[Borgbackup] ACLs not preserved

ralph strebbing blackbirdralph at gmail.com
Mon Feb 28 09:44:09 EST 2022


On Sun, Feb 27, 2022 at 4:05 PM Thorsten Schöning <tschoening at am-soft.de> wrote:
> You were talking about testing with a local BORG repo preserving ACLs
> in some former mail. How exactly did you test that? At filesrv1, so
> with the same version of BORG etc. like in PUSH/PULL, only with a
> local repo and one running BORG instance in that case? Or somewhere
> else only to see if BORG keeps ACLs at all?
Correct, I tested with a local repo on filesrv1 in another directory.

> The thing is that the BORG instance on filesrv1 needs to be able to
> read the ACLs. When that works with a local repo on filesrv1, there
> should be something really strange going on. From my understanding of
> how BORG works, the instance in the client is doing all the heavy
> stuff like knowing about metadata, ACLs etc. and only forwards
> encrypted data to the BORG instance "hosting" the remote repo using
> "borg serve".
>
> How should that be able to not store ACLs, when it only gets encrypted
> data of something? AFAIK metadata and file contents are even handled
> the same in the end, regarding de-duplication, compression and stuff.
> "borg serve" shouldn't be able to distinguish any of that.
Well that's where I'm hung up honestly. We currently only rsync the
data from the systems to our backup server, and that preserves our
ACLs just fine, but ZFS has some issues, so we're looking to move back
to straight ext4 and use borg to enable encryption among some of its
other features.

Ralph


More information about the Borgbackup mailing list