[Borgbackup] what are the implications of this repair?

[Thorsten von Eicken]

On 07/17/2018 02:03 PM, Thomas Waldmann wrote:
> On 17.07.2018 04:06, Thorsten von Eicken wrote:
>> I've had filesystem corruption on a backup disk and I've run borg check
>> --repair with the result that I got some messages like the following one:
>>
>> big/MusicFLAC/_Thorsten/lesson2/02 Unknown - Track 02.mp3: New missing
>> file chunk detected (Byte 0-1269141). Replacing with all-zero chunk.
>>
>> I'm now wondering what this means for the next backup I'm making. I
>> still have that file at the source, so when I do a borg create will it
>> back up the file fresh or will it think it already has it and dedup it
>> including the zero block?
> If the missing chunk is still in the source data and you run a backup
> (and thus recreate the missing chunk), the new archive will be healthy.
>
> To also get the old archives (that are not healthy and have the all-zero
> replacement chunk) into a healthy state, you need to run borg check
> --repair again after that backup.
Thanks for the info, that's pretty cool, nicely designed!

In another repo I wasn't as lucky, the index files got corrupted, and 
instead of ~20 archives only one remained. The messages I got from check 
--repair are like:

Analyzing archive photos-2018-05-31T03:57-07:00 (7/17)
Archive metadata block is missing!

After the check-repair this is one of the repos that is no longer listed 
by "borg list". Do I gather correctly that there is no way to 
reconstruct those indexes? And at what point does all the data that is 
no longer referenced get garbage collected, I presume that happens the 
next time I run borg prune?

Thanks!!!

Thorsten
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/borgbackup/attachments/20180717/eb86383c/attachment-0001.html>