*BSD News Article 59033


Return to BSD News archive

#! rnews 3562 bsd
Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!news.bhp.com.au!mel.dit.csiro.au!munnari.OZ.AU!news.ecn.uoknor.edu!paladin.american.edu!europa.chnt.gtegsc.com!news.kreonet.re.kr!overload.lbl.gov!lll-winken.llnl.gov!uwm.edu!vixen.cso.uiuc.edu!newsfeed.internetmci.com!xmission!newsfeed.kdcol.net!usenet
From: dave@mail.kdcol.com (Dave E Martin)
Newsgroups: comp.unix.bsd.freebsd.misc,comp.unix.solaris,news.software.nntp
Subject: INND takethis really long pauses problem (more info)
Date: Mon, 15 Jan 1996 05:31:40 GMT
Organization: KDC-Online
Lines: 64
Distribution: inet
Message-ID: <30f9e36c.10513600@newsfeed.kdcol.com>
References: <30e8909b.1118570368@newsfeed.kdcol.com> <30ec12f8.10926492@news.kdcol.com>
Reply-To: dave@mail.kdcol.com
NNTP-Posting-Host: bryce.kdcol.net
X-Newsreader: Forte Agent .99c/16.141
Xref: euryale.cc.adfa.oz.au comp.unix.bsd.freebsd.misc:12143 comp.unix.solaris:56563 news.software.nntp:18870

I am adding this to comp.os.solaris in hopes of some insight, and also
see below the quote.

>On Tue, 02 Jan 1996 02:13:39 GMT, dave@mail.kdcol.com (Dave E Martin)
>wrote:

>After enabling tracing (ctlinnd trace innd on) our log indicates the
>following:

> < takethis <articleid>
> > 239 <articleid>
> < takethis <next_articleid>
> > 239 <next_articleid>

>and so on, with several per second according to the timestamp.
>But then there will be
> < takethis <later_articleid>
> 45 minute pause, according to timestamps and observed behaviour
> > 239 <later_articleid>
> then the several per second rate resumes.

>This happens several times per day, with the pauses varying in length
>(anywhere from 5-45 minutes and possibly longer).  During the pause,
>innd appears to operate normally, will accept feeds from other sites,
>will accept connections from readers, etc.

>We are running 1.4unoff3 with FreeBSD 2.1.0 and the feeding site is
>running 1.4unoff3 with a machine that reports Unix System 5 Release
>4.0 (suspected to be a sun of some sort).

>I have tried it with and without MMAP compiled in.

>We recently got behind several days in our newsfeed (due to murphy's
>law and the christmas holidays), and now we are trying to get caught
>back up.  These pauses are frustrating.  What could be causing them?

>Thanks for considering the problem

One person has stated that they have seen this behavior feeding news
from a FreeBSD box to a Solaris box and suspects unoff3.  (May I
suggest running innxmit with -d and seeing if any useful info is
provided?)

Another person has suggested that when the pauses occur, the article
being transferred is large, accounting for the pause, or that
connectivity to the other end has been lost.

First, connectivity is still present.  Other traffic to/from the site
goes fine.

Second,  It does, in fact, appear that the article at the time of the
pause is large, anywhere from .5 to 1.5 megabytes.  However.
being *extremely* pessimistic, this should take no more than 2 minutes
to transfer over our connection.  The feeding site is very close to
us, (we share a common router).  Furthermore, there doesn't appear to
be any active transfer of data during the pause.

Is there any further method of diagnosing the problem or does anyone
know what may be causing this?
----
Dave E Martin        Main groups and alt, wrapped in 2,500,000 sectors
P11 XXIII KB7VBD     of spinning cylinders, all alone on the disk.  A home
dave@mail.kdcol.com  for diplomats, hustlers, entrepreneurs, and wanderers,
DM1004               it is our last, best hope for freedom of speech.