Report loopback detection

[k4li]

Loopback deteciton (as far as i understand) should prevent Loops, that are not blocked by stp, eg on ports where stp is filtered or Devices (like my siemens phone), that somehow drops stp and thus creates a L2 loop.

From release Notes:

Code: Select all

LBD can detect and prevent L2 forwarding loops on port either in the absence of other loop-detection 
mechanisms like STP/RSTP/MSTP or when the mechanism can‟t detected it. Sometimes the 
STP/RSTP/MSTP based loop detection can‟t be used due to the following Facts 
- There is a client‟s equipment that drops or cuts the BPDUs.
- The STP protocol is restricted on edge Network
The LBD feature detects that a port has been looped back or looped. If a loop-back/loop is detected, 
the port is disabled (forced down) and the appropriate Error Log is issued.

So do you have a loop in your network, when you

a) I make a first loopback on the other swith (connected on the stack on the port 1/10 for example) : nothing
b) I break the loopback and make it again : loopback is detected, the port 1/10 is shutdown for 30 secondes : it works

In case a): do you have a l2 loop, e.g. network interface goes to 100% transmit, cpu high ...

Yes, I had only this when I do the first loopback.

or is this case just blocked by e.g. stp ?

If you have a loop (with resulting impacts on your network) i would open a SR (and please keep me updated). If this is just a "it's blocked somehow else" i would not care too much.

I just want the switch shutdown automatically the port where loopback is detected. So, it works on the second loopback, not on the first...

[devnull]

then i would open a SR and provide them with the necessary data.

[k4li]

then i would open a SR and provide them with the necessary data.

Thanks

[k4li]

Hello,

I update this post because now its working. I erase all my config and do and new conf step by step. I think I had a problem with the old config.

Thank you all for your support.

[m00n]

Hello
I need advice how to prevent loop in MC-lAG OS10k Environment ? From time to time someone connect dumb L2 switch to network and make loop on that dumb l2 switch. Tried everything, qos policies, loop-detection, bpdu guards and everything what i found here doesn't work. It is realyy disaster becouse when the loop occured MC-LAG bring down ALL aggregation and i need restart one core to restore connectivity (really no command like "no multi-chassis loop-detection 2/1" for example, why need i restart core to delete interface from disable mode ?) try to find a loop, remove the loop and restart second core to return MC-LAG synchronisation.

[devnull]

How did it kill all your network? The times i had a loop detected it was only one of the MC-LAGS shut down (if i create a link between e.g. MC-LAG LACP 96 to 97 switches) or a single lag (when someone somehow created a loop behind a single MC-LAG-LACP)

You can always use
clear multi-chassis loopback-detection, to clear a detected loop info, shutdown and reenable the ports, that got blocked
(int 1/17 adm sta dis, int 1/17 adm sta ena)
Afterwards it should "only" reblock the affected links if you did not fix the loop.

But i must admit i faced a similar problem with some real strange loop occurence, which i was not blocked by loopback detection - i had to use divide an conquer on a attached switch to identifiy the port in question. E.g. block 50% of ports, see if loop resurfaces - if yes unblock that 50% (you know it must be on the other half) and block half of the remaining ports..

If everything fails you can disable loopback detection on MC-LAG, but than you must life with the loop - which may be the better solution depending on the network..

[m00n]

Ok too much emotions was in me Loop dont kill all my network, only this aggregation when loop occured, but if it is stack when i have connect servers then i have a problems Thx for tip how to reenable port after loop detected. I used this commanf before but dont think about disable/enable port after that. Some time ago i disable loop detection on 10k but it was problematic with mac address leraning. IN my case computetrs with WIndows Vista and Windows 7 cant find default gateway (VIP address) i had to make static arp with mac address and ip of default gateway. Strange was that on linux amchines and Windows XP there was no problem.

[devnull]

Yeah i know that - in my case it killed a MLAG to a whole building, which was also not nice.
Never had your problem with mac-learning, has it been fixed in newer releases?

If you can test try with loopback-detection on edge switch (reduce transmit interval, use current code) - maybe you need to slow down loopbackdetection interval on the 6900?
e.g. 5 sec transmit-interval on edge, 10 secs on MC-LAG?

otherwise: if you can reproduce a loop that is not catched by loopback-detection: open a ticket to get that fixed..