Recently this question came up with a client that I’ve been working with. In this case the problem was that disk space alarms were going off for the disk which held the database backups. The proposed solution was to change the log backup retention from 24 hours to 16 hours.
The problem with this solution, which I pointed out is that from 5pm until midnight you now have no way to do any log restore operations which means that you’ve lost your point in time restore capabilities.
Log and full backups shouldn’t be deleted from the disk until the full backup from the next cycle of backups has been completed. This way you are sure that you’ve always got a full backup to restore from. In a perfect world I’d like to see 2-3 days worth of full and t-log backups kept on disk so that if there is ever a problem with a corrupt database backup you can restore from the prior full backup and then roll two cycles of t-log backups forward.
Denny
