brian@radio.astro.utoronto.ca (Brian Glendenning) (11/29/89)
I have been running a demand-based expire when I switched to Cnews
alpha (with input batches arriving via mail) a while back. That is,
before I processed a set of batches, I'd look at the disk space
situation, if it was insufficient to handle the queued batches (on a
different file system) it would run an expire, if it was still
insufficient it would expire again harder etc.
As I see it there are two potential advantages to this:
1) You might be able to expire less often on average. (but with
the combination Cnews+dbz(incore) this is hardly a consideration!)
2) You can run your news spool directories close to the limit
without worrying about it. We have a smallish, ~30M, news
partition that I don't want to expire more conservatively than
I need to, and running it to the limit without it needing
any attention is a real win to me.
[ 3) If a feed goes down you can read old news for a long time! ]
I have recently moved our inbound feed to a nntp based feed. Works
nicely. I am going to add a #ifdef MINFREEPROG to specify a program to
run to try to free up disk space if it drops below minfree. This seems
like the cleanest approach to my current set up, i.e. why encumber
newsrun when nntpd knows the disk situation.
Comments? Are there better solutions to demand based expire? Does
this sound useful enough that I should send the (trivial) patches to
someone, or is this a local hack that shouldn't otherwise see the
light of day?
--
Brian Glendenning - Radio astronomy, University of Toronto
brian@radio.astro.utoronto.ca uunet!utai!radio!brian glendenn@utorphys.bitnet