* Wilkinson, Tim <Tim.Wilkinson AT dsto.defence.gov DOT au> [2006-05-18 10:47]:
> Hi,
>  
> I'm trying to understand how this actually works as I think it works a
> little differently from how I thought it worked. We have a couple of
> polices that do a full backup weekly, with the frequency set to 1 week.
> However, the jobs didn't run on the weekend; there were no errors, they
> just didn't run.
> The only reason I can think of this happening is the job falling outside
> of the start window. From what I've read, the frequency is the minimum
> amount of tim that must pass before the next schedule runs, which means
> that we should probably set the weekly jobs to having a frequency of 6
> days to avoid it falling outside the start window. Would this be
> correct?

Yes.  Generally speaking, you should make the frequency just larger than
the expected time to complete the backup (e.g. a weekly backup should
have a frequency of 2 days, a daily should have a frequency of 18
hours).  This allows the backup sufficient time to complete, but the
window of availablilty will open well before the next scheduled backup.

Remember, the frequency is measured from the _end_ of the last succesful
backup.  A weekly backup that has a frequency of 1 week with a 10 hour
backup window that takes 12 hours to complete will "walk" outside of its
window so that the following week, when it gets it's first chance to
check, will have an open backup window but the frequency will still be
within the "1 week" period, and it won't run.  

Your backups check two things: am I in the start window, AND has it been
longer than the listed frequency.  If either of those is false, it will
not run.

-- 
David Rock
david AT graniteweb DOT com