🚀 go-pugleaf

 
I have a FDDI ring with about 20 RS6000's and 4 cisco AGS routers on it.  If I "bounce" the ring the FDDI stack and the IP stack on the RS6000 will stop talking to each other.

"Bouncing the ring" means:
	-disconnecting and reconnecting an FDDI cable on any system on the ring.
	-rebooting one of the RS6000's (or several of them)
	-issuing the command "clear int fddi 0" on one of the routers.
	-or any combination of these.

Performing one of these actions and waiting for a period of at least five minutes does not cause any problem.  Performing several of these actions in a fairly short time will cause one or more of the RS6000's to exhibit the 'disconnect' systems.

"FDDI/IP disconnect" means the RS6000 with the problem:
	-can not communicate via IP over the FDDI interface meaning:
		NO response to ping queries sent to the problem host
		NO response to ping queries sent from the problem host
		NO LLC FDDI packets from the problem host recorded on a 
			lanalyzer.
	-the upstream and downstream neighbors of the problem host show normal 
		FDDI connections.  (thru, etc)
	-the command 'fddistate 0' on the problem host shows normal FDDI 				connections.
	-there is no change in the ARP table on the problem host or any other
		system connected to the network.
	-there is no change in the routing table on the problem host or any
		other system connected to the network.
	-a FDDI lanalyzer continues to record some packets from the 
		problem host.
	-After a period of about 20 minutes, without any operator actions, 
		normal IP traffic will resume.
	-If the problem host is rebooted, normal traffic will resume.

The cisco routers are now running OS 8.3(3.2) but they have run other newer and older versions of the OS.  The cisco systems never have any trouble maintaining communication over the ring.

Recording traffic when the problem shows that the ring enters a token claim state and then a beacon state.  The DTI lanalyzer buffer is filled almost immediatly having captured over 50K packets in about a second.  It is filled before the renegotiation process is completed so I can't provide much more information on what is really happening on the ring.
 
Has anyone else seen this problem? 

Does anyone have any suggestions on what additional information I should try to gather so I can make an intelligent problem report to IBM?

Please send Email
Thanks
jzwiebel@pgl-devsvr.den.mmc.com