[news.software.b] *SERIOUS* expire problem...

george@taca.ece.cmu.edu (George Cebulka) (03/08/91)

    
    I'm having a problem with articles not getting expired. I have many articles
hanging around *way* past their expire times. This is causing my spool partition
to fill up (I'm at 96% now).
    When I run "newsfrom 02/24/91 02/28/91" it returns exactly one article.
However, when I cd into a spool directroy for a newsgroup there are many the
creation dates show many files which fit into this time frame. Most of them 
seem to be from around the Feb/24-Feb/28 time frame. It would almost seem that
few, if any, articles are getting deleted since that Feb 26.
	Thank you
	    George Cebulka, ECE Facilities

    Here is a copy of my explist file.
 # modified version, no archiving
 #
 # hold onto history lines 14 days, nobody gets >365 days
 /expired/			x	30	-
 #/bounds/			x	0-1-120	-


 # default:  7 days and ditch'em
 cmu.ece.general			x	0-45-46	-
 cmu.ece.class			x	0-120-121	-
 ece				x	0-120-121	-
 cmu				x	0-21-22	-
 # Some of the stuff in news.announce is important. Let it hang around for a while.
 news.software.nntp		x	0-35-36	-
 news.software.b			x	0-35-36	-
 news.announce.conferences	x	0-30-31	-
 news.announce.important		x	0-120-121	-
 news.announce.newusers		x	0-120-121	-
 news.announce.newgroups		x	0-15-16	-
 # Newsgroups we don't want to keep for more than a week.
 #
 # Newgroups that we don't want or need to keep more than 3 days
 control				x	0-2-2	-
 # This is the default. Keep stuff for a week, then get rid of it
 bionet				x	0-7-7	-
 talk				x	0-7-7	-
 # Stuff to get rid of early. This is an emergency measure.
 comp				x	0-6-6	-
 comp.binaries			x	0-3-4	-
 comp.dcom			x	0-3-4	-
 comp.lang			x	0-3-4	-
 comp.amiga			x	0-2-3	-
 comp.sys			x	0-3-4	-
 comp.sources			x	0-4-5	-
 comp.unix			x	0-4-5	-
 comp.windows			x	0-4-5	-
 comp.mail			x	0-3-4	-
 alt				x	0-6-6	-
 alt.activism			x	0-3-4	-
 alt.flame			x	0-3-4	-
 alt.folklore			x	0-3-4	-
 alt.sex			x	0-3-4	-
 alt.sources			x	0-3-4	-
 alt.tv				x	0-3-4	-
 soc				x	0-3-4	-
 #
 all				x	0-7-7	-

 Here is a partial listing of two of my directories.
 /nnspool/comp/binaries/ibm/pc
 total 2779
 drwxrwxr-x  3 news         1536 Mar  6 19:40 .
 drwxrwxr-x  3 news          512 Nov 26 11:53 ..
 -rw-rw-r--  1 news        25014 Feb 26 03:44 3752
 -rw-rw-r--  1 news        20978 Feb 26 03:44 3753
 -rw-rw-r--  1 news        44563 Feb 26 03:52 3754
 -rw-rw-r--  1 news        45034 Feb 26 03:52 3755
 -rw-rw-r--  1 news         5662 Feb 26 03:52 3756
 -rw-rw-r--  1 news        21029 Feb 26 03:52 3757
 -rw-rw-r--  1 news         7251 Feb 26 03:52 3758
 -rw-rw-r--  1 news        42878 Feb 26 16:33 3759
 -rw-rw-r--  1 news        40865 Feb 26 16:41 3760
 -rw-rw-r--  1 news        40890 Feb 26 17:34 3761
 -rw-rw-r--  1 news        40896 Feb 26 17:34 3762
 -rw-rw-r--  1 news        40103 Feb 26 17:34 3763
 -rw-rw-r--  1 news        22032 Feb 26 17:51 3764
 -rw-rw-r--  1 news        15755 Feb 26 17:51 3765
 -rw-rw-r--  1 news        46780 Mar  1 11:52 3766
 -rw-rw-r--  1 news        46457 Mar  1 12:03 3767
 -rw-rw-r--  1 news        45993 Mar  1 12:31 3768
 total 241

 /nnspool/soc/culture/japan
 drwxrwxr-x  2 news         1536 Mar  7 19:36 .
 drwxrwxr-x 34 news         1024 Feb 18 02:16 ..
 -rw-rw-r--  1 news         6362 Dec 18 09:04 6553
 -rw-rw-r--  1 news         2085 Dec 18 09:32 6554
 -rw-rw-r--  1 news         2706 Dec 18 10:03 6555
 -rw-rw-r--  1 news          959 Jan 14 16:40 6728
 -rw-rw-r--  8 news         3237 Feb 24 01:46 6995
 -rw-rw-r--  1 news         1967 Feb 24 03:30 6996
 -rw-rw-r--  1 news         1647 Feb 24 05:21 6997
 -rw-rw-r--  8 news         2308 Feb 24 18:02 6998
 -rw-rw-r--  1 news         1935 Feb 24 20:03 6999
 -rw-rw-r--  3 news         1503 Feb 25 11:01 7000
 -rw-rw-r--  1 news         1429 Feb 25 12:17 7001
 -rw-rw-r--  1 news         1144 Feb 25 19:48 7002
 -rw-rw-r--  8 news         2082 Feb 26 00:39 7003
 -rw-rw-r--  1 news         1067 Feb 26 00:50 7004
 -rw-rw-r--  1 news         2731 Feb 26 00:54 7005
 -rw-rw-r--  1 news         1331 Feb 26 01:36 7006
 -rw-rw-r--  1 news          757 Feb 26 02:04 7007
 -rw-rw-r--  1 news          683 Feb 26 02:19 7008
 -rw-rw-r--  1 news          816 Feb 26 10:17 7009
 -rw-rw-r--  1 news         1592 Feb 26 11:50 7010
 -rw-rw-r--  1 news         3342 Feb 26 14:16 7011
 -rw-rw-r--  1 news         1031 Feb 26 17:46 7012
 -rw-rw-r--  1 news         2935 Feb 26 19:33 7013
 -rw-rw-r--  6 news         2609 Feb 26 19:44 7014
 -rw-rw-r--  7 news         1918 Feb 26 19:44 7015
 -rw-rw-r--  8 news         7830 Feb 26 19:44 7016
 -rw-rw-r--  8 news         2270 Feb 26 19:44 7017
 -rw-rw-r--  8 news         4068 Feb 26 19:44 7018
 -rw-rw-r--  9 news         5286 Feb 26 21:32 7019
 -rw-rw-r--  1 news         1944 Feb 28 17:03 7033
		 .
		 .
		 .
 -rw-rw-r--  1 news         1059 Mar  1 00:46 7034








-- 
George Cebulka, ECE Facilities

molenda@s1.msi.umn.edu (Jason Molenda) (03/08/91)

george@taca.ece.cmu.edu (George Cebulka) writes:

>  I'm having a problem with articles not getting expired. [...]
>    When I run "newsfrom 02/24/91 02/28/91" it returns exactly one article.
>However, when I cd into a spool directroy for a newsgroup there are many the
>creation dates show many files which fit into this time frame. Most of them 
>seem to be from around the Feb/24-Feb/28 time frame. It would almost seem that
>few, if any, articles are getting deleted since that Feb 26.

News Hiccup.  It happens all the time, although there is No Good Reason
for it to; in theory it shouldn't happen at all.

The problem is most likely that the messages are not listed in the
history file (/usr/lib/news/history, usually).  Expire really just
scans through the history file when it is deciding what articles should
be expired.  If, by some magic fate, an article is not listed in
the history file, then expire will never delete it.  newsfrom probably
does the same thing (I've never used it so I'm not sure exactly how
it works).  Again, if the messages aren't in the history file, newsfrom
won't list them.

The easiest solution is to run 'addmissing' if you are running a
recent version of Cnews (I forget when addmissing was added.. fairly
recently I believe).  This will run around the spool directory checking
for messages that Cnews doesn't know about (or something to that effect).

Alternatively, run a manual find(1) to delete messages older than <x>
days and delete them.  The problem with this is that if you archive
any newsgroups or keep them longer than <x> days, you'll have to do
some weird stuff to keep find(1) away from those messages.
-- 
More information that you couldn't have existed another day without, from:
Jason Molenda, Tech Support, Iris & News Admin, Minnesota Supercomputer Inst
molenda@s1.msi.umn.edu || molenda%msi.umn.edu@umnacvx.bitnet
"And remember: Evil will always prevail, because Good is dumb." -- Spaceballs

molenda@s1.msi.umn.edu (Jason Molenda) (03/08/91)

molenda@s1.msi.umn.edu (Jason Molenda) writes:

>george@taca.ece.cmu.edu (George Cebulka) writes:

>>  I'm having a problem with articles not getting expired. [...]

>The easiest solution is to run 'addmissing' if you are running a

>Alternatively, run a manual find(1) to delete messages older than <x>
>days and delete them. 

My thinking cap wasn't on.  There is a third (rather undesirable) option.
You can rebuild your history file.  It really isn't _that_ bad on
most systems these days; just set up an atjob to do it at some obscenely
late hour of the night and you probably won't even notice.

The only drawback is that you have increased chances of getting
slightly-old articles at your site that have already been expired
but would have still been in the history file normally.  Not usually
a big problem because your neighbour sites will usually 'shield' you
from any serious problems with older articles. (to an extent.. it
isn't Good Policy to rely on your neighbours to keep their expired
Message-IDs around a long time.)
-- 
More information that you couldn't have existed another day without, from:
Jason Molenda, Tech Support, Iris & News Admin, Minnesota Supercomputer Inst
molenda@s1.msi.umn.edu || molenda%msi.umn.edu@umnacvx.bitnet
"And remember: Evil will always prevail, because Good is dumb." -- Spaceballs

henry@zoo.toronto.edu (Henry Spencer) (03/09/91)

In article <1991Mar8.015811.243@fs7.ece.cmu.edu> george@taca.ece.cmu.edu (George Cebulka) writes:
>However, when I cd into a spool directroy for a newsgroup there are many the
>creation dates show many files which fit into this time frame. Most of them 
>seem to be from around the Feb/24-Feb/28 time frame. It would almost seem that
>few, if any, articles are getting deleted since that Feb 26.

It sounds like something is badly wrong with your news system, and the
history file lacks entries for those articles.  That is *not* a normal
situation; you should investigate at once.  Are current articles getting
history entries properly?  If not, that's the problem.  (In particular,
if you are running System V by any chance, check to see if your history
has hit the default ulimit.)

Running addmissing is the preferred solution.  Rebuilding your history
file is more drastic but would have the same effect in this situation.
However, you really need to understand the problem first, or there is
a good chance it will recur.
-- 
"But this *is* the simplified version   | Henry Spencer @ U of Toronto Zoology
for the general public."     -S. Harris |  henry@zoo.toronto.edu  utzoo!henry