From owner-freebsd-hackers Tue Oct 8 15:31:26 1996 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id PAA07254 for hackers-outgoing; Tue, 8 Oct 1996 15:31:26 -0700 (PDT) Received: from csla.csl.sri.com (csla.csl.sri.com [192.12.33.2]) by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id PAA07224 for ; Tue, 8 Oct 1996 15:31:18 -0700 (PDT) Received: from impulse.csl.sri.com (impulse.csl.sri.com [130.107.15.11]) by csla.csl.sri.com (8.7.3/8.7.3) with ESMTP id PAA25379 for ; Tue, 8 Oct 1996 15:28:37 -0700 (PDT) Received: from impulse.csl.sri.com (localhost [127.0.0.1]) by impulse.csl.sri.com (8.7.3/8.7.3) with ESMTP id PAA05742 for ; Tue, 8 Oct 1996 15:28:03 -0700 (PDT) Message-Id: <199610082228.PAA05742@impulse.csl.sri.com> To: freebsd-hackers@freebsd.org Subject: NFS problems with dual-network-interface machines Date: Tue, 08 Oct 1996 15:28:02 -0700 From: Fred Gilham Sender: owner-hackers@freebsd.org X-Loop: FreeBSD.org Precedence: bulk Hello, I've got a site where I'm using a bunch of FreeBSD boxes as file servers and as routers to segment a class B network into class C networks. So each box has two interfaces. I've noticed problems where putting the addresses of both interfaces in DNS sometimes causes NFS clients to connect to the `wrong' address and NFS hangs. The tcpdump program shows that the client is getting icmp port unreachable messages. I.e. machine alpha might be a router/file server with two addresses: 130.107.4.200 130.107.15.200 and machine beta might be on a different subnet having address 130.107.17.234. (both alpha and beta are freebsd boxes.) Machine beta is connected through a router similar to alpha to the 130.107.4 network. When machine beta tries to mount one of machine alpha's file systems, and it gets 130.107.15.200 from DNS as the address of alpha, port unreachable errors occur. This seems wrong to me, since machine alpha should be able to route from one of its addresses to the other. I've verified that this is the problem by putting alpha's 130.107.4.200 address in the host table on machine beta. Once I do this, everything works smoothly. However, this means putting all the file servers in all the host tables of all the clients. We're trying to run minimal host tables on the clients and use DNS for host lookups. Am I doing something wrong? Is there a canonical way to deal with this that I don't know about? Or is this a bug? Thanks for any help, -Fred Gilham gilham@csl.sri.com