• That wouldn't be enough to recover the database. If there was a long running transaction that started before the full backup did, only backing up the transaction log generated during the backup would not allow the transaction to be rolled back correctly during the restore operation

    I was under the impression that at the start of a backup, a checkpoint is issued and a marker is placed in the trans log indicating that a database wide lock was obtained (to ensure there are no running transactions), this lock is then released and the backup is performed. At the end of the backup, another checkpoint is issued (another lock obtained and released) and the section of the transaction log between the two checkpoints is backed up. Has this changed between 2005 and 2008 or did I dream all that 😀