possible to do a Full backup every  night for time reasons or
   reasons of tape capacity. 
\end{itemize}

\subsubsection*{Practical Details}
\index[general]{Details!Practical }
\index[general]{Practical Details }
\addcontentsline{toc}{subsubsection}{Practical Details}

The simplest way to ``force'' Bacula to use a different tape each day is to
define a different Pool for each day of the the week a backup is done. In
addition, you will need to specify appropriate Job and File retention periods
so that Bacula will relabel and overwrite the tape each week rather than
appending to it. Nic Bellamy has supplied an actual working model of this
which we include here. 

What is important is to create a different Pool for each day of the week, and
on the {\bf run} statement in the Schedule, to specify which Pool is to be
used. He has one Schedule that accomplishes this, and a second Schedule that
does the same thing for the Catalog backup run each day after the main backup
(Priorities were not available when this script was written). In addition, he
uses a {\bf Max Start Delay} of 22 hours so that if the wrong tape is
premounted by the operator, the job will be automatically canceled, and the
backup cycle will re-synchronize the next day. He has named his Friday Pool
{\bf WeeklyPool} because in that Pool, he wishes to have several tapes to be
able to restore to a time older than one week. 

And finally, in his Storage daemon's Device resource, he has {\bf Automatic
Mount = yes} and {\bf Always Open = No}. This is necessary for the tape
ejection to work in his {\bf end\_of\_backup.sh} script below. 

For example, his bacula-dir.conf file looks like the following: 

\footnotesize
\begin{verbatim}
 
# /etc/bacula/bacula-dir.conf
#
# Bacula Director Configuration file
#
Director {
  Name = ServerName
  DIRport = 9101
  QueryFile = "/etc/bacula/query.sql"
  WorkingDirectory = "/var/lib/bacula"
  PidDirectory = "/var/run"
  SubSysDirectory = "/var/lock/subsys"
  Maximum Concurrent Jobs = 1
  Password = "console-pass"
  Messages = Standard
}
#
# Define the main nightly save backup job
#
Job {
  Name = "NightlySave"
  Type = Backup
  Client = ServerName
  FileSet = "Full Set"
  Schedule = "WeeklyCycle"
  Storage = Tape
  Messages = Standard
  Pool = Default
  Write Bootstrap = "/var/lib/bacula/NightlySave.bsr"
  Max Start Delay = 22h
}
# Backup the catalog database (after the nightly save)
Job {
  Name = "BackupCatalog"
  Type = Backup
  Client = ServerName
  FileSet = "Catalog"
  Schedule = "WeeklyCycleAfterBackup"
  Storage = Tape
  Messages = Standard
  Pool = Default
 # This creates an ASCII copy of the catalog
  RunBeforeJob = "/usr/lib/bacula/make_catalog_backup -u bacula"
 # This deletes the copy of the catalog, and ejects the tape
  RunAfterJob  = "/etc/bacula/end_of_backup.sh"
  Write Bootstrap = "/var/lib/bacula/BackupCatalog.bsr"
  Max Start Delay = 22h
}
# Standard Restore template, changed by Console program
Job {
  Name = "RestoreFiles"
  Type = Restore
  Client = ServerName
  FileSet = "Full Set"
  Storage = Tape
  Messages = Standard
  Pool = Default
  Where = /tmp/bacula-restores
}
# List of files to be backed up
FileSet {
  Name = "Full Set"
  Include = signature=MD5 {
    /
    /data
  }
  Exclude = { /proc /tmp /.journal }
}
#
# When to do the backups
#
Schedule {
  Name = "WeeklyCycle"
  Run = Level=Full Pool=MondayPool Monday at 8:00pm
  Run = Level=Full Pool=TuesdayPool Tuesday at 8:00pm
  Run = Level=Full Pool=WednesdayPool Wednesday at 8:00pm
  Run = Level=Full Pool=ThursdayPool Thursday at 8:00pm
  Run = Level=Full Pool=WeeklyPool Friday at 8:00pm
}
# This does the catalog. It starts after the WeeklyCycle
Schedule {
  Name = "WeeklyCycleAfterBackup"
  Run = Level=Full Pool=MondayPool Monday at 8:15pm
  Run = Level=Full Pool=TuesdayPool Tuesday at 8:15pm
  Run = Level=Full Pool=WednesdayPool Wednesday at 8:15pm
  Run = Level=Full Pool=ThursdayPool Thursday at 8:15pm
  Run = Level=Full Pool=WeeklyPool Friday at 8:15pm
}
# This is the backup of the catalog
FileSet {
  Name = "Catalog"
  Include = signature=MD5 {
     /var/lib/bacula/bacula.sql
  }
}
# Client (File Services) to backup
Client {
  Name = ServerName
  Address = dionysus
  FDPort = 9102
  Catalog = MyCatalog
  Password = "client-pass"
  File Retention = 30d
  Job Retention = 30d
  AutoPrune = yes
}
# Definition of file storage device
Storage {
  Name = Tape
  Address = dionysus
  SDPort = 9103
  Password = "storage-pass"
  Device = Tandberg
  Media Type = MLR1
}
# Generic catalog service
Catalog {
  Name = MyCatalog
  dbname = bacula; user = bacula; password = ""
}
# Reasonable message delivery -- send almost all to email address
#  and to the console
Messages {
  Name = Standard
  mailcommand = "/usr/sbin/bsmtp -h localhost -f \"\(Bacula\) %r\"
     -s \"Bacula: %t %e of %c %l\" %r"
  operatorcommand = "/usr/sbin/bsmtp -h localhost -f \"\(Bacula\) %r\"
     -s \"Bacula: Intervention needed for %j\" %r"
  mail = root@localhost = all, !skipped
  operator = root@localhost = mount
  console = all, !skipped, !saved
  append = "/var/lib/bacula/log" = all, !skipped
}
    
# Pool definitions
#
# Default Pool for jobs, but will hold no actual volumes
Pool {
  Name = Default
  Pool Type = Backup
}
Pool {
  Name = MondayPool
  Pool Type = Backup
  Recycle = yes
  AutoPrune = yes
  Volume Retention = 6d
  Accept Any Volume = yes
  Maximum Volume Jobs = 2
}
Pool {
  Name = TuesdayPool
  Pool Type = Backup
  Recycle = yes
  AutoPrune = yes
  Volume Retention = 6d
  Accept Any Volume = yes
  Maximum Volume Jobs = 2
}
Pool {
  Name = WednesdayPool
  Pool Type = Backup
  Recycle = yes
  AutoPrune = yes
  Volume Retention = 6d
  Accept Any Volume = yes
  Maximum Volume Jobs = 2
}
Pool {
  Name = ThursdayPool
  Pool Type = Backup
  Recycle = yes
  AutoPrune = yes
  Volume Retention = 6d
  Accept Any Volume = yes
  Maximum Volume Jobs = 2
}
Pool {
  Name = WeeklyPool
  Pool Type = Backup
  Recycle = yes
  AutoPrune = yes
  Volume Retention = 12d
  Accept Any Volume = yes
  Maximum Volume Jobs = 2
}
# EOF
\end{verbatim}
\normalsize

Note, the mailcommand and operatorcommand should be on a single line each.
They were split to preserve the proper page width. In order to get Bacula to
release the tape after the nightly backup, he uses a {\bf RunAfterJob} script
that deletes the ASCII copy of the database back and then rewinds and ejects
the tape. The following is a copy of {\bf end\_of\_backup.sh} 

\footnotesize
\begin{verbatim}
#! /bin/sh
/usr/lib/bacula/delete_catalog_backup
mt rewind
mt eject
exit 0
\end{verbatim}
\normalsize

Finally, if you list his Volumes, you get something like the following: 

\footnotesize
\begin{verbatim}
*list media
Using default Catalog name=MyCatalog DB=bacula
Pool: WeeklyPool
+-----+-----------+-------+--------+-----------+-----------------+-------+------+
| MeId| VolumeName| MedTyp| VolStat| VolBytes  | LastWritten     | VolRet| Recyc|
+-----+-----------+-------+--------+-----------+-----------------+-------+------+
| 5   | Friday_1  | MLR1  | Used   | 2157171998| 2003-07-11 20:20| 103680| 1    |
| 6   | Friday_2  | MLR1  | Append | 0         | 0               | 103680| 1    |
+-----+-----------+-------+--------+-----------+-----------------+-------+------+
Pool: MondayPool
+-----+-----------+-------+--------+-----------+-----------------+-------+------+
| MeId| VolumeName| MedTyp| VolStat| VolBytes  | LastWritten     | VolRet| Recyc|
+-----+-----------+-------+--------+-----------+-----------------+-------+------+
| 2   | Monday    | MLR1  | Used   | 2260942092| 2003-07-14 20:20| 518400| 1    |
+-----+-----------+-------+--------+-----------+-----------------+-------+------+
Pool: TuesdayPool
+-----+-----------+-------+--------+-----------+-----------------+-------+------+
| MeId| VolumeName| MedTyp| VolStat| VolBytes  | LastWritten     | VolRet| Recyc|
+-----+-----------+-------+--------+-----------+-----------------+-------+------+
| 3   | Tuesday   | MLR1  | Used   | 2268180300| 2003-07-15 20:20| 518400| 1    |
+-----+-----------+-------+--------+-----------+-----------------+-------+------+
Pool: WednesdayPool
+-----+-----------+-------+--------+-----------+-----------------+-------+------+
| MeId| VolumeName| MedTyp| VolStat| VolBytes  | LastWritten     | VolRet| Recyc|
+-----+-----------+-------+--------+-----------+-----------------+-------+------+
| 4   | Wednesday | MLR1  | Used   | 2138871127| 2003-07-09 20:2 | 518400| 1    |
+-----+-----------+-------+--------+-----------+-----------------+-------+------+
Pool: ThursdayPool
+-----+-----------+-------+--------+-----------+-----------------+-------+------+
| MeId| VolumeName| MedTyp| VolStat| VolBytes  | LastWritten     | VolRet| Recyc|
+-----+-----------+-------+--------+-----------+-----------------+-------+------+
| 1   | Thursday  | MLR1  | Used   | 2146276461| 2003-07-10 20:50| 518400| 1    |
+-----+-----------+-------+--------+-----------+-----------------+-------+------+
Pool: Default
No results to list.
\end{verbatim}
\normalsize

Note, I have truncated a number of the columns so that the information fits on
the width of a page.