Conference 7.286::fddi

1324.0. "DECnet, FDDI and Fibronix' Brouters" by GIDDAY::STANISLAUS () Sat Apr 23 1994 18:41

    
	Has anyone had any experience on working with Fibronix FDDI Bridges 
and Routers Model 8xxx for routing DECnet. These Brouters have FDDI A and B
ports and a maximum of 6 Ethernet Ports. A customer has an FDDI ring in a 
campus with 19 of the above 8xxx series products out of which 7 can operate as 
MAC layer bridges only and 12 can operate as bridges-cum-multiprotocol 
routers.

	The customer has a LAVc cluster with 2 Large VAX systems - VAX 4 and 
VAX5 and about 30 VAXstation 4000s and about 1 or 2 Alpha Workstations.
They are all interconnected via the above FDDI bridges and brouters.
There is one Cisco 3000 router on one of the Ethernets routing DECnet and IP
and bridging other protocols. The Cisco 3000'd DECnet address is 1.501.
VAX 4 is 1.6 and VAX 5 is 1.15. These two nodes are routing IV level 1 
routers. Cisco 3000 on the Ethernet is also routing IV level 1.


Config 1: 7 bridges and 12 routers by Fibronix were all configured as bridges 
only. When this was done everything worked fine - LAVc, IP, IPX, DECnet etc. 
between stations. In this config 1.501 (Cisco) is a DECnet L1 router.

Config 2: 7 bridges and 10 of the 12 Brouters configured as Multiprotocol 
Brouters routing DECnet and IP and the remaining 2 brouters behaving as 
bridges only. These Brouters addresses are 1.601 to 1.610 for the 10 of them.
They are all level 1 routers.

In this configuration LAVc and IP still work fine for providing end-to-end 
connectivity from one Ethernet via FDDI ring to a remote Ethernet. But DECnet 
does not provide a RELIABLE connectivity between one Ethernet and a remote 
Ethernet via FDDI ring. A set host that works between a Workstation called 
VAX42 and VAX4 one minute will fail to set host a few minutes later.

Symptom 1 in Config 2:

	The designated router as seen by routing VAX systems 4 and 5
will be 1.610 (Fibronix) to begin with. Just around 10 or 12 minutes later the 
designated router will be 1.609 and suddenly it will be 1.608 and so on 
changing erratically quite often and each time the designated router changes, 
the adjacent node changes as seen by VAX 4 and VAX 5.

Symptom 2 in Config 2:

	The VAXstaion 4000 end nodes (all part of a large MI cluster) do not 
see any designated router at all.

Symptom 3 in Config 2:

	While we had problems with DECnet (set host) there was no problem with 
LAVc or IP. Because LAVc was running fine I do not doubt the physical or data 
link layer of the Fibronix Bridges and Brouters or Ethernet cabling or FDDI 
cabling. I think the DECnet implementation within the Fibronix Brouters may
be causing the above symptom.

	I traced DECnet packets from a workstation (VAX42) to VAX 5 (the cluster
boot node) while setting host from VAX 42 to VAX 5. VAX42 sent the 'Connect 
Initiate' to VAX 5's MAC layer address directly (since it did not see any
adjacent or designated or any DECnet router.) The 'Connect Ack' from VAX 5
went to the designated router for VAX 5 (say 1.609 - fibronix), and since 
1.609 does not know about 1.42 (VAX 42), it did not forward the 'Connect Ack'
to VAX 42 and hence the remote node VAX 5 is unreachable for VAX 42. Hence VAX 
42 does a 'Retransmit Connect Initiate' and keeps doing the same thing over 
and over again till it gives up with the 'remote node is unreachable' message.

Anyway the workaround is to turn off all DECnet routing on all Fibronix 
brouters. Then the Cisco 3000 (1.501) is seen as the designated router for VAX 
4, VAX 5, and all VAXstations and everything work fine even for DECnet when 
the setup is just a bridged set up.

Alphonse

    

    How R ya mate
    New Jersey Transit has a very similar config and is seeing starnge
    problems. Their local support Eng. would be able to shed more light
    on what he has discovered. Contact me and I will provide more details.
    							Al E