BaculaFS 0.1.7

Bacula Filesystem in USErspace

  Add it to your Download Basket!

 Add it to your Watch List!


Rate it!

What's new in BaculaFS 0.1.7:

  • Added batch extract mode.
Read full changelog
send us
an update
GPL v3 
Avi Rozen
ROOT \ System \ Filesystems
BaculaFS is a filesystem that exposes the Bacula catalog and storage as a Filesystem in USErspace (FUSE).

BaculaFS is a tool, independent of Bacula, that represents the Bacula catalog and backup storage media as a read-only filesystem in userspace.

BaculaFS is specifically designed to cater for the following use-cases:

* maintaining a remote snapshot of the files in the backup storage using rsync
* auditing the contents of backup jobs, without resorting to SQL queries
* comparing backup jobs (using several mount points)


Extract the source code archive to a temporary directory, cd to this directory and run

python install


baculafs [mountpoint] [options]

 --version show program's version number and exit
 -h, --help show this help message and exit
 -o opt,[opt...] mount options
 -o driver=mysql|postgresql|sqlite|sqlite3
 database driver [default: sqlite3]
 -o host=HOST database server address [default: localhost]
 -o port=PORT database server port
 -o database=PATH database name [default: bacula]
 -o username=USERNAME database user name [default: bacula]
 -o password=PASSWORD database password
 -o conf=PATH storage daemon configuration file [default:
 -o client=CLIENT file daemon name
 -o fileset=FILESET backup fileset
 -o device=DEVICE storage device name [default: FileStorage]
 -o datetime='YYYY-MM-DD hh:mm:ss'
 snapshot date/time [default: now]
 -o recent_job select contents of most recent job only [default:
 -o joblist='JOBID1 JOBID2 ...'
 select contents of specified list of jobs
 -o cleanup clean cache directory upon umount [default: False]
 -o move_root make absolute path symlinks point to path under
 mount point [default: False]
 -o prefetch_attrs read and parse attributes for all files upon
 filesystem initialization [default: False]
 -o prefetch_symlinks extract all symbolic links upon filesystem
 initialization (implies prefetch_attrs) [default:
 -o prefetch_regex=REGEX
 extract all objects that match REGEX upon
 filesystem initialization (implies prefetch_attrs)
 -o prefetch_recent extract contents of most recent non-full job upon
 filesystem initialization (implies
 prefetch_symlinks) [default: False]
 -o prefetch_diff=PATH extract files that do not match files at PATH
 (hint: speeds up rsync; implies prefetch_symlinks)
 -o prefetch_everything
 extract everything upon filesystem initialization
 (complete restore to cache) [default: False]
 -o user_cache_path=PATH
 user specified cache path (hint: combine this with
 one of the prefetch options) [default: none]
 -o logging=debug|info|warning|critical|error
 logging level [default: warning]
 -o syslog log to both syslog and console [default: False]

FUSE options:
 -d -o debug enable debug output (implies -f)
 -f foreground operation
 -s disable multi-threaded operation

 -o allow_other allow access to other users
 -o allow_root allow access to root
 -o nonempty allow mounts over non-empty file/dir
 -o default_permissions enable permission checking by kernel
 -o fsname=NAME set filesystem name
 -o subtype=NAME set filesystem type
 -o large_read issue large read requests (2.4 only)
 -o max_read=N set maximum size of read requests

 -o hard_remove immediate removal (don't hide files)
 -o use_ino let filesystem set inode numbers
 -o readdir_ino try to fill in d_ino in readdir
 -o direct_io use direct I/O
 -o kernel_cache cache files in kernel
 -o [no]auto_cache enable caching based on modification times (off)
 -o umask=M set file permissions (octal)
 -o uid=N set file owner
 -o gid=N set file group
 -o entry_timeout=T cache timeout for names (1.0s)
 -o negative_timeout=T cache timeout for deleted names (0.0s)
 -o attr_timeout=T cache timeout for attributes (1.0s)
 -o ac_attr_timeout=T auto cache timeout for attributes (attr_timeout)
 -o intr allow requests to be interrupted
 -o intr_signal=NUM signal to send on interrupt (10)
 -o modules=M1[:M2...] names of modules to push onto filesystem stack

 -o max_write=N set maximum size of write requests
 -o max_readahead=N set maximum readahead
 -o async_read perform reads asynchronously (default)
 -o sync_read perform reads synchronously
 -o atomic_o_trunc enable atomic open+truncate support
 -o big_writes enable larger than 4kB writes
 -o no_remote_lock disable remote file locking

Module options:

 -o subdir=DIR prepend this directory to all paths (mandatory)
 -o [no]rellinks transform absolute symlinks to relative

 -o from_code=CHARSET original encoding of file names (default: UTF-8)
 -o to_code=CHARSET new encoding of the file names (default: UTF-8)



BaculaFS starts by running several SQL queries against the Bacula catalog. This is done to determine the list of files that belong to the most recent backup for a given client and fileset.

BaculaFS can also be told to represent a backup snapshot corresponding to a specified date and time, or a list of backup job ids.

Following this, BaculaFS may run bextract once to prefetch and cache symbolic links and actual file contents, depending on user specified command line options.

At this point the filesystem is ready.


Opening a file for reading causes BaculaFS to run bextract in order to extract the file from the storage device. If this operation succeeds, the file is cached for subsequent read operations.

Bacula storage is not designed for random access file retrieval, so it is important to select a suitable cache prefetch strategy beforehand. Running bextract once, during filesystem initialization, to extract several files, is much more efficient than running it several times to extract each individual file, when accessed at a later stage.

For example:

 * use -o prefetch_attrs for storage space usage analysis (e.g. with Baobab)
 * use -o prefetch_symlinks for any manual filesystem traversal with command line or GUI tools (find, mc, etc.)
 * use -o prefetch_diff with rsync

The cache may be cleaned up automatically upon un-mounting the filesystem, with -o cleanup. It may also be reused between mount operations with -o user_cache_path.


BaculaFS uses extended file attributes to expose Bacula specific information for each file in the filesystem. These extended attributes are all grouped in the user.baculafs namespace:


The root directory has several more attributes, that expose filesystem instance-specific information:


and several more attributes for monitoring the file extraction process:



If the storage device is a tape drive then it's possible that BaculaFS will attempt to retrieve a file from a volume that's on an unmounted tape. BaculaFS will then set user.baculafs.bextract.state to *user intervention required*, and will wait for user intervention.

The user should then mount the tape containing the missing volume and set the state to run, to make BaculaFS retry the operation:

attr -s baculafs.bextract.state -V run < mount-point >

Please note that this feature has undergone only rudimentary testing. Expect breakage.

Last updated on January 3rd, 2011

requirements limitations

#Bacula filesystem #FUSE filesystem #Bacula catalog #Bacula #filesystem #FUSE #catalog

Add your review!