Empty MAC address table

[Jenni McWright]

So this morning I took a call regarding sporatic workstations not being able to ping other devices within the same subnet. Upon trying to trace out one of the troublesome MAC address to a port, I discovered that the switch, while still operational, had a completely empty mac-address table. I looked at the log and there wasn't anything detrimental and no reloads. Port were showing connected, VLANs were active, and I was able to telnet into the box without incident. The switch is a 48 port 2950 (WS-C2950G-48-EI) and we have one to two handfuls of them within our network, but I've never seen this issue before. To get everyone functional again, I just replaced the switch, rather than try to troubleshoot it in the closet.

Any thoughts on what would cause this? I'd normally just open a case, but for some reason this particular switch didn't make it on our maintenance renewal list last year (which of course means no RMA either).

Thanks in advance

[IAN WHITMORE]

Weird. Can't says I've ever had that problem with the 2950-EI and we have over 50 still in our production network. And you say there was nothing in the logs like link flapping or ports in err-disable, bad trunk negotaition or something like that? And now it's not in your network you can't troubleshoot it (takes time and lots of angry users).

Can't you plug it in somewhere in your office and play around with it? Cos if it wasn't the switch causing the problem...it could come back. But if you can verify the switch is dead you should be safe.

Regards,

Ian

[kapathak]

Hello,

This could have been due to a Layer 2 loop or a corrupt CAM table. When a switch receives BPDUs with TCN bit set to 1, the switch will flush out MAC entries. A huge continuous burst of TCNs might cause this behavior. We have no way to verify what is the cause now. I have also seen this issue happen due to corrupt CAMs which are recovered either by a reboot and if they can't, by switch replacement.

Cheers!

Kapil

[Souvik Ghosh]

what is the version of IOS ? this could be a bug,

@ kapil- just one small correction , i think when the switch receives a TCN from the root bridge it does not flush the cam but reduces the aging time from default 300 secs to 15secs.