Nortel 8600

[JohnJohn]

Our current config is an 8600 at the core out to 450 stacks in the closets. Since this was implemented we have started to see Netware users lose drive mappings during normal operation. They have no problem with IP sessions only IPX.

 

[Tremor]

Ensure if you just installed these switches that you have disabled spanning tree on the host ports. The delay of setting up spanning tree will cause problems with netware trying to find its servers. Cisco calls it "portfast" (which must be enabled to reduce spantree delay...)

 

[JohnJohn]

I tried that, Nortel calls it fast learn. No help.

 

[gunfighter]

I had the same problem. I even disabled spanning tree to no avail. I found that the autonegotion on the switch did correctly set the speed and duplex with certain NIC cards like compaq's netflex 3p. I went into the workstations and hard set the speed and duplex of the NIC and then checked the switch port settings. This should work for you

 

[JohnJohn]

Being that we have over 60K workstations we hard set the switch. I assume that would get the same results??? I have tested it both ways and no help. I think it is a Novell problem. We also see diffent mappings across WAN connection and even at the core out to Local VLANs. Seems the Novell client doesn't want to see servers across a routed link correctly. Novell has had a similure problem and are looking into ours. We currently have over 160K objects in our tree and from what I can see three diffent DS versions which I think the problem maybe. I'll let you know what they find. Thanks for the input.

 

[GLHEC]

We had similar problems, I think. If it is happening every 15 minutes from bootup you should look for the watchdog packet negotiation/keep alives. Just another lovely Novell "feature to keep only active users mapped to drives". Capture traffic on both sides of the 8600. make sure your server is sending them/ make sure your workstation is not recieving them. If this is the case, there is an option on your netware server to choose routing options: nlsp or ripXX something like that. I forget which we had to change to, but I think that did the trick. Of course this is assuming that the switch is routing IPX. Nortel could not help us, they said they build a test lab and could not replicate it. We build our own test server and after 2 weeks figured it out. I hope this helps and that it is not too late.

 

[Numptie]

I have built a network consisting of 3 core 8600 switches connected in a ring with a further 8 8600 switches dual homed to two of the core switches. All of the 8600s are connected using single port VLANs with routing interfaces and OSPF enabled as the routing protocol. Users connect to 450T or BPS switches which in turn connect to single port VLANs on an 8600 switch. All VLANs were left in STG1(as STP not required) and STP is disabled at the port level but enabled at the group level on every 8600. The problem is that if a 450T/BPS is accidently connected to an 8600 with STP enabled the whole network dies within seconds. Why do the STP BDPUs not get dropped at the 8600 port the 450T is connecting to? Is the solution to put all 8600 interconnecting ports into STG2? I have tried disabling STP at the group level to no effect.