TW
Label "hw-issue" (or kernel / fs issue) - cause of corruption Label "bug" because `borg check --repair` crashed.
I analysed a somehow similar issue there: https://github.com/borgbackup/borg/issues/7221#issuecomment-1363193614 That "unpack requires a buffer of 41 bytes" points to a "short or empty read" where borg expected to be able to...
Can you read the file `37` (like e.g. copy it to /tmp)? Does it start with the magic string "BORG_SEG"? That sounds like that magic is not at the beginning...
That looks like a photo from an iphone, not like a borg segment file. Make sure hw and kernel is ok (maybe use the old kernel, before upgrade), then run...
Sure you must unmount to run `fsck` but there's no alternative.
Due to deduplication, a segment file usually contains chunks referenced by multiple backup archives. So, if you lose a segment file, it might affect multiple archives. Also, not checking/fixing (fsck)...
The tmp files come from borg trying to repair that segment file. Usually that repair should succeed (but here, it does not - likely because of the corrupted filesystem) and...
This is trying to prepare a bigger change in how borg works: giving up with precise refcounting and removing chunk size information from the chunks index/cache (so borg could just...
Did you read the next sentence from the docs after the one you quoted? It should make clear how that is meant.
borg needs enough memory for its hashtables / indexes: - repo index (server side when running client/server) - chunks index (client side) - files cache (client side) There are some...