Gigabit ethernet ports disable and never comes back up- loop in the network

[Scott Fella]

I had a loop in a network that caused all my uplink ports on my 6509, 4006, 3550-12G, 3560-24 and 2940s to disable. I thought that spanning tree was reconfiguring, but after a few minutes I knew the ports were not going to come back. I saw the error on the 4006 stating port was flapping... then the whole network went down. When the ports shut down or disables itself the only way I was able to get a link was to reboot all the switches excep for the 6509 and 4006. Can someone explain why the ports disable itself and is there a way I can prevent this from happening. My client has hub connected to hubs and other switches which we can not touch. I'm afraid someonee is going to plug a hub into the network and create another loop. All I want to know is, the only way to get the ports back up is to reboot?

Thanks

[glen.grant]

You should be able to just renenable the ports if they went err-disable . If you were using udld on the uplinks it's possible the ports shut themselves down if you had a loop because the uplinks missed hello packets due to the loop. What state are the ports in when you look at them using the "show port" or show interface status" commands? If they are err-disabling , yes you can turn off udld on a per port basis I believe.

[Scott Fella]

They did go err-disable, but I was not able to get in each switch since the uplink ports were down. Had to console in..... Should the port stay down?

All that was configured on the trunk ports was trunking and dot1q trunking. The problem is that they have so many hubs (this is a high school with 8 IDFs that support around 15 - 20 other switch enclosures. The hubs we have to leave in place (teachers work area and labs) and make sure we keep those on the network per the client. Cables are ran in the ceiling from one room to another so it looks liek a regular drop that other PC's are using. All 10/100 ports are configured with portfast and this might be the biggest problem. But not know what drops go to where maked it hard to not enable portfast on all ports.

UDLD is disabled by default so what can I do....

[amit-singh]

Hi,

You can use auto error disable recovery feature.

errdisable recovery {cause {all | arp-inspection | bpduguard | channel-misconfig | dhcp-rate-limit | dtp-flap | gbic-invalid | l2ptguard | link-flap | loopback | pagp-flap | psecure-violation | security-violation | udld | vmps} | {interval interval}

http://www.cisco.com/univercd/cc/td/doc/product/lan/cat3750/12225sec/3750cr/cli1.htm#wp2233388

regards,

-amit singh

[Scott Fella]

Thanks, I will have to try that.

[Scott Fella]

On the 2940 switches, the interface showed up/up. But I did not have connectivity to the network. I issued the shut and no shut on the interface and that did not help. Once I rebooted the switch, connectivity was back. This was on about 180 2940's.

[sivakumar.ks]

Is the connectivity holding or once again getting dropped.Use packet snipper tool and mirror that port and check why it is dropping out. For me it looks like 5-4-3-2-1 rule issue.But I am not sure as you said you aware of that rule.

or give show interface command for that GB interface and check for crc error.

[sivakumar.ks]

Hi, are you following 5-4-3-2-1 rules in your network

http://compnetworking.about.com/od/networkdesign/l/blfaq018.htm

[Scott Fella]

I understand that rule, the issue I'm having is that they have old infrastructure that we have no control over. We basically are basically integrating equipment in the MDF, IDF and equipment enclosures that are located in each room. The old infrastructure has links from one hub in another room to other hubs in other rooms which are connected to the equipment enclosures. The data cable in these enclosues and the long patch cords that my client uses are the same color so its hard to detect if it's actually used for a PC or printer or if its an uplink from a hub.

Thanks for the input.