Unterabschnitte


Automated Disk Backup

If you manage five or ten machines and have a nice tape backup, you don't need Pools, and you may wonder what they are good for. In this chapter, you will see that Pools can help you optimize disk storage space. The same techniques can be applied to a shop that has multiple tape drives, or that wants to mount various different Volumes to meet their needs.

The rest of this chapter will give an example involving backup to disk Volumes, but most of the information applies equally well to tape Volumes.

The Problem

A site that I administer (a charitable organization) had a tape DDS-3 tape drive that was failing. The exact reason for the failure is still unknown. Worse yet, their full backup size is about 15GB whereas the capacity of their broken DDS-3 was at best 8GB (rated 6/12). A new DDS-4 tape drive and the necessary cassettes was more expensive than their budget could handle.

The Solution

They want to maintain six months of backup data, and be able to access the old files on a daily basis for a week, a weekly basis for a month, then monthly for six months. In addition, offsite capability was not needed (well perhaps it really is, but it was never used). Their daily changes amount to about 300MB on the average, or about 2GB per week.

As a consequence, the total volume of data they need to keep to meet their needs is about 100GB (15GB x 6 + 2GB x 5 + 0.3 x 7) = 102.1GB.

The chosen solution was to buy a 120GB hard disk for next to nothing -- far less than 1/10th the price of a tape drive and the cassettes to handle the same amount of data, and to have Bacula write to disk files.

The rest of this chapter will explain how to setup Bacula so that it would automatically manage a set of disk files with the minimum sysadmin intervention. The system has been running since 22 January 2004 until today (23 June 2007) with no intervention, with the exception of adding a second 120GB hard disk after a year because their needs grew over that time to more than the 120GB (168GB to be exact). The only other intervention I have made is a periodic (about once a year) Bacula upgrade.

Overall Design

Getting Bacula to write to disk rather than tape in the simplest case is rather easy, and is documented in the previous chapter. In addition, all the directives discussed here are explained in that chapter. We'll leave it to you to look at the details there. If you haven't read it and are not familiar with Pools, you probably should at least read it once quickly for the ideas before continuing here.

One needs to consider about what happens if we have only a single large Bacula Volume defined on our hard disk. Everything works fine until the Volume fills, then Bacula will ask you to mount a new Volume. This same problem applies to the use of tape Volumes if your tape fills. Being a hard disk and the only one you have, this will be a bit of a problem. It should be obvious that it is better to use a number of smaller Volumes and arrange for Bacula to automatically recycle them so that the disk storage space can be reused. The other problem with a single Volume, is that until version 2.0.0, Bacula did not seek within a disk Volume, so restoring a single file can take more time than one would expect.

As mentioned, the solution is to have multiple Volumes, or files on the disk. To do so, we need to limit the use and thus the size of a single Volume, by time, by number of jobs, or by size. Any of these would work, but we chose to limit the use of a single Volume by putting a single job in each Volume with the exception of Volumes containing Incremental backup where there will be 6 jobs (a week's worth of data) per volume. The details of this will be discussed shortly. This is a single client backup, so if you have multiple clients you will need to multiply those numbers by the number of clients, or use a different system for switching volumes, such as limiting the volume size.

The next problem to resolve is recycling of Volumes. As you noted from above, the requirements are to be able to restore monthly for 6 months, weekly for a month, and daily for a week. So to simplify things, why not do a Full save once a month, a Differential save once a week, and Incremental saves daily. Now since each of these different kinds of saves needs to remain valid for differing periods, the simplest way to do this (and possibly the only) is to have a separate Pool for each backup type.

The decision was to use three Pools: one for Full saves, one for Differential saves, and one for Incremental saves, and each would have a different number of volumes and a different Retention period to accomplish the requirements.

Full Pool

Putting a single Full backup on each Volume, will require six Full save Volumes, and a retention period of six months. The Pool needed to do that is:

Pool {
  Name = Full-Pool
  Pool Type = Backup
  Recycle = yes
  AutoPrune = yes
  Volume Retention = 6 months
  Maximum Volume Jobs = 1
  Label Format = Full-
  Maximum Volumes = 9
}

Since these are disk Volumes, no space is lost by having separate Volumes for each backup (done once a month in this case). The items to note are the retention period of six months (i.e. they are recycled after six months), that there is one job per volume (Maximum Volume Jobs = 1), the volumes will be labeled Full-0001, ... Full-0006 automatically. One could have labeled these manually from the start, but why not use the features of Bacula.

Six months after the first volume is used, it will be subject to pruning and thus recycling, so with a maximum of 9 volumes, there should always be 3 volumes available (note, they may all be marked used, but they will be marked purged and recycled as needed).

If you have two clients, you would want to set Maximum Volume Jobs to 2 instead of one, or set a limit on the size of the Volumes, and possibly increase the maximum number of Volumes.

Differential Pool

For the Differential backup Pool, we choose a retention period of a bit longer than a month and ensure that there is at least one Volume for each of the maximum of five weeks in a month. So the following works:

Pool {
  Name = Diff-Pool
  Pool Type = Backup
  Recycle = yes
  AutoPrune = yes
  Volume Retention = 40 days
  Maximum Volume Jobs = 1
  Label Format = Diff-
  Maximum Volumes = 10
}

As you can see, the Differential Pool can grow to a maximum of 9 volumes, and the Volumes are retained 40 days and thereafter they can be recycled. Finally there is one job per volume. This, of course, could be tightened up a lot, but the expense here is a few GB which is not too serious.

If a new volume is used every week, after 40 days, one will have used 7 volumes, and there should then always be 3 volumes that can be purged and recycled.

See the discussion above concering the Full pool for how to handle multiple clients.

Incremental Pool

Finally, here is the resource for the Incremental Pool:

Pool {
  Name = Inc-Pool
  Pool Type = Backup
  Recycle = yes
  AutoPrune = yes
  Volume Retention = 20 days
  Maximum Volume Jobs = 6
  Label Format = Inc-
  Maximum Volumes = 7
}

We keep the data for 20 days rather than just a week as the needs require. To reduce the proliferation of volume names, we keep a week's worth of data (6 incremental backups) in each Volume. In practice, the retention period should be set to just a bit more than a week and keep only two or three volumes instead of five. Again, the lost is very little and as the system reaches the full steady state, we can adjust these values so that the total disk usage doesn't exceed the disk capacity.

If you have two clients, the simplest thing to do is to increase the maximum volume jobs from 6 to 12. As mentioned above, it is also possible limit the size of the volumes. However, in that case, you will need to have a better idea of the volume or add sufficient volumes to the pool so that you will be assured that in the next cycle (after 20 days) there is at least one volume that is pruned and can be recycled.

The Actual Conf Files

The following example shows you the actual files used, with only a few minor modifications to simplify things.

The Director's configuration file is as follows:

Director {          # define myself
  Name = bacula-dir
  DIRport = 9101
  QueryFile = "/home/bacula/bin/query.sql"
  WorkingDirectory = "/home/bacula/working"
  PidDirectory = "/home/bacula/working"
  Maximum Concurrent Jobs = 1
  Password = " *** CHANGE ME ***"
  Messages = Standard
}
#   By default, this job will back up to disk in /tmp
Job {
  Name = client
  Type = Backup
  Client = client-fd
  FileSet = "Full Set"
  Schedule = "WeeklyCycle"
  Storage = File
  Messages = Standard
  Pool = Default
  Full Backup Pool = Full-Pool
  Incremental Backup Pool = Inc-Pool
  Differential Backup Pool = Diff-Pool
  Write Bootstrap = "/home/bacula/working/client.bsr"
  Priority = 10
}

# Backup the catalog database (after the nightly save)
Job {
  Name = "BackupCatalog"
  Type = Backup
  Client = client-fd
  FileSet="Catalog"
  Schedule = "WeeklyCycleAfterBackup"
  Storage = File
  Messages = Standard
  Pool = Default
  # This creates an ASCII copy of the catalog
  # WARNING!!! Passing the password via the command line is insecure.
  # see comments in make_catalog_backup for details.
  RunBeforeJob = "/home/bacula/bin/make_catalog_backup bacula bacula"
  # This deletes the copy of the catalog
  RunAfterJob  = "/home/bacula/bin/delete_catalog_backup"
  Write Bootstrap = "/home/bacula/working/BackupCatalog.bsr"
  Priority = 11                   # run after main backup
}

# Standard Restore template, to be changed by Console program
Job {
  Name = "RestoreFiles"
  Type = Restore
  Client = havana-fd
  FileSet="Full Set"
  Storage = File
  Messages = Standard
  Pool = Default
  Where = /tmp/bacula-restores
}



# List of files to be backed up
FileSet {
  Name = "Full Set"
  Include = { Options { signature=SHA1; compression=GZIP9 }
    File = /
    File = /usr
    File = /home
    File = /boot
    File = /var
    File = /opt
  }
  Exclude = {
    File = /proc
    File = /tmp
    File = /.journal
    File = /.fsck
    ...
  }
}
Schedule {
  Name = "WeeklyCycle"
  Run = Level=Full 1st sun at 2:05
  Run = Level=Differential 2nd-5th sun at 2:05
  Run = Level=Incremental mon-sat at 2:05
}

# This schedule does the catalog. It starts after the WeeklyCycle
Schedule {
  Name = "WeeklyCycleAfterBackup"
  Run = Level=Full sun-sat at 2:10
}

# This is the backup of the catalog
FileSet {
  Name = "Catalog"
  Include { Options { signature=MD5 }
    File = /home/bacula/working/bacula.sql
  }
}

Client {
  Name = client-fd
  Address = client
  FDPort = 9102
  Catalog = MyCatalog
  Password = " *** CHANGE ME ***"
  AutoPrune = yes      # Prune expired Jobs/Files
  Job Retention = 6 months
  File Retention = 60 days
}

Storage {
  Name = File
  Address = localhost
  SDPort = 9103
  Password = " *** CHANGE ME ***"
  Device = FileStorage
  Media Type = File
}

Catalog {
  Name = MyCatalog
  dbname = bacula; user = bacula; password = ""
}

Pool {
  Name = Full-Pool
  Pool Type = Backup
  Recycle = yes           # automatically recycle Volumes
  AutoPrune = yes         # Prune expired volumes
  Volume Retention = 6 months
  Maximum Volume Jobs = 1
  Label Format = Full-
  Maximum Volumes = 9
}

Pool {
  Name = Inc-Pool
  Pool Type = Backup
  Recycle = yes           # automatically recycle Volumes
  AutoPrune = yes         # Prune expired volumes
  Volume Retention = 20 days
  Maximum Volume Jobs = 6
  Label Format = Inc-
  Maximum Volumes = 7
}

Pool {
  Name = Diff-Pool
  Pool Type = Backup
  Recycle = yes
  AutoPrune = yes
  Volume Retention = 40 days
  Maximum Volume Jobs = 1
  Label Format = Diff-
  Maximum Volumes = 10
}

Messages {
  Name = Standard
  mailcommand = "bsmtp -h mail.domain.com -f \"\(Bacula\) %r\"
      -s \"Bacula: %t %e of %c %l\" %r"
  operatorcommand = "bsmtp -h mail.domain.com -f \"\(Bacula\) %r\"
      -s \"Bacula: Intervention needed for %j\" %r"
  mail = root@domain.com = all, !skipped
  operator = root@domain.com = mount
  console = all, !skipped, !saved
  append = "/home/bacula/bin/log" = all, !skipped
}

and the Storage daemon's configuration file is:

Storage {               # definition of myself
  Name = bacula-sd
  SDPort = 9103       # Director's port
  WorkingDirectory = "/home/bacula/working"
  Pid Directory = "/home/bacula/working"
}
Director {
  Name = bacula-dir
  Password = " *** CHANGE ME ***"
}
Device {
  Name = FileStorage
  Media Type = File
  Archive Device = /files/bacula
  LabelMedia = yes;    # lets Bacula label unlabeled media
  Random Access = Yes;
  AutomaticMount = yes;   # when device opened, read it
  RemovableMedia = no;
  AlwaysOpen = no;
}
Messages {
  Name = Standard
  director = bacula-dir = all
}

eric 2009-05-06