Going about beta-testing a program... advice?

Thu Sep 23 12:29:26 UTC 2004

On Thu, Sep 23, 2004 at 12:06:50AM -0400, Madison Kelly wrote:
>   Thanks for the feedback! Here is a feature list off the top of my 
> head (I will do a proper one a little closer to releasing it). It is 
> what I have written so far for the help system though I haven't properly 
> edited it yet: (for the record it does not yet support CD/DVD-R drives 
> though that is planned. How soon it will be implemented depends on 
> demand). If you are interested I can post a few screen shots. I don't 
> want to post those without a request first though for fear of sounding 
> like I am being too pushy.
> 
>  [-= Warning, marketting-type spiel follows :) =-]
> 
> It is a backup program:
> - It can backup any mounted filesystem.
> - Any partition with a UUID can be specified as a source or destination 
> (assigned by 'blkid', part of the 'e2fsprogs' package).
> - Files and directories can be selected or excluded from backup using a 
> familiar tree-based file browser.
> 
> It is an archival program:
> - Source and destination partitions can be labeled with a descriptive 
> label such as a catalogue number.
> - Any partition can be search even if the physical medium is offline.
> - Search results show the descriptive label of the media containing the 
> file or directory making locating the proper medium simple.
> - Detailed information is stored on every file and directory.
> - Files copied onto a destination manually or otherwise will be added to 
> the searchable catalogue on the next update.
> - File and directory information in the database is frequently and 
> automatically updated.
> 
> It is easy to use:
> - Many modern distributions like Fedora Core pre-install everything 
> 'TLE-BU' needs to run.
> - It installs on top of a Linux machine requiring only Perl, 
> PostgresSQL, and standard shell programs.
> - It is made available to users through any webserver that supports CGI 
> scripts such as Apache.
> - Once installed, no knowledge of Linux is required and it can be 
> manipulated from any web browser under any operating system.
> - There is extensive, easy to read, help documentation throughout the 
> program.
> 
> It is easy to administer:
> - All warnings and errors are presented in an easy to understand fashion.
> - It has a build-in log viewer with detailed and easy to read entries.
> 
> It is secure:
> - Security is handled by the system (ie: 'iptables') and the web 
> browser; known and tested to be extremely secure.
> - Simply stopping the web-server can block all access to the program 
> after it has been configured.
> 
> It is very flexible:
> - Runs on 2.4.x and 2.6.x linux kernels.
> - Multiple backup jobs can be created with individual run times.
> - Multiple sources and destinations can be selected for each backup job.
> - Backup jobs do not need to be updated when files, directories, sources 
> or destination are added or removed.
> - It intelligently searches available destinations before assigning each 
> source directory:
> - It updates previously backed up copies of a source directory when 
> found maximizing efficient use of disk space.
> - Uses 'rsync' to only copy the parts of a file that have changed when 
> updating previously saved files minimizing disk activity.

I suspect you may have misunderstood how rsync works.  Rsync only
transfers changes when doing work over a network link, by reading the
whole file on each end (unless other things like time stamps and size
and such match perfectly in which case it assumes they are the same),
and calculating checksums to figure out what to transfer over the slow
network link.  When rsync is run locally to transfer between disks, it
simply checks the time stamps and decides if the file has changed or
not.  It would not make sense for rsync to do it's efficient over
network algorith since it involves reading both files entirely,
calculating checksums, then writing a new copy of the file on the
destination, reading from the original destination and filling in with
updates from the other as it goes.  When done the new file is renamed
over top of the original destination file.  Doing this would not save
disk activity, on the contrary it would require reading two files and
writing one file completely instead of a simple copy which is one read
and one write.  This is why rsync run locally simply does a copy if it
sees that a file is changed.

> - Can backup multiple sources to multiple destinations simultaneously 
> regardless of overlap.
> - Souce and destination partitions will be used in an intelligent order, 
> regardless of what order or how they are connected.
> - Source and destination partitions to not need to be mounted before use.
> - Any filesystem Linux knows how to read or write from can be used as a 
> source and any that it can write to can be used as a destination.

Does it backup all relevant attributes of the files?  Extended
atributers (filesystem specific)?  ACL?

> Data backed up by 'TLE-BU' is easily recoverable:
> - Files are directly written to the destination partition. No special 
> software in needed to read the backed-up data.
> - Data can be recovered simply by mounting the storage medium under any 
> operating system that can read it's filesystem.
> - 'TLE-BU' is NOT required to copy or recover the files!
> - 'TLE-BU' uses standard hard drives connected via USB, Firewire, IDE, 
> SerialATA, SCSI and/or Fiber-Channel.
> 
> Flexible capacity and performance:
> - It can run on anything from a laptop with a single USB drive through 
> to very large servers with dozens of drives.
> - Destination capacity is limited only by the number and size of the 
> drives, arrays and SANs your system can handle.

Well it sounds decent.

I use tob (tape oriented backup) with afio archives, splitting into 1GB
pieces that I then burn to DVD (4 pieces per DVD).  I archive acl info
with getfact -R and dump to a file and burn that seperately.

Restores simply require a linux system installed with fielsystem, and
afio available.

Lennart Sorensen
--
The Toronto Linux Users Group.      Meetings: http://tlug.ss.org
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://tlug.ss.org/subscribe.shtml