[Borgbackup] How to fix corrupted index or hints?

[Sebastian Felis]

On 11/29/22 11:12, Thomas Waldmann wrote:
> Your repo is still in a very weird state.
>
>> Further the "rsync -a --delete backup/ repo/" did not help because 
>> the cronjob for the rsync backup job with ~"rsync -a repo/ backup/" 
>> was not disabled and created another remote copy :-/
>
> I hope you still have a consistent, valid state of that repo somewhere.


So I read: My repo has a weird state which can not be repaired. Full 
stop. Mm.

I will do my homework and search for an offline backup which is not in a 
weird state...


>> But I assume that borgs internal structure is comparable to git tree 
>> which I do know.
> We have docs, if you read them, you do not to jump to conclusions / 
> assumptions.


If I have a look to 
https://borgbackup.readthedocs.io/en/stable/internals/data-structures.html#the-object-graph 
and 
https://borgbackup.readthedocs.io/en/stable/internals/data-structures.html#keys 
I do not get the match in my head between my repo objects are all clean 
and the error of corrupted segment reference count.


My understanding is:

- The data structure of borg is a cryptographic DAG. If a single bit of 
an object content would toggle, it would lead to verification error and 
the upper ids (file, dir, and archive) would change to be correct (again)
- Borg lists archives. So there is a manifest and a object graph
- The object graph is valid and all objects (archives, directories, 
files, chunks) can be referenced and do exist
- All objects are valid and their data integrity is verified

And my big question mark is: Why does the repair job fails while the 
verification of DAG and objects are fine? And why is the repo in a "read 
only" mode since creating archives fails?


Don't get me wrong or bad. I still love borg, appreciate the endless 
work and have trust in it. I am also very happy about your responses. 
Still there are some open questions on my side.


Best regards

Sebastian