borg operations always use an exclusive lock on the repository.
this goes back to a change when a deadlock issue related to lock-upgrade (from shared to exclusive lock) was resolved by just always doing exclusive locks.
always-exclusive locks can be a problem if one has long-running (or frequent) backup jobs (or in general: repo-writing operations) and in parallel one wants to do other borg operations:
- "read-only" operations, like
borg mount / list / info / extract
- write operations (this is definitely the more critical / complicated case and could be omitted to simplify)
topics to research in this context could be:
- analyze / document repo locking requirements on a rather detailed level
- analyze / document other (cache? clientside indexes?) locking requirements on a rather detailed level
- can we have a deadlock-free locking without using always-exclusive locking?
borg operations always use an exclusive lock on the repository.
this goes back to a change when a deadlock issue related to lock-upgrade (from shared to exclusive lock) was resolved by just always doing exclusive locks.
always-exclusive locks can be a problem if one has long-running (or frequent) backup jobs (or in general: repo-writing operations) and in parallel one wants to do other borg operations:
borg mount / list / info / extracttopics to research in this context could be: