Date: Thu, 7 Mar 1996 13:44:51 -0500 From: "Garrett A. Wollman" <wollman@lcs.mit.edu> To: questions@freebsd.org Cc: Robert Du Gaue <rdugaue@calweb.com> Subject: [Robert Du Gaue: routing] Message-ID: <9603071844.AA06801@halloran-eldar.lcs.mit.edu>
next in thread | raw e-mail | index | archive | help
Forwarding this to questions in case I don't have an opportunity to figure it out... Has anybody else seen this problem, and what was the solution? (Is this another instance of the `routed' bug? I can't tell since this person doesn't say what version he is running.) -GAWollman ------- start of forwarded message (RFC 934 encapsulation) ------- Message-Id: <Pine.BSF.3.91.960306080902.11887A-100000@www.calweb.com> Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Resent-To: wollman@freebsd.org Resent-Date: Wed, 06 Mar 1996 14:46:02 -0800 Resent-Message-Id: <16611.826152362@time.cdrom.com> Resent-From: "Jordan K. Hubbard" <jkh@time.cdrom.com> From: Robert Du Gaue <rdugaue@calweb.com> To: "Jordan K. Hubbard" <jkh@calweb.com> Subject: routing Date: Wed, 6 Mar 1996 08:16:21 -0800 (PST) Could you forward to your routing gurus. We have had a terrible couple of days. Our machine keep 'learning' bogus routes and we can't trace just where it's coming from. I'm trying now to give the system some fixed routes to overcome some problems, however the machines don't seem to also be learning arp entries, at least not all of them. Some machines are missing entries for various interfaces. I'm more worried about the portmasters right now, because if I do a net route to a portmaster that's not in the arp table, the route takes as a host instead of a net. I've made a file of fixed_arps to run with 'arp -f filename' command, but all entries for the portmasters that the system doesn't know about fail anyways. Here's what I'm seeing: bash# arp -a lan1.calweb.com (165.90.138.1) at 8:0:87:14:ac:2e calweb.calweb.com (165.90.138.3) at 8:0:69:8:8c:24 sun1.calweb.com (165.90.138.6) at 8:0:20:8:4e:b0 web1.calweb.com (165.90.138.10) at 0:0:c0:1c:f4:c6 web2.calweb.com (165.90.138.11) at 0:0:c0:18:6c:e web3.calweb.com (165.90.138.12) at 0:0:c0:de:8b:e web4.calweb.com (165.90.138.15) at 0:0:c0:63:c:9f mail.calweb.com (165.90.138.20) at 0:0:c0:1d:f4:c6 sac1.calweb.com (165.90.138.26) at 0:c0:5:1:d:25 sac2.calweb.com (165.90.138.27) at 0:c0:5:1:1e:48 ded1.calweb.com (165.90.138.28) at 0:c0:5:1:2e:f6 infosite.com (165.90.138.203) at 0:80:ad:14:b9:15 ? (165.90.138.255) at (incomplete) bash# cd /common bash# arp -f fixed_arps writing to routing socket: File exists writing to routing socket: File exists writing to routing socket: File exists cannot intuit interface index and type for sac3 cannot intuit interface index and type for sac4 cannot intuit interface index and type for sac5 cannot intuit interface index and type for sac6 bash# cat fixed_arps sac1 0:c0:05:01:0d:25 sac2 0:c0:05:01:1e:48 ded1 0:c0:05:01:2e:f6 sac3 0:c0:05:01:2a:d2 sac4 0:c0:05:01:36:0e sac5 0:c0:05:01:36:ea sac6 0:c0:05:01:42:bf bash# I've tried putting 'temp' and 'pub' entries on sac3-6, but still get the same thing. Very wierd. The machines at some point learn routes of like: 165.90 and 165.90.138 Usually ether to the #link address or to a portmaster. Though I've test various different things and can't find anything that would cause the PM to broadcast routes like that. If everything was dynamic I'd just to RIP off on the PMs, but we do have some fixed IP people that can come in under any PM that I think would case problems. ------- end -------
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9603071844.AA06801>