> Support had me stop Netbackup and remove the file and then start
> Netbackup.  I was told it would recreate that "pempersist" 
> file.  It did
> recreate the file but did not help get my fulls to run.  I was using
> nbpemreq -predict -date XX/XX/XXXX XX:XX:XX with different dates and
> times to try and forecast what would run.  I was very careful 
> to use the
> -updatepolicies after any changes I made.  This is important to note
> since there is a small possibility that some of the errors I saw after
> the initial failure was from the nbpemreq command.  I tried 
> to wait and
> see if the fulls ran at all, but after making a few changes 
> such as the
> pempersist and waiting a week and then trying something else, I had to
> change my policies in order to have data backed up.  
> 
> I was always able to manually run the policy.  Which I did for all
> affected policies when I discovered the issue.

Jonathan, I duplicated your symptoms today.

Took a Solaris 9 NetBackup 5.1MP4 lab server, added the simplest
possible policy to it with a frequency-based schedule and over-midnight
window.  Copied that to a new policy and changed only the freq attribute
to calendar (more on this later).  Made several variants of cal- and
freq-based, but that's not important here.  Anyway, ran them both from
bpbackup -i and they worked fine.

Upgraded the server to 6.0MP2.  All good running immediate backups.  But
the calendar-based ones (new copy, with no "immediate" history) don't
run by the scheduler.  Hmmm.  Futzed with a lot of things, but mostly
relied on blowing away pempersist.  That perks up the frequency-based,
but no-go on the calendar-based.  Soooooo, _this_ is what Jonathan
meant...  Changed all over-midnight windows (in the cal- policies first,
later on everything in the system) to not cross midnight (0000-2355).
Still no scheduled calendar-based backups.

Uninstalled MP2.  Nothing changed.  "Weird," sez I.  There _have_ to be
a ton of customers using calendar-based scheduling even though _I_ think
it rots.  What on earth could my setup have in common with yours to
where I'm seeing this on 6.0base and you on 6.0MP2?  

One possibility came to me when I started writing this mail.  Since I
don't use calendar-based scheduling, I muffed the basics:  no run days.
DOH!  Set a bunch o' rundays and the calendars fired off just fine at
2344 and 0000, as expected.

Changed the calendar schedules to span midnight (2300-2200).  They fired
up just fine.  Whew!  At least base 6.0 handles calendar-based schedules
correctly, whether they span midnight or not.  Speaking of midnight, I
didn't see the cal schedules run again at midnight.  Maybe this means
that 6.0 "fixes" the logic which befuddles first-timers, or maybe I
didn't set up correctly to determine the behavior.  Let's see what
happens when I put MP2 back in...

Okay, Jonathan, I have results:  NetBackup 6.0MP2 starts calendar-based
schedules just fine, whether the schedule spans midnight or not.  Just
got through the run of the 2300-2200 window.  (And no 0000 rerun, so it
looks as if there's now logic to treat cal-based windows as an entity,
as they are in freq-based.  This might be documented; I don't have time
to check.)  Looks like your problem is specific to your setup.  Good
luck.

The only way I've been able to duplicate your problem is by not having
appropriate rundays checked; perhaps that a good place to dig deeper.
Maybe post a bppllist.