Re: rdiff-backup and rdiff-backup-delete

[Dieter Heußner]

Thanks a lot for your e-mail and sorry for my late response.

A partition on my customer's computers has 900+ GB data which are backed up by 
rdiff-
backup on a daily basis (increment). The average daily data volume to be backed 
up (due 
to changes) is about 300 MB.

The rdiff-backup "backup window" is set to 52 weeks.

Currently I run rdiff-backup-delete for "many" files which had been backed up 
in the first 
run because rdiff-backup had been launched without exclude option.
I run rdiff-backup-delete because I do not want to compromise the integrity of 
the backup.

To my surprise, for the above mentioned partition (and its backup) the 
subdirectory "rdiff-
backup-data" contains more than 153,000 files ending on "*.gz".

To delete a single file by rdiff-backup-delete, it takes 20 - 25 minutes to 
delete it because 
the program has to open a huge number pf *.gz files, drop information abot the 
file to 
delete, and create a new version of that "*.gz" file.

Questions / new feature requests::
1. How can the performance of rdiff-backup-delete significantly be improved?
    Any hints?
2. Are there plans for the next release of rdiff-backup-delete to improve its
    performance by rewriting parts of the source code?
3. Is it possible to run multiple copies of rdiff-backup-delete simultaneously
    without any problems (deadlocks, collision, etc.)?
4. Does rdiff-backup use a database (e.g. sqlite3, etc.) so it knows in which
    one of the"*.gz" files there is information about the file to be deleted?
5. How to exclude open MS Office files from being backed up?
    The names of these files usually start with "~$".
    Do I have to escape the "$" in the exclude pattern?
    Could you provide built-in exlude-option for automatically exclude such
    files?

I think rdiff-backup is a very good tool for backup, however there is also a 
big need for an 
optimized version of rdiff-backup-delete to get rid of unwanted files and 
directories. 
Please note, the administrator is responsible to decide which files / 
directories to drop 
from rdiff-backup.

Best regards
Dieter

> > I had a chat with my customer an hour ago: They want the most recent
> > version and its immediate predecessor ("the two most recent version
> 
> of > any file") to be always available at any time.
> 
> If these are Office files, I suggest you use Office's own option: see
> https://answers.microsoft.com/en-us/msoffice/forum/all/my-computer-keeps-mak
> ing-backups-of-my-microsoft/79411263-08e6-4f7c-b67d-c75933513550 , but check
> the option rather than unchecking it. Then you can make backups, if
> desired, using any system you want, not necessarily
> rdiff-backup.
> 
> Note that what Office considers an "immediate predecessor" might not
> be what the client wants. If you take a file made yesterday, change it
> and save it, then make another change and save it two minutes later,
> the immediate predecessor is the version that you made two minutes
> ago, not the version made yesterday.

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

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