Bacula Filesystem in USErspace
Copyright © 2009, 2010 Avi Rozen <firstname.lastname@example.org>
BaculaFS is a tool, developed independently 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 or duplicity
- auditing the contents of backup jobs, without resorting to SQL queries
- comparing backup jobs (using several mount points)
Note that BaculaFS is a maintenance tool - its operation may interfere with the normal operation of a live Bacula setup (see the Limitations section below).
BaculaFS has been tested with the following set of required software packages:
- Bacula 5.0.2 with one of the following database backends:
- FUSE components:
- Python FUSE 0.2.1
- FUSE library 2.8.4
- fusermount 2.8.4
- FUSE kernel interface 7.12
- Python 2.6.6, with the following additional libraries:
- attr extended attributes utilities 2.4.44
BaculaFS requires the following in order to function:
- access to the database that’s used to store the Bacula catalog
- access to the Bacula storage device
- access to the Bacula bextract utility (bundled with the Bacula storage daemon installation)
Extract the source code archive to a temporary directory, cd to this directory and run
python setup.py install
baculafs [mountpoint] [options] 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|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 (use '-o password= ' to get a password prompt; if not provided, the password is read from the DATABASE_PASSWORD environment variable) -o conf=PATH storage daemon configuration file [default: /etc/bacula/bacula-sd.conf] -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: False] -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: False] -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_difflist=DIFFLIST extract files that do not match files in DIFFLIST (list line format: 'Day Mon DD hh:mm:ss YYYY PATH'; use '-' to read from standard input; hint: format matches output of 'duplicity list-current-files -v0 target_url'; implies prefetch_symlinks) -o prefetch_list=LIST extract files that match files in LIST (list should contains one absolute file path per line; use '-' to read from standard input; implies prefetch_symlinks) -o prefetch_everything extract everything upon filesystem initialization (complete restore to cache) [default: False] -o batch_list list files to be prefetched and exit [default: False] -o batch_bsr dump contnets of bsr file for extracting prefetched files and exit [default: False] -o batch_extract extract prefetched files to mount point and exit [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: info] -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: [subdir] -o subdir=DIR prepend this directory to all paths (mandatory) -o [no]rellinks transform absolute symlinks to relative [iconv] -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.
- 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 may be used in batch mode with -o batch_extract in order to extract files from the Bacula storage device and then exit without mounting the filesystem. The mountpoint specified at the command line is then treated as the destination directory for extracted files.
The list of files, to be extracted in batch mode, is determined by the various cache prefetch options, and may be dumped with -o batch_list.
The bootstrap file that is generated, in order to extract the files, can also be dumped to standard output with -o batch_bsr.
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:
user.baculafs.FileIndex user.baculafs.JobId user.baculafs.LStat user.baculafs.MD5
Note that user.baculafs.MD5 shows whatever digest Bacula was configured to calculate for the file, be it MD5, SHA1, SHA256 or SHA512.
The root directory has several more attributes, that expose filesystem instance-specific information:
user.baculafs.cache_prefix user.baculafs.client user.baculafs.datetime user.baculafs.fileset user.baculafs.joblist
and several more attributes for monitoring the file extraction process:
user.baculafs.bextract.failures user.baculafs.bextract.path user.baculafs.bextract.pending user.baculafs.bextract.retries user.baculafs.bextract.state user.baculafs.bextract.volume
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.
Mount the most recent backup snapshot for Bacula client client-fd and fileset client-fileset; SQLite database backend (default):
baculafs -o client=client-fd,fileset=client-fileset /path/to/mount/point
Mount the contents of the specified list of backup jobs; MySQL database backend; prompt for password and prefetch to cache all symbolic links:
baculafs -o joblist='1001 1003',client=client-fd,fileset=client-fileset \ -o driver=mysql,password=,prefetch_symlinks /path/to/mount/point
Mount the contents of the most recent backup job only:
baculafs -o recent_job,client=client-fd,fileset=client-fileset /path/to/mount/point
Mount the contents of the fileset snapshot before the specified date/time:
baculafs -o datetime='2009-05-23 00:00:00' \ -o client=client-fd,fileset=client-fileset /path/to/mount/point
Allow other users to access filesystem, set logging level to debug and stay in foreground, so that bextract messages may be examined; assume a single Bacula fileset is defined for client client-fd:
baculafs -f -o allow_other,client=client-fd,logging=debug /path/to/mount/point
Prefetch to cache files that do not exist, or whose modification time or size differ from those at the specified directory (i.e. files that would be sent to that directory using rsync); and cleanup cache after un-mounting:
baculafs -o prefetch_diff=/path/to/rsync/target,cleanup \ -o client=client-fd /path/to/mount/point
Batch update a snapshot of the current files in backup (the filesystem is not mounted):
baculafs -o batch_extract,prefetch_diff=/path/to/snapshot,cleanup \ -o client=client-fd /path/to/snapshot/
(this is not as accurate as mounting the filesystem, like in the previous example, and then updating the snapshot with rsync - but it is faster).
BaculaFS will abort if no backup job was run for the client/fileset combination selected by the user.
BaculaFS will abort if the target mount point is in use. Note, however, that mount point availability is checked by FUSE only after the (potentially lengthy) initialization of BaculaFS.
Browsing a BaculaFS mount point with a file manager like Nautilus is liable to be very slow. This is because the file manager reads data from each file being listed, in order to determine its type, generate a thumbnail etc. This, in turn, triggers BaculaFS to extract the files, one at a time, from the Bacula storage device to its cache directory.
Access to the storage device by different instances of BaculaFS is serialized by locking the storage daemon configuration file. This means that you can mount several views of the backup catalog (e.g. accessing backup snapshots of different clients or snapshots from the same client but from different dates).
There are at least two issues with this locking mechanism that you should note:
- the lock is advisory, meaning that it does not prevent the Bacula storage daemon itself from accessing the storage device while in use by BaculaFS
- depending on your setup, the lock may not work if the storage daemon configuration file is accessed via NFS
BaculaFS depends on bextract and thus inherits its limitations:
- BaculaFS can be used with Windows filesets, but it does not reproduce any Windows specific file attributes
- Encrypted backup files are not supported
- BaculaFS must be started with enough permissions (typically as root) in order to allow bextract to extract files from the Bacula storage
Version 0.1.7 (2010-12-30)
- fixed: compatibility issues with bacula v2.4.4 and FUSE 7.8
- fixed: synthesize missing inode numbers with -o use_ino
- modified: decode value of user.baculafs.MD5 extended file attribute
- added: cache prefetch specified list of files
- added: changelog to README
- added: batch extract mode
Version 0.1.6 (2010-09-19)
- fixed cache prefetch by regex
Version 0.1.5 (2010-07-06)
- fixed: removed reference to obsolete db field Copy
- fixed: recent_job option with MySQL
- added: read database password from environment variable DATABASE_PASSWORD
- added: usage examples to README
Version 0.1.4 (2010-02-07)
- added: cache prefetch based on duplicity file listing
Version 0.1.3 (2010-01-13)
- fixed: (again) prefetch restore of files split between volumes
- fixed: missing import sys
- fixed: spurious linebreaks in debug log
Version 0.1.2 (2010-01-13)
- fixed: prefetch_recent when joblist contains a single non full job
- fixed: prefetch restore of files split between volumes
- added: copyright, trademark and license blurbs
Version 0.1.1 (2010-01-07)
- workaround: subtle extraction bug (fix forthcoming)
Version 0.1.1 (2010-01-06)
- initial public release
Please report problems via the BaculaFS issue tracking system: http://code.google.com/p/baculafs/issues/list
BaculaFS contains SQL queries that were adapted from Bacula, Copyright © 2000-2010 Free Software Foundation Europe e.V.
BaculaFS is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/.
|[*]||Bacula is a registered trademark of Kern Sibbald.|