mr@ogre.cica.indiana.edu (Michael Regoli) (03/05/91)
Configuration:
MicroVAX 3500 running 4.3 BSD (cica.cica.indiana.edu)
nntp 1.5.11
Cnews (13 Jan 1990 Version)
nn 6.4.12
I've got a problem and I'm having a hard time pinning it down.
nn is requesting articles in daemon mode (nnmaster -r) every ten
minutes. Sometime during the collection, the connection is lost with
the host. (nnmaster is running on the host, so it's talking to
itself.)
These are the messages in nnmaster's /usr/local/lib/nn/Log file:
...
N: Mar 4 15:06 (M): Lost connection to server cica.cica.indiana.edu:
No such file or directory
C: Mar 4 15:06 (M): Collect: 250 art, 99 gr, 279 s
N: Mar 4 15:20 (M): Lost connection to server cica.cica.indiana.edu:
No such file or directory
C: Mar 4 15:20 (M): Collect: 268 art, 110 gr, 228 s
N: Mar 4 15:33 (M): Lost connection to server cica.cica.indiana.edu:
No such file or directory
C: Mar 4 15:33 (M): Collect: 261 art, 88 gr, 148 s
...
Now, wouldn't it be lovely if we knew which "file or directory" is
missing? :) Has anyone seen this? Have any ideas on what is going
on? nntpd is not logging any strange messages in SYSLOG either.
(By the way, we've got the MAXLOADAVG at 20, so nntpd will not be
phazed by higher loads.) We've got plenty of diskfree on the spool
device as well.
Thanks, thanks, thanks!
--
michael regoli
mr@ogre.cica.indiana.edu
regoli@iubacs.BITNET
..rutgers!iuvax!ogre!mrjwright@cfht.hawaii.edu (Jim Wright) (03/06/91)
mr@ogre.cica.indiana.edu (Michael Regoli) writes: >... >N: Mar 4 15:06 (M): Lost connection to server cica.cica.indiana.edu: > No such file or directory >C: Mar 4 15:06 (M): Collect: 250 art, 99 gr, 279 s >N: Mar 4 15:20 (M): Lost connection to server cica.cica.indiana.edu: > No such file or directory >C: Mar 4 15:20 (M): Collect: 268 art, 110 gr, 228 s >N: Mar 4 15:33 (M): Lost connection to server cica.cica.indiana.edu: > No such file or directory >C: Mar 4 15:33 (M): Collect: 261 art, 88 gr, 148 s >... I was having the same trouble after changing from 6.4.11 to 6.4.12. Not only did it generate these messages, it failed to collect many groups at all. And many other groups it would only collect about 5% of the posted articles. I was on the verge of junking nn! I just changed from 6.4.12 to 6.4.13. In the process, I (intentionally) deleted GROUPS, MASTER, ACTIVE and DATA/*. Then I did an nnmaster -D. When it got to the group alt.desert-storm, it sucked over all the headers and then said that there were 0 articles! From then on every group reported 0 articles. By the speed at which it went through the articles, it was obvious that the entire header was not being sent. I ^C'd out and editted the GROUPS file to kill alt.desert-storm, removed alt.[d-z]* files from the DATA/ directory, and restarted the nnmaster. The same procedure was required for the group junk. Our nntp feed is running an old version of nntp. I asked them about upgrading, and they said "someday..." All seems to be working now, but I'm going to keep a close eye on this. It seems to me that nn is not reacting correctly to some sort of error message or bogus nntp message. The error message produced by nn is less than helpful. -- Jim Wright jwright@cfht.hawaii.edu Canada-France-Hawaii Telescope Corp.
dave@boingo.med.jhu.edu (David Heath) (03/07/91)
mr@ogre.cica.indiana.edu (Michael Regoli) writes: >In <jwright.668210717@cfht.hawaii.edu> jwright@cfht.hawaii.edu (Jim >Wright) writes: >>mr@ogre.cica.indiana.edu (Michael Regoli) writes: >>>... >>>N: Mar 4 15:06 (M): Lost connection to server cica.cica.indiana.edu: >>> No such file or directory >>I was having the same trouble after changing from 6.4.11 to 6.4.12. >>Not only did it generate these messages, it failed to collect many >>groups at all. And many other groups it would only collect about 5% >>of the posted articles. I was on the verge of junking nn! >All is well now that I've recompiled nn to grab articles directly >from the spool rather than from nntp itself. (Now looking back, I'm >not certain why I enabled nntp support on the server machine to begin >with. It must have been one of those days.) >>I just changed from 6.4.12 to 6.4.13. >Now I just realized I missed patch 13! Thanks, Jim-bo! ;) >>It seems to me that nn is not reacting correctly to some sort of error >>message or bogus nntp message. Actually, this sounds like a problem with nntp. It has a hard-coded limit of 4096 articles/group. When nnmaster tries to enter a group with more than 4096 articles (like junk on some systems), the remote nntpserver dies and breaks the connection. You can tell nnmaster not to collect the junk group by starting the daemon as: nnmaster -l -r C !junk "" -dave heath dave@boingo.med.jhu.edu
sob@tmc.edu (Stan Barber) (03/07/91)
In article <1991Mar6.195636.2408@boingo.med.jhu.edu> dave@boingo.med.jhu.edu (David Heath) writes: >Actually, this sounds like a problem with nntp. It has a hard-coded limit >of 4096 articles/group. When nnmaster tries to enter a group with more than >4096 articles (like junk on some systems), the remote nntpserver dies and >breaks the connection. This is incorrect in two respects: 1. NNTP prior to 1.5.11 had a user-setable limit of the number of articles per group. The default was 4096. There is nothing in the code that kept it from being larger. Most people just didn't reset the size. 2. NNTP 1.5.11 has a feature to remove this limit altogether. It is kinda beta since we didn't have a chance to test it throughly, but no one has sent nntp@tmc.edu mail saying it didn't work. -- Stan internet: sob@bcm.tmc.edu Director, Networking Olan uucp: {rutgers,mailrus}!bcm!sob and Systems Support Barber Opinions expressed are only mine. Baylor College of Medicine