[news.admin] rn and NFS; large search strings and mangled .newsrc's

jkimball@SRC.Honeywell.COM (John Kimball) (11/30/89)

(Maybe a news.software.readers would make sense -- where *should* you post
something about newsreader software?)

We've been having a few persistant problems with rn.

Context:  lots of Sun3 workstations, one serving as a news server.  All
running SunOS 4.0.3.  News reading is done by NFS-mounting the spool
tree;  news posting is done with the NNTP mini-inews.  People read news
with rn (patchlevel 40) or with a homegroup GNUemacs newsreader mode.
Many people also use a homegrown 'news biff' icon which periodically
does an 'rn -c'.

Here's the problems we've seen with rn:

1) When the news server gets swamped (high load average), rn processes
   frequently lock up.  'ps' shows them in 'D' state, "disk (or other
   short term) waits".  These guys are basically unkillable.

   This doesn't seem to happen to other processes doing NFS accesses.  Does
   rn do something wierd?  I don't know where to look.

2) Possibly related:  occasionally 'rn' suddenly believes that your .newsrc
   doesn't exist, and so it merrily creates you a new one (blowing away the
   one which did exist.)  Often it is the 'news biff' invocation of "rn -c"
   which does this.  I tracked down where the message ("running newsetup")
   is coming from, in rcstuff.c -- a .newsrc is built when fopen(rcname,"r")
   returns "Nullfp".  Why would it return Nullfp when the file exists?

   I've put in a patch which refuses to do the newsetup if rn was invoked
   "-c", but I'd like to know why rn and NFS aren't getting along.

3) Likely completely unrelated:  I have some very long search patterns in my
   .rnmac file. Recently they've gotten long enough that my .newsrc file is
   getting corrupted -- a random newsgroup line is not written out correctly,
   instead a string containing pieces of a pattern is written in its place.
   Has anybody seen this before?  (Bumping up LBUFLEN doesn't seem to help.)

Thanks in advance!
						John Kimball

Inet: jkimball@src.honeywell.com          Honeywell Systems and Research Center
      postmaster@src.honeywell.com, etc   Computer Sciences/Software Technology
uucp: <any-smart-host>!srcsip!jkimball    3660 Technology Drive, MN65-2100
phone: 612-782-7343  fax: 612-782-7438    Minneapolis, MN  55418-1006