Subsections

New Features in 9.0.0

Note: The first beta versions are released as version 7.9.0, and the first production release will be 9.0.0.

Maximum Virtual Full Interval Option

Two new director directives have been added:

   Max Virtual Full Interval
and
   Virtual Full Backup Pool

The Max Virtual Full Interval directive should behave similar to the Max Full Interval, but for Virtual Full jobs. If Bacula sees that there has not been a Full backup in Max Virtual Full Interval time then it will upgrade the job to Virtual Full. If you have both Max Full Interval and Max Virtual Full Interval set then Max Full Interval should take precedence.

The Virtual Full Backup Pool directive allows one to change the pool as well. You probably want to use these two directives in conjunction with each other but that may depend on the specifics of one's setup. If you set the Max Full Interval without setting Max Virtual Full Interval then Bacula will use whatever the "default" pool is set to which is the same behavior as with the Max Full Interval.

Progressive Virtual Full

In Bacula version 9.0.0, we have added a new Directive named Backups To Keep that permits you to implement Progressive Virtual Fulls within Bacula. Sometimes this feature is known as Incremental Forever with Consolidation.

Figure: Backup Sequence Slides Forward One Day, Each Day
Image pvf-slidingbackups

To implement the Progressive Virtual Full feature, simply add the Backups To Keep directive to your Virtual Full backup Job resource. The value specified on the directive indicates the number of backup jobs that should not be merged into the Virtual Full (i.e. the number of backup jobs that should remain after the Virtual Full has completed. The default is zero, which reverts to a standard Virtual Full than consolidates all the backup jobs that it finds.

Backups To Keep Directive

The new BackupsToKeep directive is specified in the Job Resource and has the form:

   Backups To Keep = 30

where the value (30 in the above figure and example) is the number of backups to retain. When this directive is present during a Virtual Full (it is ignored for other Job types), it will look for the most recent Full backup that has more subsequent backups than the value specified. In the above example the Job will simply terminate unless there is a Full back followed by at least 31 backups of either level Differential or Incremental.

Assuming that the last Full backup is followed by 32 Incremental backups, a Virtual Full will be run that consolidates the Full with the first two Incrementals that were run after the Full. The result is that you will end up with a Full followed by 30 Incremental backups. The Job Resource in bacula-dir.conf to accomplish this would be:

   Job {
     Name = "VFull"
     Type = Backup
     Level = VirtualFull
     Client = "my-fd"
     File Set = "FullSet"
     Accurate = Yes
     Backups To Keep = 10
   }

Delete Consolidated Jobs

The new directive Delete Consolidated Jobs expects a yes or no value that if set to yes will cause any old Job that is consolidated during a Virtual Full to be deleted. In the example above we saw that a Full plus one other job (either an Incremental or Differential) were consolidated into a new Full backup. The original Full plus the other Job consolidated will be deleted. The default value is no.

Virtual Full Compatibility

Virtual Full as well as Progressive Virtual Full works with any standard backup Job.

However, it should be noted that Virtual Full jobs are not compatible with any plugins that you may be using.

TapeAlert Enhancements

There are some significant enhancements to the TapeAlert feature of Bacula. Several directives are used slightly differently, which unfortunately causes a compatibility problem with the old TapeAlert implementation. Consequently, if you are already using TapeAlert, you must modify your bacula-sd.conf in order for Tape Alerts to work. See below for the details ...

What is New

First, you must define a Alert Command directive in the Device resource that calls the new tapealert script that is installed in the scripts directory (normally: /opt/bacula/scripts). It is defined as follows:

Device {
  Name = ...
  Archive Device = /dev/nst0
  Alert Command = "/opt/bacula/scripts/tapealert %l"
  Control Device = /dev/sg1 # must be SCSI ctl for /dev/nst0
  ...
}

In addition the Control Device directive in the Storage Daemon's conf file must be specified in each Device resource to permit Bacula to detect tape alerts on a specific devices (normally only tape devices).

Once the above mentioned two directives (Alert Command and Control Device) are in place in each of your Device resources, Bacula will check for tape alerts at two points:

At each of the above times, Bacula will call the new tapealert script, which uses the tapeinfo program. The tapeinfo utility is part of the apt sg3-utils and rpm sg3_utils packages that must be installed on your systems. Then after each alert that Bacula finds for that drive, Bacula will emit a Job message that is either INFO, WARNING, or FATAL depending on the designation in the Tape Alert published by the T10 Technical Committee on SCSI Storage Interfaces (www.t10.org). For the specification, please see: www.t10.org/ftp/t10/document.02/02-142r0.pdf

As a somewhat extreme example, if tape alerts 3, 5, and 39 are set, you will get the following output in your backup job.

     17-Nov 13:37 rufus-sd JobId 1: Error: block.c:287 
     Write error at 0:17 on device "tape"
     (/home/kern/bacula/k/regress/working/ach/drive0)
     Vol=TestVolume001. ERR=Input/output error.

     17-Nov 13:37 rufus-sd JobId 1: Fatal error: Alert:
     Volume="TestVolume001" alert=3: ERR=The operation has stopped because
     an error has occurred while reading or writing data which the drive
     cannot correct.  The drive had a hard read or write error

     17-Nov 13:37 rufus-sd JobId 1: Fatal error: Alert:
     Volume="TestVolume001" alert=5: ERR=The tape is damaged or the drive
     is faulty.  Call the tape drive supplier helpline.  The drive can no
     longer read data from the tape

     17-Nov 13:37 rufus-sd JobId 1: Warning: Disabled Device "tape"
     (/home/kern/bacula/k/regress/working/ach/drive0) due to tape alert=39.

     17-Nov 13:37 rufus-sd JobId 1: Warning: Alert: Volume="TestVolume001"
     alert=39: ERR=The tape drive may have a fault.  Check for availability
     of diagnostic information and run extended diagnostics if applicable.
     The drive may have had a failure which may be identified by stored
     diagnostic information or by running extended diagnostics (eg Send
     Diagnostic).  Check the tape drive users manual for instructions on
     running extended diagnostic tests and retrieving diagnostic data.

Without the tape alert feature enabled, you would only get the first error message above, which is the error return Bacula received when it gets the error. Notice also, that in the above output the alert number 5 is a critical error, which causes two things to happen. First the tape drive is disabled, and second the Job is failed.

If you attempt to run another Job using the Device that has been disabled, you will get a message similar to the following:

17-Nov 15:08 rufus-sd JobId 2: Warning:
     Device "tape" requested by DIR is disabled.

and the Job may be failed if no other drive can be found.

Once the problem with the tape drive has been corrected, you can clear the tape alerts and re-enable the device with the Bacula bconsole command such as the following:

  enable Storage=Tape

Note, when you enable the device, the list of prior tape alerts for that drive will be discarded.

Since is is possible to miss tape alerts, Bacula maintains a temporary list of the last 8 alerts, and each time Bacula calls the tapealert script, it will keep up to 10 alert status codes. Normally there will only be one or two alert errors for each call to the tapealert script.

Once a drive has one or more tape alerts, you can see them by using the bconsole status command as follows:
status storage=Tape
which produces the following output:
Device Vtape is "tape" (/home/kern/bacula/k/regress/working/ach/drive0)
mounted with:
    Volume:      TestVolume001
    Pool:        Default
    Media type:  tape
    Device is disabled. User command.
    Total Bytes Read=0 Blocks Read=1 Bytes/block=0
    Positioned at File=1 Block=0
    Critical Alert: at 17-Nov-2016 15:08:01 Volume="TestVolume001"
       alert=Hard Error
    Critical Alert: at 17-Nov-2016 15:08:01 Volume="TestVolume001"
       alert=Read Failure
    Warning Alert: at 17-Nov-2016 15:08:01 Volume="TestVolume001"
       alert=Diagnostics Required
if you want to see the long message associated with each of the alerts, simply set the debug level to 10 or more and re-issue the status command:
setdebug storage=Tape level=10
status storage=Tape
    ...
    Critical Alert: at 17-Nov-2016 15:08:01 Volume="TestVolume001"
      flags=0x0 alert=The operation has stopped because an error has occurred
       while reading or writing data which the drive cannot correct. The drive had
       a hard read or write error
    Critical Alert: at 17-Nov-2016 15:08:01 Volume="TestVolume001"
       flags=0x0 alert=The tape is damaged or the drive is faulty. Call the tape
       drive supplier helpline.  The drive can no longer read data from the tape
    Warning Alert: at 17-Nov-2016 15:08:01 Volume="TestVolume001" flags=0x1
       alert=The tape drive may have a fault. Check for availability of diagnostic
       information and run extended diagnostics if applicable.   The drive may
       have had a failure which may be identified by stored diagnostic information
       or by running extended diagnostics (eg Send Diagnostic). Check the tape
       drive users manual for instructions on running extended diagnostic tests
       and retrieving diagnostic data.
    ...
The next time you enable the Device by either using bconsole or you restart the Storage Daemon, all the saved alert messages will be discarded.

Handling of Alerts

Tape Alerts numbered 7,8,13,14,20,22,52,53, and 54 will cause Bacula to disable the current Volume.

Tape Alerts numbered 14,20,29,30,31,38, and 39 will cause Bacula to disable the drive.

Please note certain tape alerts such as 14 have multiple effects (disable the Volume and disable the drive).

New Console ACL Directives

By default, if a Console ACL directive is not set, Bacula will assume that the ACL list is empty. If the current Bacula Director configuration uses restricted Consoles and allows restore jobs, it is mandatory to configure the new directives.

DirectoryACL

This directive is used to specify a list of directories that can be accessed by a restore session. Without this directive, a restricted console cannot restore any file. Multiple directories names may be specified by separating them with commas, and/or by specifying multiple DirectoryACL directives. For example, the directive may be specified as:

    DirectoryACL = /home/bacula/, "/etc/", "/home/test/*"

With the above specification, the console can access the following directories:

But not to the following files or directories:

If a directory starts with a Windows pattern (ex: c:/), Bacula will automatically ignore the case when checking directory names.

New Bconsole ``list'' Command Behavior

The bconsole list commands can now be used safely from a restricted bconsole session. The information displayed will respect the ACL configured for the Console session. For example, if a restricted Console has access to JobA, JobB and JobC, information about JobD will not appear in the list jobs command.

New Console ACL Directives

It is now possible to configure a restricted Console to distinguish Backup and Restore job permissions. The BackupClientACL can restrict backup jobs on a specific set of clients, while the RestoreClientACL can restrict restore jobs.

# cat /opt/bacula/etc/bacula-dir.conf
...

Console {
 Name = fd-cons             # Name of the FD Console
 Password = yyy
...
 ClientACL = localhost-fd           # everything allowed
 RestoreClientACL = test-fd         # restore only
 BackupClientACL = production-fd    # backup only
}

The ClientACL directive takes precedence over the RestoreClientACL and the BackupClientACL. In the Console resource resource above, it means that the bconsole linked to the Console named "fd-cons" will be able to run:

At the restore time, jobs for client ``localhost-fd'', ``test-fd'' and ``production-fd'' will be available.

If *all* is set for ClientACL, backup and restore will be allowed for all clients, despite the use of RestoreClientACL or "BackupClientACL.


Client Initiated Backup

A console program such as the new tray-monitor or bconsole can now be configured to connect a File Daemon. There are many new features available (see the New Tray Monitor section below), but probably the most important is the ability for the user to initiate a backup of his own machine. The connection established by the FD to the Director for the backup will be used by the Director for the backup, thus not only can clients (users) initiate backups, but a File Daemon that is NATed (cannot be reached by the Director) can now be backed up without using advanced tunneling techniques providing that the File Daemon can connect to the Director.

The flow of information is shown in the picture below: Image nat

Configuring Client Initiated Backup

In order to ensure security, there are a number of new directives that must be enabled in the new tray-monitor, the File Daemon and in the Director. A typical configuration might look like the following:

# cat /opt/bacula/etc/bacula-dir.conf
...

Console {
 Name = fd-cons             # Name of the FD Console
 Password = yyy

 # These commands are used by the tray-monitor, it is possible to restrict
 CommandACL = run, restore, wait, .status, .jobs, .clients
 CommandACL = .storages, .pools, .filesets, .defaults, .estimate

 # Adapt for your needs
 jobacl = *all*
 poolacl = *all*
 clientacl = *all*
 storageacl = *all*
 catalogacl = *all*
 filesetacl = *all*
}

# cat /opt/bacula/etc/bacula-fd.conf
...

Console {              # Console to connect the Director
  Name = fd-cons
  DIRPort = 9101
  address = localhost
  Password = "yyy"
}

Director {
  Name = remote-cons   # Name of the tray monitor/bconsole
  Password = "xxx"     # Password of the tray monitor/bconsole
  Remote = yes         # Allow to use send commands to the Console defined
}

cat /opt/bacula/etc/bconsole-remote.conf
....

Director {
  Name = localhost-fd
  address = localhost        # Specify the FD address
  DIRport = 9102             # Specify the FD Port
  Password = "notused"
}

Console {
  Name = remote-cons         # Name used in the auth process
  Password = "xxx"
}

cat ~/.bacula-tray-monitor.conf
Monitor {
  Name = remote-cons
}

Client {
  Name = localhost-fd
  address = localhost     # Specify the FD address
  Port = 9102             # Specify the FD Port
  Password = "xxx"
  Remote = yes
}

Image conf-nat Image conf-nat2


A more detailed description with complete examples is available in chapter [*].

New Tray Monitor

A new tray monitor has been added to the 9.0 release, the tray monitor offers the following features:

Figure: Tray Monitor Status
Image tray-monitor-status

Figure: Tray Monitor Client Configuration
Image tray-monitor-conf-fd

Figure: Tray Monitor Run a Job
Image tray-monitor-run1

Image tray-monitor-run2

Schedule Jobs via the Tray Monitor

The Tray Monitor can scan periodically a specific directory ``Command Directory'' and process ``*.bcmd'' files to find jobs to run.

The format of the ``file.bcmd'' command file is the following:

<component name>:<run command>
<component name>:<run command>
...

<component name> = string
<run command>    = string (bconsole command line)

For example:

localhost-fd: run job=backup-localhost-fd level=full
localhost-dir: run job=BackupCatalog

The command file should contain at least one command. The component specified in the first part of the command line should be defined in the tray monitor. Once the command file is detected by the tray monitor, a popup is displayed to the user and it is possible for the user to cancel the job directly.

The file can be created with tools such as ``cron'' or the ``task scheduler'' on Windows. It is possible to verify the network connection at that time to avoid network errors.

#!/bin/sh
if ping -c 1 director &> /dev/null
then
   echo "my-dir: run job=backup" > /path/to/commands/backup.bcmd
fi

Accurate Option for Verify ``Volume Data'' Job

Since Bacula version 8.4.1, it has been possible to have a Verify Job configured with level=Data that will reread all records from a job and optionally check the size and the checksum of all files. Starting with

Bacula version 9.0, it is now possible to use the accurate option to check catalog records at the same time. When using a Verify job with level=Data and accurate=yes can replace the level=VolumeToCatalog option.

For more information on how to setup a Verify Data job, see label:verifyvolumedata.

To run a Verify Job with the accurate option, it is possible to set the option in the Job definition or set use the accurate=yes on the command line.

* run job=VerifyData jobid=10 accurate=yes

FileDaemon Saved Messages Resource Destination

It is now possible to send the list of all saved files to a Messages resource with the saved message type. It is not recommended to send this flow of information to the director and/or the catalog when the client FileSet is pretty large. To avoid side effects, the all keyword doesn't include the saved message type. The saved message type should be explicitely set.

# cat /opt/bacula/etc/bacula-fd.conf
...
Messages {
  Name = Standard
  director = mydirector-dir = all, !terminate, !restored, !saved
  append = /opt/bacula/working/bacula-fd.log = all, saved, restored
}

Minor Enhancements

New Bconsole ".estimate" Command

The new .estimate command can be used to get statistics about a job to run. The command uses the database to approximate the size and the number of files of the next job. On a PostgreSQL database, the command uses regression slope to compute values. On SQLite or MySQL, where these statistical functions are not available, the command uses a simple ``average'' estimation. The correlation number is given for each value.

*.estimate job=backup
level=I
nbjob=0
corrbytes=0
jobbytes=0
corrfiles=0
jobfiles=0
duration=0
job=backup

*.estimate job=backup level=F
level=F
nbjob=1
corrbytes=0
jobbytes=210937774
corrfiles=0
jobfiles=2545
duration=0
job=backup

Traceback and Lockdump

After the reception of a signal, traceback and lockdump information are now stored in the same file.

Bconsole ``list jobs'' command options

The list jobs bconsole command now accepts new command line options:

Minor Enhancements

New Bconsole "Tee All" Command

The ``@tall'' command allows logging all input/output from a console session.

*@tall /tmp/log
*st dir
...

Bconsole ``list jobs'' command options

The list jobs bconsole command now accepts new command line options:

New Bconsole "Tee All" Command

The ``@tall'' command allows logging all input/output from a console session.

*@tall /tmp/log
*st dir
...

New Job Edit Codes %I

In various places such as RunScripts, you have now access to %I to get the JobId of the copy or migration job started by a migrate job.

Job {
  Name = Migrate-Job
  Type = Migrate
  ...
  RunAfter = "echo New JobId is %I"
}

.api version 2

In Bacula version 9.0 and later, we introduced a new .api version to help external tools to parse various Bacula bconsole output.

The api_opts option can use the following arguments:

C
Clear current options
tn
Use a specific time format (1 ISO format, 2 Unix Timestamp, 3 Default Bacula time format)
sn
Use a specific separator between items (new line by default).
Sn
Use a specific separator between objects (new line by default).
o
Convert all keywords to lowercase and convert all non isalpha characters to _

  .api 2 api_opts=t1s43S35
  .status dir running
==================================
jobid=10
job=AJob
...

New Debug Options

In Bacula version 9.0 and later, we introduced a new options parameter for the setdebug bconsole command.

The following arguments to the new option parameter are available to control debug functions.

0
Clear debug flags
i
Turn off, ignore bwrite() errors on restore on File Daemon
d
Turn off decomp of BackupRead() streams on File Daemon
t
Turn on timestamps in traces
T
Turn off timestamps in traces

c
Truncate trace file if trace file is activated

l
Turn on recoding events on P() and V()
p
Turn on the display of the event ring when doing a bactrace

The following command will enable debugging for the File Daemon, truncate an existing trace file, and turn on timestamps when writing to the trace file.

* setdebug level=10 trace=1 options=ct fd

It is now possible to use a class of debug messages called tags to control the debug output of Bacula daemons.

all
Display all debug messages
bvfs
Display BVFS debug messages
sql
Display SQL related debug messages
memory
Display memory and poolmem allocation messages
scheduler
Display scheduler related debug messages

* setdebug level=10 tags=bvfs,sql,memory
* setdebug level=10 tags=!bvfs

# bacula-dir -t -d 200,bvfs,sql

The tags option is composed of a list of tags. Tags are separated by ``,'' or ``+'' or ``-'' or ``!''. To disable a specific tag, use ``-'' or ``!'' in front of the tag. Note that more tags are planned for future versions.

Communication Line Compression

Bacula version 9.0.0 and later now includes communication line compression. It is turned on by default, and if the two Bacula components (Dir, FD, SD, bconsole) are both version 6.6.0 or greater, communication line compression) will be enabled, by default. If for some reason, you do not want communication line compression, you may disable it with the following directive:

Comm Compression = no

This directive can appear in the following resources:

bacula-dir.conf: Director resource
bacula-fd.conf Client (or FileDaemon) resource
bacula-sd.conf: Storage resource
bconsole.conf: Console resource
bat.conf: Console resource

In many cases, the volume of data transmitted across the communications line can be reduced by a factor of three when this directive is enabled (default) In the case that the compression is not effective, Bacula turns it off on a. record by record basis.

If you are backing up data that is already compressed the comm line compression will not be effective, and you are likely to end up with an average compression ratio that is very small. In this case, Bacula reports None in the Job report.

Deduplication Optimized Volumes

This version of Bacula includes a new alternative (or additional) volume format that optimizes the placement of files so that an underlying deduplicating filesystem such as ZFS can optimally deduplicate the backup data that is written by Bacula. These are called Deduplication Optimized Volumes or Aligned Volumes for short. The details of how to use this feature and its considerations are in the Deduplication Optimized Volumes whitepaper.

This feature is available if you have Bacula Community produced binaries and the Aligned Volumes plugin.

baculabackupreport

I have added a new script called baculabackupreport to the scripts directory. This script was written by Bill Arlofski. It prints a backup summary of the backups that occurred in the prior number of hours specified on the command line. You need to edit the first few lines of the file to ensure that your email address is correct and the database type you are using is correct (default is PostgreSQL). Once you do that, you can manually implement it with:

/opt/bacula/scripts/baculabackupreport 24

I have put the above line in my scripts/delete_catalog_backup script so that it will be mailed to me nightly.

New Message Identification Format

We are starting to add unique message indentifiers to each message (other than debug and the Job report) that Bacula prints. At the current time only two files in the Storage Daemon have these message identifiers and over time with subsequent releases we will modify all messages.

The message identifier will be kept unique for each message and once assigned to a message it will not change even if the text of the message changes. This means that the message identifier will be the same no matter what language the text is displayed in, and more importantly, it will allow us to make listing of the messages with in some cases, additional explanation or instructions on how to correct the problem. All this will take several years since it is a lot of work and requires some new programs that are not yet written to manage these message identifiers.

The format of the message identifier is:

   [AAnnnn]
where A is an upper case character and nnnn is a four digit number, where the first charcter indicates the software component (daemon); the second letter indicates the severity, and the number is unique for a given componet and severity.

For example:

   [SF0001]

The first charcter representing the componend at the current time one of the following:

   S      Storage daemon
   D      Director
   F      File daemon

The second character representing the severity or level can be:

   A      Abort 
   F      Fatal
   E      Erropr
   W      Warning
   S      Security
   I      Info
   D      Debug
   O      OK (i.e. operation completed normallly)

So in the example above [SF0001] indicates it is a message id, because of the brackets and because it is at the beginning of the message, and that it was generated by the Storage daemon as a fatal error.

As mentioned above it will take some time to implement these message ids everywhere, and over time we may add more component letters and more severity levels as needed.

Kern Sibbald 2017-07-02