biocca@bevsun.bev.lbl.gov (Alan Biocca) (01/17/91)
I have experienced two problems with the trn package. The first was the mthreads database builder overloaded our nntp server so badly that it ceased to unbatch incoming news and overflowed its disk. This I solved by putting a delay in mthreads to limit the rate it can request headers at. The second problem is that if trn loses the connection to the nntp server (typically a timeout when a user leaves trn running but doing nothing for awhile) it incorrectly interprets the errors and unsubscribes the user from all newsgroups. I have not investigated this second problem, and potential users here have indicated they'd rather wait than switch until this is resolved. I have installed one set of (standard) patches to the software, bringing it to version 1.1 if I remember correctly. Comments? Solutions? Thanks, Alan K Biocca WB6ZQZ
zeeff@b-tech.ann-arbor.mi.us (Jon Zeeff) (01/17/91)
>I have experienced two problems with the trn package. The first was >the mthreads database builder overloaded our nntp server so badly that >it ceased to unbatch incoming news and overflowed its disk. This I >solved by putting a delay in mthreads to limit the rate it can request >headers at. I found that performance was bad when the nntp server didn't run updatemin (with C News). I believe it causes requests for thousands of articles that don't exist. Also, the sever might want to run a "find /usr/spool/news -mtime +30 -exec rm {} \;" now and then. -- Jon Zeeff (NIC handle JZ) zeeff@b-tech.ann-arbor.mi.us
palkovic@linac.fnal.gov (John A. Palkovic) (01/18/91)
In article <PPT+T9+@b-tech.uucp> zeeff@b-tech.ann-arbor.mi.us (Jon Zeeff) writes: >Also, the sever might want to run a "find /usr/spool/news >-mtime +30 -exec rm {} \;" now and then. It is advisable to exclude the news subdirectory of /usr/spool/news from this find command so that all the important articles in news.announce.newusers are not removed. -John -- palkovic@linac.fnal.gov Hmmm... Deleted the trailing cruft.
henry@zoo.toronto.edu (Henry Spencer) (01/18/91)
In article <PPT+T9+@b-tech.uucp> zeeff@b-tech.ann-arbor.mi.us (Jon Zeeff) writes: >I found that performance was bad when the nntp server didn't run >updatemin (with C News). I believe it causes requests for thousands >of articles that don't exist. This is true. The problem is that NNTP provides no easy way to ask for a crucial piece of information: "which of the articles in the min..max range actually exist?". Keeping min up to date, with upact or updatemin, helps. (It does not actually solve the problem, since the min..max range is not necessarily densely populated, but it helps. The only real fix is to fix the NNTP protocol.) >Also, the sever might want to run a "find /usr/spool/news >-mtime +30 -exec rm {} \;" now and then. Shouldn't be necessary under C News, unless your system crashes a lot. C News doesn't lose articles, unless the underlying system fails on it. Also, again under C News, a better way to do this is to run the (new) "addmissing" program occasionally. That adds any orphaned articles to the history database, so they can expire normally, without disturbing articles that legitimately have an expiry time longer than 30 days. It's also typically faster. -- If the Space Shuttle was the answer, | Henry Spencer at U of Toronto Zoology what was the question? | henry@zoo.toronto.edu utzoo!henry