I learned a lot about routing when I posted my original question, and my
Solaris systems will be better for it, but nothing solved my problem.
It did get solved last night, however. The network people discovered that my
Solaris 2.5.1 system, a SparcServer 20 on 10baseT, was not responding to
"arp" requests. Any system on the same segment which already had the system
in its arp table could talk to it, but nobody else could. I checked
(repeatedly) the netmask and broadcast mask, and all was in order.
The system is connected directly to a switch, so the network folks decided to
connect it to a hub so they could watch the traffic with a sniffer. When
they unplugged the cable and reconnected it to the hub, the system started
working. They reconnected it to the switch, and it STILL worked. All has
been well since. I would expect such behavior if the netmask was wrong, but
I've never heard of a situation where an Ethernet connection would function
normally except that it wouldn't respond to arps.
Original question and summary are included below.
Thanks again to everyone who replied.
From: Rodney Wines <Rodney.Wines@ahqps.alcatel.fr>
To: Sun Managers List <email@example.com>
Date: Monday, February 15, 1999 5:58 PM
Subject: SUMMARY: Router Question
>I'm still having problems with my Solaris 2.5.1 system, but I certainly got
>my questions answered about routing. Thanks to everyone who replied. It
>a LONG list.
>I think that the heart of the problem is that there are problems with our
>Xylan switch. My 2.5.1 system worked for about 5 minutes earlier today, but
>has stopped working again. A netstat -rn shows that it has the correct
>default router defined, and it has the same default router as the system
>As for my original question about how routes get defined (assuming I am
>summarizing correctly), at boot time, " ... /etc/defaultrouter is ment to do
>exactly that: adds default route to the routing table while going through
>boot sequence (it is the one marked with "G" for gateway in the netstat -rn
>command output). This default route will enable the box to talk to any IP
>nodes which are not on the local subnet. This file is consulted in the
>"single user mode" part of boot sequence. Go to /etc/rcS.d and execute 'grep
>defaultrouter *'. This will enlighten you how it is being used." The quote
>was from Greg Sawicki.
>And, according to Raymond Wong, "Both 2.5.1 and 2.6 go through pretty much
>the same process for determining route. They check for /etc/defaultrouter,
>or check for some existing default route setting (i.e. a manual route
>has already been run in the script) If they don't find those, they try
>running router discovery (in.rdisc), and if that doesn't work they run
>(in.routed) to listen for continuuing RIP."
>I think the network people are tinkering with our switch even as I write
>this. I'll keep my fingers crossed ...
>From: Rodney Wines <Rodney.Wines@ahqps.alcatel.fr>
>To: Sun Managers List <firstname.lastname@example.org>
>Date: Sunday, February 14, 1999 11:18 PM
>Subject: Router Question
>>I've got two systems, one running Solaris 2.5.1, and the other Solaris 2.6.
>>The systems are sitting side by side on the same network segment. Both
>>systems USED to work fine. However, we have had some network problems
>>recently. The 2.6 system is working fine, and a "netstat -r" shows it
>>the proper router. The 2.5.1 system cannot talk to any system that isn't
>>its network segment. An ifconfig -a shows that both systems have the same
>>broadcast and subnet masks.
>>I tried a sys-unconfig on the misbehaving 2.5.1 system, but I wasn't
>>for a router when the system booted. I created a /etc/defaultrouter file,
>>and added the correct router, and the system works sort of. However, the
>>system doesn't have this file, and I never needed it before.
>>How is my router being defined, and does anybody have any suggestions for
>>places I can look to discover why I suddenly need a /etc/defaultrouter
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:13:15 CDT