>In my usual style for "archive" backups, I go straight to tape by
>specifying no holding disk at all. ... I would think the entries without a
>starttime would go ahead and write to tape as there is no holding
>disk for which to wait. Yet, things still didn't appear to work this
>way when I attempted things first in this fashion.
I see what the problem is now. There is a cascade of tests when looking
for something to do and not going direct to tape (is there enough
bandwidth, is there enough disk space, etc). Those tests are ordered
in such a way that, even though you didn't have any holding disk, the
delayed entries were noticed and a flag set saying "I saw some things
that are delayed". That put Amanda to sleep waiting on the starttime,
at which point it noticed "oops, I can't fit this in the holding disk"
and finally got going.
My previous idea of going ahead with direct to tape might fix this, or
the fix might involve only setting that flag if there is any chance of
putting the delayed backup in the holding disk. But how to figure that
out is going to take more than the couple of minutes I spent glancing
at that section of code.
I've added more notes to the TODO list so at least the issue won't
get lost. Thanks for pestering me (seriously :-) about this until the
real issue finally popped out.
>Paul Yeatman
John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
