4 Documentation to do: (a little bit at a time)
5 - Document running a test version.
6 - Document query file format.
7 - Document static linking
8 - Document fixing tape file after crash (update Media set VolFiles=xx
9 where MediaId=yy; update status to append).
12 Testing to do: (painful)
13 - that console command line options work
14 - blocksize recognition code.
16 - test and fix < code and | code.
19 - Get correct error status from run_program or open_bpipe().
20 - Figure out how to allow multiple simultaneous file Volumes on
22 - Why are save/restore of device different sizes (sparse?)
25 - Implement some why for the Console to dynamically create
27 - Cancel does not work for restore in FD.
28 - Restore to a particular time -- e.g. before date, after date.
30 - Implement disk spooling
31 - Implement finer multiprocessing options.
32 - Solaris -I on tar for include list
33 - Add whereabslinks to where or not where absolute links
34 - Figure out some way to automatically backup all local partitions
35 - Enable avoid backing up archive device (findlib/find_one.c:128)
36 - Implement FileOptions (see end of this document)
37 - Implement Bacula plugins -- design API
38 - Make hash table for linked files in findlib/find_one.c:161
39 - Make bcopy read through bad tape records.
40 - Need a verbose mode in restore, perhaps to bsr.
41 - bscan without -v is too quiet -- perhaps show jobs.
42 - Add code to reject whole blocks if not wanted on restore.
43 - Implement multiple simultaneous file Volumes on a single device.
44 - Start working on Base jobs.
45 - Make sure the MaxVolFiles is fully implemented in SD
46 - Flush all the daemon messages at the end of every job.
47 - Check if both CatalogFiles and UseCatalog are set to SD.
48 - Check if we can increase Bacula FD priorty in Win2000
49 - Need return status on read_cb() from read_records(). Need multiple
50 records -- one per Job, maybe a JCR or some other structure with
52 - Figure out how to do a bare metal Windows restore
53 - Fix read_record to handle multiple sessions.
54 - Program files (i.e. execute a program to read/write files).
55 Pass read date of last backup, size of file last time.
56 - Put system type returned by FD into catalog.
57 - Possibly add email to Watchdog if drive is unmounted too
58 long and a job is waiting on the drive.
59 - Strip trailing slashes from Include directory names in the FD.
60 - Use read_record.c in SD code.
61 - Why don't we get an error message from Win32 FD when bootstrap
62 file cannot be created for restore command?
63 - Need to specify MaximumConcurrentJobs in the Job resource.
64 - When Marking a file in Restore that is a hard link, also
65 mark the link so that the data will be reloaded.
66 - Restore program that errors in SD due to no tape reports
67 OK incorrectly in output.
68 - After unmount, if restore job started, ask to mount.
69 - Make Restore report an error if FD or SD term codes are not OK.
70 - Convert all %x substitution variables, which are hard to remember
71 and read to %(variable-name). Idea from TMDA.
72 - Add JobLevel in FD status (but make sure it is defined).
73 - Make Pool resource handle Counter resources.
74 - Remove NextId for SQLite. Optimize.
75 - Strip trailing / from Include
76 - Move all SQL statements into a single location.
77 - Cleanup db_update_media and db_update_pool
78 - Add UA rc and history files.
79 - put termcap (used by console) in ./configure and
80 allow -with-termcap-dir.
81 - Enhance time and size scanning routines.
82 - Fix Autoprune for Volumes to respect need for full save.
83 - Fix Win32 config file definition name on /install
84 - No READLINE_SRC if found in alternate directory.
85 - Add Client FS/OS id (Linux, Win95/98, ...).
86 - Test a second language e.g. french.
87 - Compare tape to Client files (attributes, or attributes and data)
88 - Make all database Ids 64 bit.
89 - Write an applet for Linux.
90 - Add estimate to Console commands
91 - Find solution to blank filename (i.e. path only) problem.
92 - Implement new daemon communications protocol.
93 - Remove PoolId from Job table, it exists in Media.
94 - Allow console commands to detach or run in background.
95 - Fix status delay on storage daemon during rewind.
96 - Add SD message variables to control operator wait time
97 - Maximum Operator Wait
98 - Minimum Message Interval
99 - Maximum Message Interval
100 - Send Operator message when cannot read tape label.
101 - Verify level=Volume (scan only), level=Data (compare of data to file).
102 Verify level=Catalog, level=InitCatalog
104 - Add keyword search to show command in Console.
105 - Fix Win2000 error with no messages during startup.
106 - Events : tape has more than xxx bytes.
107 - Restrict characters permitted in a Resource name.
108 - Complete code in Bacula Resources -- this will permit
109 reading a new config file at any time.
110 - Handle ctl-c in Console
111 - Implement LabelTemplate (at least first cut).
112 - Implement script driven addition of File daemon to config files.
113 - Think about how to make Bacula work better with File (non-tape) archives.
115 - Implement new serialize subroutines
116 send(socket, "string", &Vol, "uint32", &i, NULL)
117 - Audit all UA commands to ensure that we always prompt where possible.
118 - If ./btape is called without /dev, assume argument is a Storage resource name.
119 - Put memory utilization in Status output of each daemon
120 if full status requested or if some level of debug on.
121 - Make database type selectable by .conf files i.e. at runtime
122 - gethostbyname failure in bnet_connect() continues
123 generating errors -- should stop.
124 - Set flag for uname -a. Add to Volume label.
125 - Implement throttled work queue.
126 - Check for EOT at ENOSPC or EIO or ENXIO (unix Pc)
127 - Allow multiple Storage specifications (or multiple names on
128 a single Storage specification) in the Job record. Thus a job
129 can be backed up to a number of storage devices.
130 - Implement dump label to UA
131 - Concept of VolumeSet during restore which is a list
132 of Volume names needed.
133 - Restore files modified after date
134 - Restore file modified before date
135 - Emergency restore info:
137 - Backup working directory
139 - Restore -- do nothing but show what would happen
140 - SET LD_RUN_PATH=$HOME/mysql/lib/mysql
141 - Implement Restore FileSet=
142 - Create a protocol.h and protocol.c where all protocol messages
144 - If SD cannot open a drive, make it periodically retry.
145 - Remove duplicate fields from jcr (e.g. jcr.level and jcr.jr.Level, ...).
146 - Timout a job or terminate if link goes down, or reopen link and query.
147 - Find general solution for sscanf size problems (as well
148 as sprintf. Do at run time?
149 - Concept of precious tapes (cannot be reused).
150 - Make bcopy copy with a single tape drive.
151 - Permit changing ownership during restore.
153 - Autolabel should be specified by DIR instead of SD.
154 - Find out how to get the system tape block limits, e.g.:
155 Apr 22 21:22:10 polymatou kernel: st1: Block limits 1 - 245760 bytes.
156 Apr 22 21:22:10 polymatou kernel: st0: Block limits 2 - 16777214 bytes.
159 - AutoScan (check checksum of tape)
160 - Format command = "format /dev/nst0"
164 - Seek resolution (usually corresponds to buffer size)
165 - EODErrorCode=ENOSPC or code
166 - Partial Read error code
167 - Partial write error code
168 - Nonformatted read error
169 - Nonformatted write error
170 - WriteProtected error
174 - IgnoreCloseErrors=yes
184 - FD sends unsaved file list to Director at end of job (see
186 - File daemon should build list of files skipped, and then
187 at end of save retry and report any errors.
188 - Write a Storage daemon that uses pipes and
189 standard Unix programs to write to the tape.
191 - Need something that monitors the JCR queue and
192 times out jobs by asking the deamons where they are.
193 - Enhance Jmsg code to permit buffering and saving to disk.
194 - device driver = "xxxx" for drives.
195 - restart: paranoid: read label fsf to
196 eom read append block, and go
197 super-paranoid: read label, read all files
198 in between, read append block, and go
199 verify: backspace, read append block, and go
200 permissive: same as above but frees drive
201 if tape is not valid.
203 - Ensure that /dev/null works
204 - Need report class for messages. Perhaps
205 report resource where report=group of messages
206 - enhance scan_attrib and rename scan_jobtype, and
207 fill in code for "since" option
208 - Director needs a time after which the report status is sent
209 anyway -- or better yet, a retry time for the job.
210 Don't reschedule a job if previous incarnation is still running.
211 - Some way to automatically backup everything is needed????
212 - Need a structure for pending actions:
214 - termination status (part of buffered msgs?)
215 - Concept of grouping Storage devices and job can use
216 any of a number of devices
218 Read, Write, Clean, Delete
219 - Login to Bacula; Bacula users with different permissions:
220 owner, group, user, quotas
221 - Store info on each file system type (probably in the job header on tape.
222 This could be the output of df; or perhaps some sort of /etc/mtab record.
225 - Design at hierarchial storage for Bacula.
226 - Implement FSM (File System Modules).
227 - Identify unchanged or "system" files and save them to a
228 special tape thus removing them from the standard
229 backup FileSet -- BASE backup.
230 - Turn virutally all sprintfs into snprintfs.
231 - Heartbeat between daemons.
232 - Audit M_ error codes to ensure they are correct and consistent.
233 - Add variable break characters to lex analyzer.
234 Either a bit mask or a string of chars so that
235 the caller can change the break characters.
236 - Make a single T_BREAK to replace T_COMMA, etc.
237 - Ensure that File daemon and Storage daemon can
238 continue a save if the Director goes down (this
239 is NOT currently the case). Must detect socket error,
240 buffer messages for later.
241 - Enhance time/duration input to allow multiple qualifiers e.g. 3d2h
245 Bacula Projects Roadmap
247 last update 5 January 2003
249 Item 1: Multiple simultaneous Jobs. (done)
250 Done -- Restore part needs better implementation to work correctly
252 What: Permit multiple simultaneous jobs in Bacula.
254 Why: An enterprise level solution needs to go fast without the
255 need for the system administrator to carefully tweak
256 timing. Based on the benchmarks, during a full
257 backup, NetWorker typically hit 10 times the bandwidth to
258 the tape compared to Bacula--largely. This is probably due to
259 running parallel jobs and multi-threaded filling of buffers
260 and writing them to tape. This should also make things work
261 better when you have a mix of fast and slow machines backing
264 Notes: Bacula was designed to run multiple simultaneous jobs. Thus
265 implementing this is a matter of some small cleanups and
269 Item 2: Make the Storage daemon use intermediate file storage to buffer data.
270 Deferred -- not necessary yet.
272 What: If data is coming into the SD too fast, buffer it to
273 disk if the user has configured this option.
275 Why: This would be nice, especially if it more or less falls out
276 when implementing (1) above. If not, it probably should not
277 be given a high priority because fundamentally the backup time
278 is limited by the tape bandwidth. Even though you may finish a
279 client job quicker by spilling to disk, you still have to
280 eventually get it onto tape. If intermediate disk buffering
281 allows us to improve write bandwidth to tape, it may make
284 Notes: Whether or not this is implemented will depend upon performance
285 testing after item 1 is implemented.
288 Item 3: Write the bscan program -- also write a bcopy program.
291 What: Write a program that reads a Bacula tape and puts all the
292 appropriate data into the catalog. This allows recovery
293 from a tape that is no longer in the database, or it allows
294 re-creation of a database if lost.
296 Why: This is a fundamental robustness and disaster recovery tool
297 which will increase the comfort level of a sysadmin
298 considering adopting Bacula.
300 Notes: A skeleton of this program already exists, but much work
301 needs to be done. Implementing this will also make apparent
302 any deficiencies in the current Bacula tape format.
305 Item 4: Implement Base jobs.
307 What: A base job is sort of like a Full save except that you
308 will want the FileSet to contain only files that are unlikely
309 to change in the future (i.e. a snapshot of most of your
310 system after installing it). After the base job has been run,
311 when you are doing a Full save, you can specify to exclude
312 all files saved by the base job that have not been modified.
314 Why: This is something none of the competition does, as far as we know
315 (except BackupPC, which is a Perl program that saves to disk
316 only). It is big win for the user, it makes Bacula stand out
317 as offering a unique optimization that immediately saves time
320 Notes: Big savings in tape usage. Will require more resources because
321 the e. DIR must send FD a list of files/attribs, and the FD must
322 search the list and compare it for each file to be saved.
325 Item 5: Implement Label templates
327 What: This is a mechanism whereby Bacula can automatically create
328 a tape label for new tapes according to a detailed specification
329 provided by the user.
331 Why: It is a major convenience item for folks who use automated label
334 Notes: Bacula already has a working form of automatic tape label
335 creation, but it is very crude. The design for the complete
336 tape labeling project is already documented in the manual.
339 Item 6: Write a regression script.
342 What: This is an automatic script that runs and tests as many features
343 of Bacula as possible. The output is compared to previous
344 versions of Bacula and any differences are reported.
346 Why: This is an enormous help in preventing introduction of new
347 errors in parts of the program that already work correctly.
349 Notes: This probably should be ranked higher, it's something the typical
350 user doesn't see. Depending on how it's implemented, it may
351 make sense to defer it until the archival tape format and
352 user interface mature.
355 Item 7: GUI for interactive restore
356 Item 8: GUI for interactive backup
358 What: The current interactive restore is implemented with a tty
359 interface. It would be much nicer to be able to "see" the
360 list of files backed up in typical GUI tree format.
361 The same mechanism could also be used for creating
362 ad-hoc backup FileSets (item 8).
364 Why: Ease of use -- especially for the end user.
366 Notes: Rather than implementing in Gtk, we probably should go directly
367 for a Browser implementation, even if doing so meant the
368 capability wouldn't be available until much later. Not only
369 is there the question of Windows sites, most
370 Solaris/HP/IRIX, etc, shops can't currently run Gtk programs
371 without installing lots of stuff admins are very wary about.
372 Real sysadmins will always use the command line anyway, and
373 the user who's doing an interactive restore or backup of his
374 own files will in most cases be on a Windows machine running
378 Item 9: Add SSL to daemon communications.
380 What: This provides for secure communications between the daemons.
382 Why: This would allow doing backup across the Internet without
383 privacy concerns (or with much less concern).
385 Notes: The vast majority of near term potential users will be backing up
386 a single site over a LAN and, correctly or not, they probably
387 won't be concerned with security, at least not enough to go to
388 the trouble to set up keys, etc. to screw things down. We suspect
389 that many users genuinely interested in multi-site backup
390 already run some form of VPN software in their internetwork
391 connections, and are willing to delegate security to that layer.
394 Item 10: Define definitive tape format.
397 What: Define that definitive tape format that will not change
398 for the next millennium.
400 Why: Stability, security.
402 Notes: See notes for item 11 below.
405 Item 11: New daemon communication protocol.
407 What: The current daemon to daemon protocol is basically an ASCII
408 printf() and sending the buffer. On the receiving end, the
409 buffer is sscanf()ed to unpack it. The new scheme would
410 be a binary format that allows quick packing and unpacking
411 of any data type with named fields.
413 Why: Using binary packing would be faster. Named fields will permit
414 error checking to ensure that what is sent is what the
415 receiver really wants.
417 Notes: These are internal improvements in the interest of the
418 long-term stability and evolution of the program. On the one
419 hand, the sooner they're done, the less code we have to rip
420 up when the time comes to install them. On the other hand, they
421 don't bring an immediately perceptible benefit to potential
422 users. Item 10 and possibly item 11 should be deferred until Bacula
423 is well established with a growing user community more or
424 less happy with the feature set. At that time, it will make a
425 good "next generation" upgrade in the interest of data
431 ======================================================
433 It is somewhat like a Full save becomes an incremental since
434 the Base job (or jobs) plus other non-base files.
436 - New BaseFile table that contains:
437 JobId, BaseJobId, FileId (from Base).
438 i.e. for each base file that exists but is not saved because
439 it has not changed, the File daemon sends the JobId, BaseId,
440 and FileId back to the Director who creates the DB entry.
441 - To initiate a Base save, the Director sends the FD
442 the FileId, and full filename for each file in the Base.
443 - When the FD finds a Base file, he requests the Director to
444 send him the full File entry (stat packet plus MD5), or
445 conversely, the FD sends it to the Director and the Director
446 says yes or no. This can be quite rapid if the FileId is kept
447 by the FD for each Base Filename.
448 - It is probably better to have the comparison done by the FD
449 despite the fact that the File entry must be sent across the
451 - An alternative would be to send the FD the whole File entry
452 from the start. The disadvantage is that it requires a lot of
453 space. The advantage is that it requires less communications
455 - The Job record must be updated to indicate that one or more
457 - At end of Job, FD returns:
458 1. Count of base files/bytes not written to tape (i.e. matches)
459 2. Count of base file that were saved i.e. had changed.
460 - No tape record would be written for a Base file that matches, in the
461 same way that no tape record is written for Incremental jobs where
462 the file is not saved because it is unchanged.
463 - On a restore, all the Base file records must explicitly be
464 found from the BaseFile tape. I.e. for each Full save that is marked
465 to have one or more Base Jobs, search the BaseFile for all occurrences
467 - An optimization might be to make the BaseFile have:
473 This would avoid the need to explicitly fetch each File record for
474 the Base job. The Base Job record will be fetched to get the
475 VolSessionId and VolSessionTime.
476 =========================================================
478 ==========================================================
480 For each Incremental job that is run, there may be files that
481 were found but not saved because they were locked (this applies
482 only to Windows). Such a system could send back to the Director
483 a list of Unsaved files.
485 - New UnSavedFiles table that contains:
489 - Then in the next Incremental job, the list of Unsaved Files will be
490 feed to the FD, who will ensure that they are explicitly chosen even
491 if standard date/time check would not have selected them.
492 =============================================================
496 =============================================================
498 Request For Comments For File Backup Options
501 Subject: File Backup Options
504 A few days ago, a Bacula user who is backing up to file volumes and
505 using compression asked if it was possible to suppress compressing
506 all .gz files since it was a waste of CPU time. Although Bacula
507 currently permits using different options (compression, ...) on
508 a directory by directory basis, it cannot do it on a file by
509 file basis, which is clearly what was desired.
511 Proposed Implementation:
512 To solve this problem, I propose the following:
514 - Add a new Director resource type called FileOptions.
516 - The FileOptions resource will have records for all
517 options that can currently be specified on the Include record
518 (in a FileSet). Examples below.
520 - The FileOptions resource will permit an exclude option as well
521 as a number of additional options.
523 - The heart of the FileOptions resource is the ability to
524 supply any number of ApplyTo records which specify POSIX
525 regular expressions. These ApplyTo regular expressions are
526 applied to the fully qualified filename (path and all). If
527 one matches, then the FileOptions will be used.
529 - When an ApplyTo specification matches an included file, the
530 options specified in the FileOptions resource will override
531 the default options specified on the Include record.
533 - Include records will be modified to permit referencing one or
534 more FileOptions resources. The FileOptions will be used
535 in the order listed on the Include record and the first
536 one that matches will be applied.
538 - Options (or specifications) currently supplied on the Include
539 record will be deprecated (i.e. removed in a later version a
540 year or so from now).
542 - The Exclude record will be deprecated as the same functionality
543 can be obtained by using an Exclude = yes in the FileOptions.
546 The following records can appear in the FileOptions resource. An
547 asterisk preceding the name indicates a feature not currently
551 - Compression= (GZIP, ...)
552 - Signature= (MD5, SHA1, ...)
554 - OneFs= (yes/no) - remain on one filesystem
555 - Recurse= (yes/no) - recurse into subdirectories
556 - Sparse= (yes/no) - do sparse file backup
557 - *Exclude= (yes/no) - exclude file from being saved
558 - *Reader= (filename) - external read (backup) program
559 - *Plugin= (filename) - read/write plugin module
562 - verify= (ipnougsamc5) - verify options
565 - replace= (always/ifnewer/ifolder/never) - replace options currently
567 - *Writer= (filename) - external write (restore) program
571 Currently options specifying compression, MD5 signatures, recursion,
572 ... of a FileSet are supplied on the Include record. These will now
573 all be collected into a FileOptions resource, which will be
574 specified on the Include in place of the options. Multiple FileOptions
575 may be specified. Since the FileOptions contain regular expressions
576 that are applied to the full filename, this will give the ability
577 to specify backup options on a file by file basis to whatever level
586 Include = compression=GZIP signature=MD5 {
595 Include = FileOptions=Opts {
606 That's a lot more to do the same thing, but it gives the ability to
607 apply options on a file by file basis. For example, suppose you
608 want to compress all files but not any file with extensions .gz or .Z.
609 You could do so as follows:
613 Include = FileOptions=NoCompress FileOptions=Opts {
621 ApplyTo = /*.?*/ # matches all files
626 # Note multiple ApplyTos are ORed
627 ApplyTo = /*.gz/ # matches .gz files */
628 ApplyTo = /*.Z/ # matches .Z files */
631 Now, since the NoCompress FileOptions is specified first on the
632 Include line, any *.gz or *.Z file will have an MD5 signature computed,
633 but will not be compressed. For all other files, the NoCompress will not
634 match, so the Opts options will be used which will include GZIP
638 - Is it necessary to provide some means of ANDing regular expressions
639 and negation? (not currently planned)
641 e.g. ApplyTo = /*.gz/ && !/big.gz/
643 - I see that Networker has a "null" module which, if specified, does not
644 backup the file, but does make an record of the file in the catalog
645 so that the catalog will reflect an exact picture of the filesystem.
646 The result is that the file can be "seen" when "browsing" the save
647 sets, but it cannot be restored.
649 Is this really useful? Should it be implemented in Bacula?
652 After implementing the above, the user will be able to specify
653 on a file by file basis (using regular expressions) what options are
654 applied for the backup.
655 ====================================
657 Done: (see kernsdone for more)
658 - Add EOM records? No, not at this time. The current system works and
660 - Add VolumeUseDuration and MaximumVolumeJobs to Pool db record and
662 - Add VOLUME_CAT_INFO to the EOS tape record (as
663 well as to the EOD record). -- No, not at this time.
664 - Put MaximumVolumeSize in Director (MaximumVolumeJobs, MaximumVolumeFiles,
666 - Enhance schedule to have 1stSat, ...
667 - Make sure catalog doesn't keep growing.
668 - On I/O error, write EOF, then try to write again ? No, keep it simple.
669 - Figure out how compress everything except .gz,... files.
670 Implement FileOptions.
671 - Put Bacula version somewhere in Job stream, probably Start Session Labels.
672 - Fix start/end blocks for File devices
673 - Make Job err if WriteBootstrap fails.
674 - Test that mod of restore options works.
675 - Test that week position schedule code works.
676 - Make BSR accept count (total files to be restored).
677 - Add code to fast seek to proper place on tape/file when doing Restore.
678 - Replace popen() and pclose() -- fail safe and timeout, no SIG dep.
679 - Add code to put VolFile in bsr for restore command.
680 - Volumes can be listed multiple times in Restore volume list.
681 - Add watchdog timeout for child processes start_child_timer()
683 - Get rid of bscan.c:534 error message (one time only).
684 - Print some statistics when get EOF on device in bscan -- feedback
685 to let user know it is working.
686 - DateWritten field on tape may be wrong.
687 - Ensure that restore of differential jobs works (check SQL).
688 - Count number of ignored messages in bscan and print when first SOS is found.
689 - Test that EndFile/Block are correctly updated at end of tape
690 (in view of new block reading code).
691 - Test watchdog child timer code.
692 - Test new BSR code (mostly done).
693 - Work more on how to to a Bacula restore beginning with
694 just a Bacula tape and a boot floppy (bare metal recovery).
695 - Restore options (overwrite, overwrite if older,
696 overwrite if newer, never overwrite, ...)
697 - Fill all fields in Vol/Job Header -- ensure that everything
698 needed is written to tape. Think about restore to Catalog
699 from tape. Client record needs improving.
700 - Implement ./configure --with-client-only
701 - Finish up static linking
702 - that restore options work in FD
703 - Backup of raw partitions
704 - Document nofollow flag
705 - Make sure restore options are documented
706 - Should we dump a SOS when starting a new tape? (no -- too complicated)
707 - Fix gethostbyname() to use gethostbyname_r() -- no added mutex
708 - Add HOST to Volume label.
709 - Need to save contents of FileSet to tape? -- no not now
710 - Fix db_get_fileset in cats/sql_get.c for multiple records.
711 - Fix catalog filename truncation in sql_get and sql_create. Use
712 only a single filename split routine.
713 - Figure out how to save the catalog (possibly a special FileSet).
714 - Figure out how to restore the catalog.
715 - Make sure differential handled correctly
716 - Look at ua_prune.c in detail. Why did JobType work at all??????