Reusing chunks / brackup server

[Ask Bjørn Hansen]

On Aug 12, 2009, at 18:25, William Wolf wrote:

>> 3) Is there a practical way to re-use chunks from different backup  
>> sources;
>> or is that how it works by default? (Think backing up a bunch of  
>> different
>> boxes; many files/chunks will be the same).  I was thinking of using
>> rsnapshot (like we do now) to a central host and then 'brackup'  
>> that box;
>> but that'd probably end up being too many files for brackup to  
>> efficiently
>> handle.   (Our two rsnapshot backup servers are storing ~1.5TB each  
>> with
>> about 22M inodes in use each).   I guess the question is: How do  
>> you use
>> brackup?
>
> Someone correct me if I'm wrong, but I believe some people on this
> list have talked about this before, but the main issue was keeping the
> inventorydb etc synchronized across all the machines.  The inventorydb
> is what brackup uses to "remember" what chunks are already on the
> server, so you need to do the backups at different times on each
> server, and synchronize the inventorydb's before each sync or
> something similar.  Doing all this, I believe it should work.

Yeah, I figured that'd be the issue.

Given infinite time (or not already having infinite projects to do)  
it'd be fun to make a "brackup server" that would be a brackup target  
(working as a proxy to the real target) and then be the central  
maintainer of the inventorydb.

It'd do sha1 checking of the stored (proxied) data to let you share  
the brackup target with less risk from a rogue user getting access to  
the configuration on one of the sources.


   - ask