Switch shows itself on cdp

[ncfb-awarsame]

hello,

I have 3 348G switches connected in eiasy chain fashion on gig0/0 and gig0/1 ports. My problem is that when I do "show cdp neighbor" command, it show switch seeing itself on some of fastethernet ports.  fastethernet ports are all connected PCs. the closet is wired with old type 1 cable for token ring, but later conneverted into ethernet.

Please if any has anybody have a clue of this, let me know. 

thank you.

WITCHNGS2   Fas 0/24           173          S I      WS-C2950G-Fas 0/24
WITCHNGS2   Fas 0/15           173          S I      WS-C2950G-Fas 0/15
SWITCHNGS2   Fas 0/9            173          S I      WS-C2950G-Fas 0/9
WITCHNGS2   Fas 0/8            173          S I      WS-C2950G-Fas 0/8

[Peter Paluch]

Hello,

I would say that the most reasonable explanation is that the CDP frames are somehow "reflected" back. Are you absolutely sure that on these ports (Fa0/24, Fa0/15, Fa0/8), only end hosts are connected? Is it a direct connection by cable only, or is there any device between the port and the end host, no matter how insignificant it may seem?

Best regards,

Peter

[ncfb-awarsame]

thanks Peter,

The only thing that is in between switch ports and end hosts  is type 1 batch panel.      Type 1 cables run from cubes to the patch panel in closet, then token ring to Ethernet convertor adapters and cat 5 cables are used between switch ports to patch panel.

I'm suspecting the type 1 patch panel, but then it is not happening on ports.

thank you.

[royalblues]

You are correct

The problem you see is due to the type 1 cabling... we have seen this happen with patch panels that convert type1 to ethernet.

when the workstation is disconnected the tran/receive pairs are looped causing the switch to see a loop.

The type 1 cables were designed for token ring so when the cable is disconnected at the host end, it forms a loop restoring the ring

HTH

Narayan

[Peter Paluch]

Narayan,

That would make sense. I have no practical experience with Token Ring equipment - thank you for joining in!

Nevertheless, something here confuses me: On Catalyst switches, every switchport emits LOOP frames until explicitly disabled. If a port receives a LOOP frame it has originated itself, it will be immediately put into self-looped errdisabled state. Why did not that happen in this network?

Best regards,

Peter

[glen.grant]

   We have seen this when they use gigastack gbics and you tie them all together . Ugly setup and tough to troubleshoot and find anything including tracking down addresses .  I bet if you look closely they are using this setup .

[ncfb-awarsame]

Thank you all,

some of the switches use gigastack gbics, but this issue is also happening on those switches not using gigastack  gbics.  Narayan  comments makes more sense. If I look those ports closely, I don't find  mac addresses associated with them, yet they show line up protocol up. I'm also seeing BPDU Guard triggered on end host ports randamly and I did not find any device capable of sending BPDU, so most likely the type 1 cables and patch panel are causing loops.

.

ahmed