Some issues with rdiff-backup-delete

[Dieter Heußner]

In the context of rdiff-backup the option „--remove-older-than timeInterval“ 
removes old 
increment files thus freeing up space on backup media. A recommended 
timeInterval is 
„52W“ which stands for „52 weeks“. After having run the rdiff-backup with this 
option, it is 
not possible to restore a file back in time beyond the timeInterval.


To remove a file and all its associated history information from the 
rdiff-backup repository 
you need to run the command rdiff-backup-delete. Apart from the option 
--dry-run“ you 
have to specify the path which you want to be deleted. NEVER delete manually a 
file from 
the rdiff-backup repository.


My experiences and a question:


Running some tests with rdiff-backup-delete I was surprised that the deletion 
of a single 
file (including all its associated history information) takes quite a lot of 
time (several 
minutes).


There are some reasons for the „unusual“ amount of time:
1. The speed of your computer: A lot of *.gz files must be extracted, the 
history 
information for the file to be deleted must also be dropped, and new *.gz files 
be 
constructed to preserve the integrity of the repository
2. The total number of these *.gz files
3. The speed of your backup media


My first test with rdiff-backup-delete was with a single file.


In the next test I specified wildcards, e.g.
rdiff-backup-delete /long/directory/path/abc.*
It took roughly the same amount of time as above (single file). This is 
understandable 
because opening and closing *.gz files must be done nevertheless (and this 
contributes 
mainly to the time to proceed).


However I have a use case for which I did not find a proper solution yet. Maybe 
somebody 
can point me to a practical solution.


In the rdiff-backup repository I have a lot of files of which the basenames end 
on „.~xyz“ 
(dot tilde string "xyz"), but these files are spread across many, many, many 
different 
directories in the repository.

Is there a way to delete all the files with a *single* rdiff-backup-delete 
command. Please 
note that the different directories may have different nesting levels.


If there is not yet a practical solution, one had to find the names of all 
directories in which 
the given files reside, and run the deletion job for every directory 
separately, one after the 
other. If a single run takes /n/ minutes (roughly), and if there were /m/ 
different 
directories, the total amount of deleting the selected files would be /m/ times 
/n/ minutes.

The rdiff-backup-delete tool is indispensable because an administrator must 
always be 
able to safely delete files (and associated history information) from the 
rdiff-backup 
repository.

I'm not arguing against rdiff-backup-delete, but I'd appreciate a practical way 
for deleting 
files in *one single* rdiff-backup-delete run.

Could somebody point me to such a solution, please?
Thank you very much in advance.

Best regards
Dieter

--------------------------------------------------------------
PGP/GPG Key fingerprint:
BF12 CD6F EDC4 9FBA C933  316B 2C81 0BEF 4324 8513
--------------------------------------------------------------

signature.asc
Description: This is a digitally signed message part.