narten@cs.purdue.EDU (Thomas Narten) (06/13/88)
In article <2512@mandrill.CWRU.Edu> edguer@mandrill (Aydin Edguer) writes: >I am trying to understand why certain sites are unreachable from my machine. >Background: [...recently joined NSFnet; prior connection via CSNET..] >Question(s): > Why would sites which were reachable become unreachable like this? > Is this due (perhaps) to the dismantling of the ARPAnet which I understand > is taking place? No. Both Purdue and decwrl are still reachable via the ARPANET. > What can I do to find out why they have become unreachable? The problem is broken routing tables, and I suspect NSFnet (or its regionals). We have been unable to reach net 128.146 (ohio-state) for a week, and the same is apparently the case for 129.22. Both of these nets are reachable from NSFnet sites, but not from us. Moreover, our butterfly core gateway has routes for these nets that point to 10.4.0.14 (psc.psc.edu). Since we can ping psc.psc.edu, I conclude that we are correctly advertising routes to the core. In other words, we (Purdue and ARPANET) are managing routes correctly. By implication, NSFnet is not correctly propagating all of the routes it learns from the ARPANET. (Note that for the most part, we can reach NSFnet sites; only a handful cause trouble.) > What can I do to try to reach them now that my router has given up? Contact your regional NOC (if there is one), and if that doesn't help, call the NSFnet NOC. The following information would be of help: 1) Do you have a route to get to Purdue (either explicit or default)? For that matter, do you have routes to other sites that are on the ARPANET? Can you ping 10.0.0.37 (purdue's arpanet connection)? What about 10.4.0.14? If both of us can ping a common gateway, but not each other....... 2) Do you get ICMP unreachable messages when attempting to reach the networks in question? The best check for this is a 4.3 system with a modified ping that prints the name of the gateway that generated the ICMP message. Alternatively, one can open TCP connection to the host in question. For instance, "ftp 128.10.2.1" will abort with a "network unreachable" error if ICMP unreachables are returned; "connection timed out" on the other hand, indicates that packets are going into a black hole. This works on 4.3 BSD (running the latest tcp/ip code) and Sun OS systems. -- Thomas Narten narten@cs.purdue.edu or {ucbvax,decvax,ihnp4}!purdue!narten
tom@TUT.CIS.OHIO-STATE.EDU (Tom Easterday) (06/13/88)
The problem that CWRU and Perdue is experiencing is due to two things.
As of 6/7/88 the link between PSN 29 here at Ohio State and PSN 21 at
Pittsburgh was deinstalled. We switched over to our NSFnet link to PSC
which is a string of Proteon gateways (one here, one in Cleveland, and
one in Pittsburgh). These gateways are running Proteon software rev
7.4 which has a known bug when using default routes. PSC is in the
process (so I've been told) of installing rev 7.4b on these gateways
which will hopefully clear things up a bit. I'm counting on it happening
early this week. When the Cleveland POP gateway is upgraded CWRU should
see an improvement. Maybe Gene Hastings at PSC would like to clarify
things if something I've said here is incorrect.
Tom Easterday
The Ohio State University
Instruction and Research Computer Center
(614)292-4843
Email: tom@tut.cis.ohio-state.edu
or tom@oscsuna.osc.ohio-state.edukarl@triceratops.cis.ohio-state.edu (Karl Kleinpaste) (06/13/88)
narten@cs.purdue.EDU writes: edguer@mandrill writes: >[...Case Western can't get out to the Rest of the World...] > What can I do to find out why they have become unreachable? The problem is broken routing tables, and I suspect NSFnet (or its regionals). We have been unable to reach net 128.146 (ohio-state) for a week, and ... By implication, NSFnet is not correctly propagating all of the routes it learns from the ARPANET. (Note that for the most part, we can reach NSFnet sites; only a handful cause trouble.) Case Western is behind the same gateway as Ohio State... The problem is out-of-date software at the gateways in Columbus, Cleveland, and Pittsburgh, from what the local folks managing the net connections have told me, resulting in routing glitches in large quantities and leaving us unable to find lots of places, e.g., I haven't seen utexas.edu in 8 days. The expectation/hope is that the matter will be cleared up within another day or two - an update from The Vendor is on the way. What a mess, --Karl