alexis@panix.uucp (Alexis Rosen) (11/30/90)
In general, I continue to be very happy with Cnews. However, I've noticed
a few really strange things and I wonder if anyone knows why they happen.
First of all, once in a while new articles never get posted. Using rn, I
followup to an article (or make a new one), do everything normally, and send
off the article. Something happens, since shelling out of rn and running ps
shows various cnews-things (like inject and anne.jones) running. But the
article just vanishes into a black hole. Reposting works.
I've seen this behavior four times in three months. Twice on the second day
after I installed Cnews, once about a month ago, and yesterday (in fact, it
was the posting about the "-L" bug in rn, in this group).
The other problem is the (very infrequent) dropped newsbatch. I've only seen
this twice. A couple of days ago, I logged on and looked around the system
with ps. It showed me that cunbatch was running, and so was "sleep 300".
This only happens when cunbatch is told by spacefor that there's not enough
room for an incoming batch. With close to 60MB free on the spool disk, this
was simply not possible, and running spacefor by hand gave a correct response-
plenty of room. but sure enough, a few minutes later there was mail from
cunbatch warning me that the batch had been lost. Why???
I am sure that spacefor is not buggy. After all, it has worked well for three
months, during which times there were days when the spool was full, as well
as when it had lots of room.
Any ideas?
---
Alexis Rosen
Owner/Sysadmin, PANIX Public Access Unix, NY
{cmcl2,apple}!panix!alexishenry@zoo.toronto.edu (Henry Spencer) (12/02/90)
In article <1990Nov30.111026.66@panix.uucp> alexis@panix.uucp (Alexis Rosen) writes: >First of all, once in a while new articles never get posted. Using rn, I >followup to an article (or make a new one), do everything normally, and send >off the article. Something happens, since shelling out of rn and running ps >shows various cnews-things (like inject and anne.jones) running. But the >article just vanishes into a black hole... >The other problem is the (very infrequent) dropped newsbatch... [symptoms] >This only happens when cunbatch is told by spacefor that there's not enough >room for an incoming batch. With close to 60MB free on the spool disk, this >was simply not possible... A possible reason for both of these is that your system is hitting the ceiling on some other resource, like processes or open files, and programs are failing as a result. We know there can be problems with spacefor if it hits such a ceiling; the result tends to be a "no space available" report. It's plausible for inews to hit similar difficulties, although I'd have to put heads together with Geoff to produce a detailed explanation. -- "The average pointer, statistically, |Henry Spencer at U of Toronto Zoology points somewhere in X." -Hugh Redelmeier| henry@zoo.toronto.edu utzoo!henry
alexis@panix.uucp (Alexis Rosen) (12/03/90)
henry@zoo.toronto.edu (Henry Spencer) writes: >alexis@panix.uucp (Alexis Rosen) writes: >>First of all, once in a while new articles never get posted. Using rn, I >>followup to an article (or make a new one), do everything normally, and send >>off the article. Something happens, since shelling out of rn and running ps >>shows various cnews-things (like inject and anne.jones) running. But the >>article just vanishes into a black hole... >>The other problem is the (very infrequent) dropped newsbatch... [symptoms] >>This only happens when cunbatch is told by spacefor that there's not enough >>room for an incoming batch. With close to 60MB free on the spool disk, this >>was simply not possible... > >A possible reason for both of these is that your system is hitting the >ceiling on some other resource, like processes or open files, and programs >are failing as a result. We know there can be problems with spacefor if >it hits such a ceiling; the result tends to be a "no space available" report. >It's plausible for inews to hit similar difficulties, although I'd have to >put heads together with Geoff to produce a detailed explanation. That's not it. I though of that too, but we don't have more than 1/3 the max number of processes possible at any given time. As for open files, this seems much more reasonable- with the uuxqt bug in A/UX 2.0 (before I wrote a wrapper script to deal with it) we were seeing that all the time. However that bug has been dealt with, so we're not coming anywhere near the maximum # of open files either. Any other ideas? As for Pnews failing, I saw it again today. I did some tests, and while I'm not too sure of anything yet, it seems to have something to do with suspending rn immediately after returning from Pnews. In other words, Pnews is no longer running, I suspend back to a shell, and the article vanishes. Any ideas what this could mean? --- Alexis Rosen Owner/Sysadmin, PANIX Public Access Unix, NY {cmcl2,apple}!panix!alexis
henry@zoo.toronto.edu (Henry Spencer) (12/04/90)
In article <1990Dec3.015827.19038@panix.uucp> alexis@panix.uucp (Alexis Rosen) writes: >As for Pnews failing, I saw it again today. I did some tests, and while I'm >not too sure of anything yet, it seems to have something to do with suspending >rn immediately after returning from Pnews. In other words, Pnews is no longer >running, I suspend back to a shell, and the article vanishes. Any ideas what >this could mean? It could mean that processes run on behalf of Pnews, e.g. the relaynews that inews fires up, are still running and are not prepared to cope with bizarre job-control signals. I can't really venture a more detailed opinion, since job control is such a hideous mess that I can't look at it without getting sick. I don't doubt that there is ample potential for trouble there. -- "The average pointer, statistically, |Henry Spencer at U of Toronto Zoology points somewhere in X." -Hugh Redelmeier| henry@zoo.toronto.edu utzoo!henry