Restarting an interrupted large borg extract

[dokbua]

Say you need to restore a large'ish, as in terabytes, failed disk drive. It seems that borg extract is supposed to be faster than borg mount + rsync, but what if something interrupts the restore halfway through - like for example a faulty internet connection or something. What is the best way to restart the process?

The docs says borg extract should be done into an empty directory, so that would require a restart from scratch? I imagine restarting from where it got interrupted would cause it to extract and overwrite the already extracted files, which could take a long time, particularly over an internet connection.

How much slower would rsync from a borg mount be?

What is the best solution, keeping in mind that a restore in this case would be considered an emergency?

[ThomasWaldmann]

borg extract has only minimal support for restoring into a non-empty directory (like e.g. if a path it wants to extract already exists, it would delete the current file first to get it out of the way and then extract from the archive). do not expect anything more complex than that.

if you have a rather unstable connection so that an interruption is likely, you might be better off with borg mount and rsync, just be aware that borg mount (the FUSE fs) does not support all metadata as borg extract does, esp. no ACLs, no (bsd)flags, no xattrs(?).

an alternative might be doing multiple well-defined and distinct partial extracts (see borg extract docs), like e.g. extracting /home/user1 first, then /home/user2.

i can't say how much slower borg mount would be for you, I've never compared them and even if I had, there are a lots of parameters influencing performance.