Are there any messages in the diag? Is it an online backup? Is it possible that somewhere along the way the archiving process stopped working - maybe the disk is full? Or that there's a transaction which spans all primaries and secondaries and cannot complete? If the database is across multiple nodes then ensure that you're looking on all nodes. There has to be a transaction out there which isn't committing. When the backup fails does it cite the log full condition as the reason?