adump-devel Mailing List for Adump Backup Suite

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 422-6466

I've put up a list of my goals for Adump 1.2.  I intend for 1.2 to be a
usable backup release.

The goals are at:  <http://www.davidb.org/adump/goals-1.2.html>

David Brown

****** Long message below.  Here is a summary.

- Adump-1.0 released, what to do for next release.
- What I think are the two important tasks.
- The rest of the tasks.
- Interest in help with this.

****** Adump-1.0 released, what to do for next release.

Now that adump-1.0 has been released, I need to figure out where to go for
the next release.  The TODO list has a lot of things, and I wanted to get
some feedback on what might be appropriate to work on first.

Right now, adump has a very simple interface.  You create a config file
(say in /etc/adump.config).  In this you describe your backup device, a log
file, the backup tapes to use (and a directory to store the database for
each), and the filesystems you wish to backup.

Adump itself takes 4 arguments, the name of the config file, The set to use
(which tape, since a set can only contain one tape at this time), which
system to backup, and what level.

Some scripts can actually automate this quite a bit, but obviously I want
it to be more automatic.

****** What I think are the two important tasks.

So here is my idea on what should be worked on next.

- Client/server model.  adumpd can be run at startup.  It does all of the
  work of backups.  The user uses a program adumpc to control what adumpd
  is going to do.  I can very easily make distributed code in Ada, but it
  has the same kind of problem that NFS and Corba have, it was designed to
  be useful and not designed to be secure.  It works fine when behind a
  firewall, but would have to be rethought if you want it sitting on an
  exposed port for network backups.  (See SSL stuff below).

  A simple starting point would be for adumpd to be run once at startup of
  the system.  Adumpc commands could issue requests to change tapes, and
  restart the daemon.  Another command would be to tell it to do this next
  bacukp.  This command could be a crontab entry.  Or, adumpd could do its
  own scheduling.

- To be fully useful, then adumpd needs to incorporate the logic as to what
  level to backup next to a given set.  The logic is very simple, all that
  is required is that I store state somewhere.

At this point, I think adump would be quite useful for automated backups.
Additional features from the following would improve it, but I'm not sure
what order to do them in.

****** The rest of the tasks.

- Support exclusion lists.  You could easily exclude the /var/cache dir,
  /tmp, browser caches, and such.  It isn't real hard to do the pattern
  matching, and I already have the hooks in place, but I have to think up
  what to do if the exclusion list is changed between incrementals.  In its
  current implementation, if you exclude files and do an incremental, those
  files will actually go away when you restore.

- Write a full media manager.  Commands should allow tapes to be labelled
  and such.  Right now, if adump doesn't recognize the tape, it overwrites
  it with the label it thinks it should have.  This is not nice.

- Support remote backups.  Both push and pull models are possible.

- Add security to backups.  One possibility is to enhance Glade to use SSL
  for all of its communication.  It still wants to open more than one port.
  The PCS interface Ada uses is quite simple, so it would still probably be
  easier to replace garlic with something else rather than try to roll my
  own RPC mechanism.  Also, I'm not sure what is going to happen with our
  crypto/export laws here in the US.

- Write a restore utility.  Once this is tested and stable, the backup
  format can be extended to add more information allowing other features to
  be implemented.

- Add archiving style support.  Keep a catalog of backed up data and allow
  quick recover of specific files.

- I know people will ask for a GUI, but I have no real desire to write one,
  or even have one.  Seems like a good job for Tcl/Tk calling adumpc with
  arguments.

****** Interest in help.

I'm looking for two kinds of help for Adump, one much more than the other.

First, I'm looking for people willing to run the program.  Those who are
willing to build something that might not build right away, and play with
backup software with a slightly crufty interface.  But, people who would
like to see good backup software.

Second, would be anyone interested in helping with the code.  This could be
as simple as, "here are patches to make it build on FooOS."  Anything more
should definately be coordinated with me though.

Thanks for any feedback,
Dave Brown

I am pleased to announce the release of Adump 1.0.

Please refer to my web page <http://www.davidb.org/adump/> for more
information.

What is adump.  Adump is intended to eventually be a full suite of backup
software.  At this point, it can do full and incremental backups, and do
simple media management.

This release is primarily intended for other interested developers to look
at, although I have been using it for about a year to backup my machine.

Best of all, Adump is entirely written in Ada95.

David Brown
ad...@da...

On Wed, Sep 27, 2000 at 08:27:35AM -0500, William L. Jones wrote:
> 
> Does it read the file system through a raw device are does it go
> threw the file system?

Adump uses the file system.  Unless the OS filesystem has special support
for raw device backups (Coda for example), then raw device backups cannot
be done reliably without bringing the system down.  This is not acceptable
for most systems.

With computers being faster now, a filesystem traversal doesn't take all
that long (especially compared to the backup).

It also will make adump significantly easier to port.

Dave Brown

Does it read the file system through a raw device are does it go
threw the file system?

Bill Jones

Since there is now another subscriber on this list, I though I would post
some status.  I've been a bit busy at work, so I haven't made it to the
first release quite as quickly as I had hoped.

Adump currently implements the following:

- A config file language, and basic config file options to describe the
  configuration of backups needed to be done.

- Multiple levels of backups (full, and incremental).

- Generation of cpio archive output.  ('cpio -H crc' style)

- The configuration file can name a log file where a time-stamped log of
  the backups performed is made.

- A 'real-time' progress meter status display.

- The start of a media manager.  Can handle appending files to a tape, and
  verifying the label on the tape.

Things I would like to finish before adump.1.0 is released:

- Write a "member header" before each archive which describes the contents
  of the dump, and for incremental backups, lists the files that were
  removed by this incremental backup.

- Reorganize the option parsing and top level structure of the program.

- Properly handle tapes that are unlabelled, or are not properly labelled.

- Implement exclusion lists on backups to prevent certain files from being
  backed up (browser cache, for example).

There are many things to do for future releases, but this looks like a
good point to make a release and let others try out the code.

David Brown

2000	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug	Sep (3)	Oct	Nov	Dec
2001	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug (1)	Sep (1)	Oct	Nov	Dec
2002	Jan	Feb (1)	Mar	Apr	May	Jun	Jul	Aug	Sep	Oct	Nov	Dec

adump-devel Mailing List for Adump Backup Suite

adump-devel — Discuss Adump development.