Kern's ToDo List
- 16 July 2007
+ 02 January 2008
Document:
- How to sync remote offices.
- Exchange backup:
http://www.microsoft.com/technet/itshowcase/content/exchbkup.mspx
+- David's priorities
+ Copypools
+ Extract capability (#25)
+ Continued enhancement of bweb
+ Threshold triggered migration jobs (not currently in list, but will be
+ needed ASAP)
+ Client triggered backups
+ Complete rework of the scheduling system (not in list)
+ Performance and usage instrumentation (not in list)
+ See email of 21Aug2007 for details.
+- Implement Diff,Inc Retention Periods
+- Look at: http://tech.groups.yahoo.com/group/cfg2html
+ and http://www.openeyet.nl/scc/ for managing customer changes
Priority:
+- Implement Despooling data status.
+=== Duplicate jobs ===
+ hese apply only to backup jobs.
+
+ 1. Allow Duplicate Jobs = Yes | No | Higher (Yes)
+
+ 2. Duplicate Job Interval = <time-interval> (0)
+
+ The defaults are in parenthesis and would produce the same behavior as today.
+
+ If Allow Duplicate Jobs is set to No, then any job starting while a job of the
+ same name is running will be canceled.
+
+ If Allow Duplicate Jobs is set to Higher, then any job starting with the same
+ or lower level will be canceled, but any job with a Higher level will start.
+ The Levels are from High to Low: Full, Differential, Incremental
+
+ Finally, if you have Duplicate Job Interval set to a non-zero value, any job
+ of the same name which starts <time-interval> after a previous job of the
+ same name would run, any one that starts within <time-interval> would be
+ subject to the above rules. Another way of looking at it is that the Allow
+ Duplicate Jobs directive will only apply after <time-interval> of when the
+ previous job finished (i.e. it is the minimum interval between jobs).
+
+ So in summary:
+
+ Allow Duplicate Jobs = Yes | No | HigherLevel | CancelLowerLevel (Yes)
+
+ Where HigherLevel cancels any waiting job but not any running job.
+ Where CancelLowerLevel is same as HigherLevel but cancels any running job or
+ waiting job.
+
+ Duplicate Job Proximity = <time-interval> (0)
+
+ Skip = Do not allow two or more jobs with the same name to run
+ simultaneously within the proximity interval. The second and subsequent
+ jobs are skipped without further processing (other than to note the job
+ and exit immediately), and are not considered errors.
+
+ Fail = The second and subsequent jobs that attempt to run during the
+ proximity interval are cancelled and treated as error-terminated jobs.
+
+ Promote = If a job is running, and a second/subsequent job of higher
+ level attempts to start, the running job is promoted to the higher level
+ of processing using the resources already allocated, and the subsequent
+ job is treated as in Skip above.
+===
+- the cd-command should allow complete paths
+ i.e. cd /foo/bar/foo/bar
+ -> if a customer mails me the path to a certain file,
+ its faster to enter the specified directory
+- Fix bpipe.c so that it does not modify results pointer.
+ ***FIXME*** calling sequence should be changed.
+- Make tree walk routines like cd, ls, ... more user friendly
+ by handling spaces better.
+=== rate design
+ jcr->last_rate
+ jcr->last_runtime
+ MA = (last_MA * 3 + rate) / 4
+ rate = (bytes - last_bytes) / (runtime - last_runtime)
+- Add a recursive mark command (rmark) to restore.
+- "Minimum Job Interval = nnn" sets minimum interval between Jobs
+ of the same level and does not permit multiple simultaneous
+ running of that Job (i.e. lets any previous invocation finish
+ before doing Interval testing).
+- Look at simplifying File exclusions.
+- New directive "Delete purged Volumes"
+- It appears to me that you have run into some sort of race
+ condition where two threads want to use the same Volume and they
+ were both given access. Normally that is no problem. However,
+ one thread wanted the particular Volume in drive 0, but it was
+ loaded into drive 1 so it decided to unload it from drive 1 and
+ then loaded it into drive 0, while the second thread went on
+ thinking that the Volume could be used in drive 1 not realizing
+ that in between time, it was loaded in drive 0.
+ I'll look at the code to see if there is some way we can avoid
+ this kind of problem. Probably the best solution is to make the
+ first thread simply start using the Volume in drive 1 rather than
+ transferring it to drive 0.
+- Complete Catalog in Pool
+- Implement Bacula plugins -- design API
+- Scripts
+- Prune by Job
+- Prune by Job Level
+- True automatic pruning
+- Duplicate Jobs
+ Run, Fail, Skip, Higher, Promote, CancelLowerLevel
+ Proximity
+ New directive.
+- Auto update of slot:
+ rufus-dir: ua_run.c:456-10 JobId=10 NewJobId=10 using pool Full priority=10
+ 02-Nov 12:58 rufus-dir JobId 10: Start Backup JobId 10, Job=kernsave.2007-11-02_12.58.03
+ 02-Nov 12:58 rufus-dir JobId 10: Using Device "DDS-4"
+ 02-Nov 12:58 rufus-sd JobId 10: Invalid slot=0 defined in catalog for Volume "Vol001" on "DDS-4" (/dev/nst0). Manual load my be required.
+ 02-Nov 12:58 rufus-sd JobId 10: 3301 Issuing autochanger "loaded? drive 0" command.
+ 02-Nov 12:58 rufus-sd JobId 10: 3302 Autochanger "loaded? drive 0", result is Slot 2.
+ 02-Nov 12:58 rufus-sd JobId 10: Wrote label to prelabeled Volume "Vol001" on device "DDS-4" (/dev/nst0)
+ 02-Nov 12:58 rufus-sd JobId 10: Alert: TapeAlert[7]: Media Life: The tape has reached the end of its useful life.
+ 02-Nov 12:58 rufus-dir JobId 10: Bacula rufus-dir 2.3.6 (26Oct07): 02-Nov-2007 12:58:51
+- Eliminate: /var is a different filesystem. Will not descend from / into /var
+- Separate Files and Directories in catalog
+- Create FileVersions table
+- Look at rsysnc for incremental updates and dedupping
+- Add MD5 or SHA1 check in SD for data validation
+- modify pruning to keep a fixed number of versions of a file,
+ if requested.
+- finish implementation of fdcalled -- see ua_run.c:105
+- Fix problem in postgresql.c in my_postgresql_query, where the
+ generation of the error message doesn't differentiate result==NULL
+ and a bad status from that result. Not only that, the result is
+ cleared on a bail_out without having generated the error message.
+- KIWI
- Implement SDErrors (must return from SD)
- Implement USB keyboard support in rescue CD.
+- Implement continue spooling while despooling.
- Remove all install temp files in Win32 PLUGINSDIR.
- Audit retention periods to make sure everything is 64 bit.
- Use E'xxx' to escape PostgreSQL strings.
> configuration string value to a CRYPTO_CIPHER_* value, if anyone is
> interested in implementing this functionality.
-- Why doesn't @"xxx abc" work in a conf file?
- Figure out some way to "automatically" backup conf changes.
- Add the OS version back to the Win32 client info.
- Restarted jobs have a NULL in the from field.
For next release:
+- Try to fix bscan not working with multiple DVD volumes bug #912.
- Look at mondo/mindi
-- Don't restore Solaris Door files:
- #define S_IFDOOR in st_mode.
- see: http://docs.sun.com/app/docs/doc/816-5173/6mbb8ae23?a=view#indexterm-360
- Make Bacula by default not backup tmpfs, procfs, sysfs, ...
- Fix hardlinked immutable files when linking a second file, the
immutable flag must be removed prior to trying to link it.
http://linuxwiki.de/Bacula (in German)
- Possibly allow SD to spool even if a tape is not mounted.
-- It appears to me that you have run into some sort of race
- condition where two threads want to use the same Volume and they
- were both given access. Normally that is no problem. However,
- one thread wanted the particular Volume in drive 0, but it was
- loaded into drive 1 so it decided to unload it from drive 1 and
- then loaded it into drive 0, while the second thread went on
- thinking that the Volume could be used in drive 1 not realizing
- that in between time, it was loaded in drive 0.
- I'll look at the code to see if there is some way we can avoid
- this kind of problem. Probably the best solution is to make the
- first thread simply start using the Volume in drive 1 rather than
- transferring it to drive 0.
- Fix re-read of last block to check if job has actually written
a block, and check if block was written by a different job
(i.e. multiple simultaneous jobs writing).
- Look into replacing autotools with cmake
http://www.cmake.org/HTML/Index.html
-=== Migration from David ===
-What I'd like to see:
-
-Job {
- Name = "<poolname>-migrate"
- Type = Migrate
- Messages = Standard
- Pool = Default
- Migration Selection Type = LowestUtil | OldestVol | PoolOccupancy |
-Client | PoolResidence | Volume | JobName | SQLquery
- Migration Selection Pattern = "regexp"
- Next Pool = <override>
-}
-
-There should be no need for a Level (migration is always Full, since you
-don't calculate differential/incremental differences for migration),
-Storage should be determined by the volume types in the pool, and Client
-is really a selection issue. Migration should always occur to the
-NextPool defined in the pool definition. If no nextpool is defined, the
-job should end with a reason of "no place to go". If Next Pool statement
-is present, we override the check in the pool definition and use the
-pool specified.
-
-Here's how I'd define Migration Selection Types:
-
-With Regexes:
-Client -- Migrate data from selected client only. Migration Selection
-Pattern regexp provides pattern to select client names, eg ^FS00* makes
-all client names starting with FS00 eligible for migration.
-
-Jobname -- Migration all jobs matching name. Migration Selection Pattern
-regexp provides pattern to select jobnames existing in pool.
-
-Volume -- Migrate all data on specified volumes. Migration Selection
-Pattern regexp provides selection criteria for volumes to be migrated.
-Volumes must exist in pool to be eligible for migration.
-
-
-With Regex optional:
-LowestUtil -- Identify the volume in the pool with the least data on it
-and empty it. No Migration Selection Pattern required.
-
-OldestVol -- Identify the LRU volume with data written, and empty it. No
-Migration Selection Pattern required.
-
-PoolOccupancy -- if pool occupancy exceeds <highmig>, migrate volumes
-(starting with most full volumes) until pool occupancy drops below
-<lowmig>. Pool highmig and lowmig values are in pool definition, no
-Migration Selection Pattern required.
-
-
-No regex:
-SQLQuery -- Migrate all jobuids returned by the supplied SQL query.
-Migration Selection Pattern contains SQL query to execute; should return
-a list of 1 or more jobuids to migrate.
-
-PoolResidence -- Migrate data sitting in pool for longer than
-PoolResidence value in pool definition. Migration Selection Pattern
-optional; if specified, override value in pool definition (value in
-minutes).
-
-
-[ possibly a Python event -- kes ]
-===
- Mount on an Autochanger with no tape in the drive causes:
Automatically selected Storage: LTO-changer
Enter autochanger drive[0]: 0
("D","Diff"),
("I","Inc");
- Show files/second in client status output.
-- Add a recursive mark command (rmark) to restore.
-- "Minimum Job Interval = nnn" sets minimum interval between Jobs
- of the same level and does not permit multiple simultaneous
- running of that Job (i.e. lets any previous invocation finish
- before doing Interval testing).
-- Look at simplifying File exclusions.
-- New directive "Delete purged Volumes"
- new pool XXX with ScratchPoolId = MyScratchPool's PoolId and
let it fill itself, and RecyclePoolId = XXX's PoolId so I can
see if it become stable and I just have to supervise
MyScratchPool
- If I want to remove this pool, I set RecyclePoolId = MyScratchPool's
PoolId, and when it is empty remove it.
-- Figure out how to recycle Scratch volumes back to the Scratch Pool.
- Add Volume=SCRTCH
- Allow Check Labels to be used with Bacula labels.
- "Resuming" a failed backup (lost line for example) by using the
backups of the same client and if we again try to start a full backup of
client backup abc bacula won't complain. That should be fixed.
-- Fix bpipe.c so that it does not modify results pointer.
- ***FIXME*** calling sequence should be changed.
- For Windows disaster recovery see http://unattended.sf.net/
- regardless of the retention period, Bacula will not prune the
last Full, Diff, or Inc File data until a month after the
- In restore don't compare byte count on a raw device -- directory
entry does not contain bytes.
-=== rate design
- jcr->last_rate
- jcr->last_runtime
- MA = (last_MA * 3 + rate) / 4
- rate = (bytes - last_bytes) / (runtime - last_runtime)
+
+
- Max Vols limit in Pool off by one?
- Implement Files/Bytes,... stats for restore job.
- Implement Total Bytes Written, ... for restore job.
- It remains to be seen how the backup performance of the DIR's will be
affected when comparing the catalog for a large filesystem.
+ 1. Use the current Director in-memory tree code (very fast), but currently in
+ memory. It probably could be paged.
+
+ 2. Use some DB such as Berkeley DB or SQLite. SQLite is already compiled and
+ built for Win32, and it is something we could compile into the program.
+
+ 3. Implement our own custom DB code.
+
+ Note, by appropriate use of Directives in the Director, we can dynamically
+ decide if the work is done in the Director or in the FD, and we can even
+ allow the user to choose.
+
+=== most recent accurate file backup/restore ===
+ Here is a sketch (i.e. more details must be filled in later) that I recently
+ made of an algorithm for doing Accurate Backup.
+
+ 1. Dir informs FD that it is doing an Accurate backup and lookup done by
+ Director.
+
+ 2. FD passes through the file system doing a normal backup based on normal
+ conditions, recording the names of all files and their attributes, and
+ indicating which files were backed up. This is very similar to what Verify
+ does.
+
+ 3. The Director receives the two lists of files at the end of the FD backup.
+ One, files backed up, and one files not backed up. It then looks up all the
+ files not backed up (using Verify style code).
+
+ 4. The Dir sends the FD a list of:
+ a. Additional files to backup (based on user specified criteria, name, size
+ inode date, hash, ...).
+ b. Files to delete.
+
+ 5. Dir deletes list of file not backed up.
+
+ 6. FD backs up additional files generates a list of those backed up and sends
+ it to the Director, which adds it to the list of files backed up. The list
+ is now complete and current.
+
+ 7. The FD generates delete records for all the files that were deleted and
+ sends to the SD.
+
+ 8. The Dir deletes the previous CurrentBackup list, and then does a
+ transaction insert of the new list that it has.
+
+ 9. The rest works as before ...
+
+ That is it.
+
+ Two new tables needed.
+ 1. CurrentBackupId table that contains Client, JobName, FileSet, and a unique
+ BackupId. This is created during a Full save, and the BackupId can be set to
+ the JobId of the Full save. It will remain the same until another Full
+ backup is done. That is when new records are added during a Differential or
+ Incremental, they must use the same BackupId.
+
+ 2. CurrentBackup table that contains essentially a File record (less a number
+ of fields, but with a few extra fields) -- e.g. a flag that the File was
+ backed up by a Full save (this permits doing a Differential). The unique
+ BackupId allows us to look up the CurrentBackup for a particular Client,
+ Jobname, FileSet using that unique BackupId as the key, so this table must be
+ indexed by the BackupId.
+
+ Note any time a file is saved by the FD other than during a Full save, the
+ Full save flag is cleared. When doing a Differential backup, if a file has
+ the Full save flag set, it is skipped, otherwise it is backed up. For an
+ Incremental backup, we check to see if the file has changed since the last
+ time we backed it up.
+
+ Deleted files should have FileIndex == 0
+
====
From David:
How about introducing a Type = MgmtPolicy job type? That job type would
format string. Then I have the tape labeled automatically with weekday
name in the correct language.
==========
-- Yes, that is surely the case. I probably should turn those into Warning
- errors. In addition, you just made me think that it might not be bad to
- add an option to check the file size after backing up the file and
- report if it changes. This would be done as an option because it would
- add extra overhead.
-
- Kern, good idea. If you do do that, mention in the output: file
- shrunk, or file expanded, just to make it obvious to the user
- (without having to the refer to file size), just how the file size
- changed.
-
- Would this option be for all file, or just one file? Or a fileset?
- Make output from status use html table tags for nicely
presenting in a browser.
- Can one write tapes faster with 8192 byte block sizes?
-> maybe its more easy to maintain this, if the
descriptions of that commands are outsourced to
a ceratin-file
-- the cd-command should allow complete paths
- i.e. cd /foo/bar/foo/bar
- -> if a customer mails me the path to a certain file,
- its faster to enter the specified directory
- if the password is not configured in bconsole.conf
you should be asked for it.
-> sometimes you like to do restore on a customer-machine
to start a job or pass its DHCP obtained IP number.
- Implement a query tape prompt/replace feature for a console
- Copy console @ code to gnome2-console
-- Make tree walk routines like cd, ls, ... more user friendly
- by handling spaces better.
- Make sure that Bacula rechecks the tape after the 20 min wait.
- Set IO_NOWAIT on Bacula TCP/IP packets.
- Try doing a raw partition backup and restore by mounting a
- Have SD compute MD5 or SHA1 and compare to what FD computes.
- Make VolumeToCatalog calculate an MD5 or SHA1 from the
actual data on the Volume and compare it.
-- Implement Bacula plugins -- design API
- Make bcopy read through bad tape records.
- Program files (i.e. execute a program to read/write files).
Pass read date of last backup, size of file last time.
==============================
Longer term to do:
+- Implement wait on multiple objects
+ - Multiple max times
+ - pthread signal
+ - socket input ready
- Design at hierarchial storage for Bacula. Migration and Clone.
- Implement FSM (File System Modules).
- Audit M_ error codes to ensure they are correct and consistent.
- Enhance time/duration input to allow multiple qualifiers e.g. 3d2h
- Add ability to backup to two Storage devices (two SD sessions) at
the same time -- e.g. onsite, offsite.
-- Add the ability to consolidate old backup sets (basically do a restore
- to tape and appropriately update the catalog). Compress Volume sets.
- Might need to spool via file is only one drive is available.
- Compress or consolidate Volumes of old possibly deleted files. Perhaps
someway to do so with every volume that has less than x% valid
files.
does the right thing.
- FD-SD quick disconnect
- Building the in memory restore tree is slow.
+- Erabt if min_block_size > max_block_size
+- Add the ability to consolidate old backup sets (basically do a restore
+ to tape and appropriately update the catalog). Compress Volume sets.
+ Might need to spool via file is only one drive is available.
+- Why doesn't @"xxx abc" work in a conf file?
+- Don't restore Solaris Door files:
+ #define S_IFDOOR in st_mode.
+ see: http://docs.sun.com/app/docs/doc/816-5173/6mbb8ae23?a=view#indexterm-360
+- Figure out how to recycle Scratch volumes back to the Scratch Pool.