hedrick@geneva.rutgers.edu (Charles Hedrick) (05/09/89)
We've been having a token ring problem that I had always thought would be impossible. Does anyone else have experience that might help diagnose it? The setup: a PC lab with about 25 PS/2's (from model 25's on up). All using IBM token ring adapters and IBM file server software. One cisco gateway used as IP gateway onto main campus network. The problem: Now on then -- typically at least once a day -- the network hangs. IP traffic stops. File service stops. The only way to make it start again is to power off some particular PC (not the same one each time). The cisco gateway -- which is the only thing on the net that know how to report errors on its console -- tells us that the ring is beaconing during the period it is down. What it looks like to us is that there is some funny state in the token ring adapter card that causes it not to pass the token on to the next unit, but also not to remove itself from the network. We thought the design of the TR adapter card was such that it would either (1) pass on packets correctly (2) remove itself from the network or (3) was broken. It doesn't look like broken hardware, because it's a different one each time. I don't think the system whose powering down fixes things is necessarily using the network at the time. All of our experience and monitoring equipment is for Ethernet. We're completely baffled by this problem.