Network Outage 4006 L3 Switch

r.kate · ‎10-13-2004

Hi ,

We have two switches conneted to each other using a fiber channel on two giga bit ports on each switch.

One of the link starting flapping and brought the channel down .The IOS we are running is

12.1(8a)EW1 .and the SUP is WS-X4014 .

Any suggestions appreciated .

Thanks

Raj

Prashanth Krishnappa · ‎10-13-2004

It is possible that the issue could be due to a GBIC/Fiber or the Sups Uplink port. Few things I would do to troubleshoot

1)To change GBICs/Fiber first on affected link

2)To enable UDLD aggresive on all fiber links

3)Have the channel in desirable mode versus ON mode.

Here is our best practices document

http://www.cisco.com/warp/public/473/185.html

12.1(8a)EW1 which is FCS IOS is very old(released 25-FEB-2002). You might want to consider upgrading

to a stable release such as 12.2(18)EW1.

Kevin Dorrell · ‎10-13-2004

I don't have a solution to your problem, but I do have some thoughts on it, if they help.

Leaving aside the reason why the link is flapping, I guess your question was more about why the link-flap caused the channel to go down.

Normally when a link flaps, the Spanning Tree gives you some protection because the link does not go into forwarding until it has been up at least 30 seconds. But if the link is part of a channel, we are in trouble; the bridge port is still forwarding because one of the links is still up, but the flapping link can disrupt the channel itself.

I know that it is possible for a flapping link to cause the port to go into errdisable. In fact, that is really what you want it to do here - to go into errdisable and stay there until you can deal with it. I know that a link-flap can make a port go into errdisable, because there is a command to specify how the port is to recover from a link-flap errdisable.

What I couldn't find was what criteria the port uses to decide it is flapping sufficiently to make it go into errdisable, nor how to tune these criteria.. Alternatively, what you want is a feature that says that a port should be link-up for a certain minimum time before it is allowed to join a channel group. Again, I couldn't find one. Can anyone help?

Kevin Dorrell

Luxembourg

Prashanth Krishnappa · ‎10-13-2004

We have seen this with channel mode "on" in CAT OS. Desirable is the way to go. We fixed this in CAT OS 6.3(3).

http://www.cisco.com/cgi-bin/bugtool/onebug.pl?bugid=CSCdv01221

From release notes

EtherChannel enhancements

An EtherChannel is preserved even if it contains only one port. In software releases prior to 6.3(1), if you have a 2-port channel and one link is removed, the remaining link is removed and added back to spanning tree, which causes a loss of connection on the channel until the link is forwarding again.

http://www.cisco.com/univercd/cc/td/doc/product/lan/cat6000/relnotes/78_11235.htm#wp35723

I dont know from top of my head where the fix went into IOS. But irrespective of this, desirable mode is the way to go.