[news.software.b] Bug in unbatching

ccea3@rivm.UUCP (Adri Verhoef) (11/15/88)

I found a dead.letter file in /usr/spool/uucp/.Xqtdir, containing:

remote execution	[uucp job rivm13C260f (11/10-15:41:24)]
	rnews 
exited with status 1


	===== stderr was =====
write of 1251 to pipe returned -1rnews: write: No space left on device
inews: : Inbound news is garbled



It's news B 2.11.14, running on a 3B2/310 (System V 2.0.4).
This 3B2 (Memory size: 2 Mb) was said to be
"out of swap space" continuously; after upgrading
the Informix version, messages like:
	sh: cannot fork: no swap space
	fork: Not enough space
	trimlib: cannot fork: no swap space
kept appearing the days before this happened.
Could a heavy application (like Informix?) cause this trouble?

But during the unbatching there probably was a lack of
disk space, not a swap-space problem, which caused the
news not to be accepted (thus to be garbled).

Further questions:
can the unbatching be halted by the news software
and, probably most difficult, revived by uuxqt,
thus causing correct unbatching at a time when
things are back in order again, according to the
news software?
Could this be solved in B-news 3.0?  eric@snark?

henry@utzoo.uucp (Henry Spencer) (11/17/88)

In article <1133@rivm05.UUCP> ccea3@rivm.UUCP (Adri Verhoef) writes:
>can the unbatching be halted by the news software
>and, probably most difficult, revived by uuxqt,
>thus causing correct unbatching at a time when
>things are back in order again, according to the
>news software?

Uuxqt won't revive anything; it wants to see the stuff run once and get
out of its hair.  The right approach is the one that is increasingly in
use today (at least, C News does it, I think modern B News can do it,
and I'd be surprised if 3.0 doesn't):  have the thing run by uuxqt just
stash the data somewhere, with news processing done separately later.
This way, a problem in news processing doesn't foul up uuxqt.

However, in your specific case, "out of swap space" means the system
feels it is overcommitted and doesn't want to start any more processes
right now.  It is *very* hard to deal with this gracefully.  Your swap
space is simply too small for the things you're running.