SUMMARY: 2.4 Networking problems

From: ksa5w@server.cs.virginia.edu
Date: Wed Jul 12 1995 - 06:31:21 CDT


I recently posed the following question about my workstation and some
networking problems I was having:

        I recently upgraded my Sparc 10 to Solaris 2.4, and everything was
        fine for a few days (ie. network responses were fast like usual). But
        beginning yesterday, networking has been (I know it sounds like a
        contradiction) consistently flaky. By this I mean:

                I reboot, it takes at least 10 seconds to receive an
                answer to a ping sent to a nearby workstation.

                I wait a certain amount of time, and then the delay gets
                longer, and longer, and then eventually I get no response
                at all. All sessions I had open on remote systems die.

        The time in between the reboot and when connectivity stops is getting
        shorter and shorter every time I reboot. It's down to less than a half
        hour of online time before I have to reboot again.

        I installed all the patches that came on the CDrom, as well as the
        recommended patches from sunsolve. And since this problem started,
        I've been applying various other patches that I didn't have applied to
        various things that *might* be related, but have proven not to be.

        [list of patches applied deleted]

The problem turned out to be a minor one, (and I feel kind of sheepish
about it :) I figured it out by running ping -s to another machine on
the network here, and every time a ping was sent, I received an
ICMP_PORT_UNREACHABLE error from a machine that seemed totally
unrelated. Turned out that that machine's IP address had been moved to
the primary nameserver position in /etc/resolv.conf, and every time a
lookup had to be done, the first nameserver had to time out, hence the
delay.

I still could not figure out, however, why things were deteriorating
over time. I applied patch 101907-05 at the suggestion of Casper Dik
and things have been ok so far.

There was still the question of why there had been a delay
when trying to telnet/rlogin etc into the machine. I realized that
since I was running tcp wrappers, they were causing a DNS lookup of
every host that connected, before login was started.

I want to thank the following people for their suggestions:

Casper Dik <casper@Holland.Sun.COM>
<joel@virtualvegas.com>
Perry Hutchison <perryh@pluto.rain.com>
Henry Unger <hunger@hitech.com>

Kira Attwood
kira@virginia.edu



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:10:29 CDT