LWONG@UVPHYS.BITNET (03/01/90)
We have a domain of 8 apollo workstations (dn3xxx, dn4xxx, dn3xx). Most are
on SR10.2 but we have 2 nodes still on SR9.7. Our DN4000 has an ethernet
controller card and acts as a gateway to the campus ethernet. The primary
network is Apollo Token Ring.
We recently installed TCP/IP on the gateway node. It starts daemons tcpd,
routed, inetd and enables telnet and ftp services. It works fine. The
problem is that I can't get the other (non-gateway) nodes in the network to
telnet to computers outside the primary domain ring. It can successfully
telnet to the gateway node, though.
To install TCP/IP on the non-gateway node (which is also my admin node, BTW,
so I know there isn't a problem with links), I start daemons llbd, tcpd,
routed (-f -q options). No inetd daemon since I don't want to enable it
to ACCEPT connections.
The interface on the gateway node for DR0 and ETH1 (I misunderstood and
configured the jumper on the ethernet controller card to be eth1 instead
of eth0) are as follows:
dr0: flags=43<UP,BROADCAST,RUNNING>
inet 128.121.72.1 netmask ff000000 broadcast 126.0.0.0
eth1: flags=43<UP,BROADCAST,RUNNING>
inet 128.189.67.2 netmask ffff0000 broadcast 128.189.0.0
macaddr 8:0:1e:1:63:91
The interface on the tcp/ip host is:
dr0: flags=43<UP,BROADCAST,RUNNING>
inet 126.122.72.4 netmask ff000000 broadcast 126.0.0.0
The tcp/ip host learns its own address from the host table at startup.
I read something awhile back in this newsgroup about bugs in the router
or in TCP/IP ... but I've lost the messages. Am I missing something???
Has anyone else done this? Is the mixed SR9 SR10 network giving me
problems? Response from the hostline on this has been poor to say the
least.
Any help would be appreciated.
Lynda Wong.dr@stl.stc.co.uk (David Rutter) (03/06/90)
>The problem is that I can't get the other (non-gateway) nodes in the network >to telnet to computers outside the primary domain ring. It can successfully >telnet to the gateway node, though. We have the same problem. We are running sr 10.1. The Apollo helpline suggested installing patch 28 (ethernet-microcode) but this didnt help. By looking at the USE field obtained from the command /usr/ucb/netstat -r, we found that when non gateway Apollo nodes attempt to connect to a Unix machine on the ethernet, the tcp/ip packets are forwarded by the gateway and received by the Unix machine. However, the reply from the Unix machine does not get forwarded through the gateway. (We also found that the USE field obtained from /usr/ucb/netstat -r does not always work - the USE count sometimes remains at zero even when we know packets are being forwarded through the gateway) By running the routing demon in the foreground on a Unix machine, we found that the routing demon on the Apollo gateway successfully forwards details of our Apollo token ring to other Unix machines on the ethernet. However it does this only once. It does not rebroadcast this information every few seconds and so the routing information soon gets deleted from the Unix machines routing tables. We tried to manually add a route to the Unix machines routing tables using /etc/route. When this was tried on a sun running routed, it worked for a few days and then stopped. /etc/route was successfully used on a Unix HP machine which wasnt running routed. Even if /etc/route did work, because of the size of our network, this is not an acceptable solution. Apollo has as yet not come up with an answer. A few weeks ago someone posted to the net that they had installed a patch that solved a similar problem. I dont think they mentioned the patch number. I would be grateful if they could reply to this message. Was it the patch 28 that we installed or another one that our Apollo branch doesnt know about? David Rutter dr@stl.stc.co.uk