Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!ucsd!ucbvax!ATC.BOEING.COM!snicoud
From: snicoud@ATC.BOEING.COM (Stephen Nicoud)
Newsgroups: comp.sys.ti.explorer
Subject: RE: hosed networks
Message-ID: <19891120215315.1.SLN@SKAGIT.atc.boeing.com>
Date: 20 Nov 89 21:53:00 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Distribution: inet
Organization: The Internet
Lines: 43


    Date: 20 Nov 89 21:01:22 GMT
    From: pasteur!jwz%teak.berkeley.edu@ucbvax.berkeley.edu.ARPANET  (Jamie Zawinski)


    This keeps happening to me: at random times, certain hosts out in net-land
    (unix machines) will appear to stop responding.  No connections can be 
    opened to them, no matter how many times hit resume or call IP:RESET.
    The problem doesn't go away after several hours.  However, some other hosts
    will respond, and I don't see the pattern.

    I touch three IP hosts regularly: spice.cs.cmu.edu, which is on the other
    side of the country; teak.berkeley.edu, which is upstairs; and 
    pasteur.berkeley.edu, which is also local.  The most common (but by no 
    means exclusive) way that my network gets wedged is that Spice and Pasteur
    will seem to stop responding, but Teak won't; and other machines have no
    problem connecting to these, so it's not the network itself.

    Warm-booting often causes this, but it happens spontaniously as well.
    Cold-booting always makes it go away.  I'm using IP 3.43.

    Does this happen to anyone else?

	    -- Annoyed.

Dear Annoyed, :-)

I don't know if this is related, but when a failed IP connection occurs
to a host, I've discovered that a property (:connection-failures) is
added to the host object's property list.

	(send (si:parse-host "HOST") :connection-failures)

Removing this property allows me to try again (otherwise, subsequent
attempts fail immediately without trying to connect).

	(remprop (si:parse-host "HOST") :connection-failures)

I'd be interested in what you do or do not find out.

Hope this black magic helps. :-)

	-- Dr. Net Voodoo