Cisco EtherChannel with media Converter !!Problem

[wizkidnono]

I have a little problem.

I have 2 Cisco 2960 switches which are connected through a etherchannel.

The physical connection is based on 2 sets of lan cable to fiber media converters.
Switch -> Lan cable -> media converter -> fiber -> media converter -> Lan cable -> switch.

This works fine until I try to "break" the connection.
I have tested this in a lab setup without the media converters, only patch cables and then it has (redundance x channels)

My problem is that when I plug out either path cable the etherchannel stops transmitting as the fiber media converter does not tell the other switch that the link is down.
I run a ping test and this stops working.
This means that i have to go to the other switch and unplug the patch there as well before the channel is restored on the next port in the etherchannel.

Is there a way of telling the cisco switch to "check" for breakage other than just the "Physical Patch link status" ?

I have tried setting the media converter to "Link Loss" but this apperently dows not work to well.
Each time i have a link failure with link loss enabled i have to go and push a button on the media converter to restore the line.
The link is broken on the other cisco so it fails over but I can not restore it as I do not have physical access to the rooms where these media converters are stored.


I belive that when i pull the cable on either port on one side, the switch goes into link fail but on the other switch it has not senced that the link is down so it keeps transmitting packets on that interface.

I guess that load ballancing would make it possible for another computer to run fine as one of the links are still up,
and I am probably load ballanced on the "old" link but I would like mine to fail over as well like in the lab setup.

Would it help to not set the etherchannel mode to on ??
Or any other config ??


This is my cisco config :

show etherchannel detail
Channel-group listing:
----------------------

Group: 5

-----------------------------------------------------------

Fa0/23 connected trunk full 100 10/100BaseTX
Fa0/24 connected trunk full 100 10/100BaseTX
Gi0/1 notconnect 1 auto auto 10/100/1000BaseTX
Gi0/2 notconnect 1 auto auto 10/100/1000BaseTX
Po5 connected trunk a-full a-100

------------------------------------------------------------



show interfaces etherchannel
----
FastEthernet0/23:
Port state = Up Mstr In-Bndl
Channel group = 5 Mode = On/FEC Gcchange = -
Port-channel = Po5 GC = - Pseudo port-channel = Po5
Port index = 0 Load = 0x00 Protocol = -

Age of the port in the current state: 00d:04h:47m:38s

----
FastEthernet0/24:
Port state = Up Mstr In-Bndl
Channel group = 5 Mode = On/FEC Gcchange = -
Port-channel = Po5 GC = - Pseudo port-channel = Po5
Port index = 0 Load = 0x00 Protocol = -

Age of the port in the current state: 00d:04h:48m:30s

----
Port-channel5:
Age of the Port-channel = 00d:07h:11m:10s
Logical slot/port = 2/5 Number of ports = 2
GC = 0x00000000 HotStandBy port = null


----------
Group state = L2
Ports: 2 Maxports = 8
Port-channels: 1 Max Port-channels = 1
Protocol: -
Ports in the group:
-------------------
Port: Fa0/23
------------

Port state = Up Mstr In-Bndl
Channel group = 5 Mode = On/FEC Gcchange = -
Port-channel = Po5 GC = - Pseudo port-channel = Po5
Port index = 0 Load = 0x00 Protocol = -

Age of the port in the current state: 00d:04h:34m:34s

Port: Fa0/24


Thanks.

 

[ADB100]

This is down to the media converters. I have no idea if this is possible but what you want is the media converters to drop the fibre link if the copper link is down. It may be possible to workaround this with UDLD since this sends and receives hellos at layer-2. Try enabling udld on the physical ports in the channel:

interface FastEthernet0/23
 udld port aggressive

I have never tried this so I am not sure whether it will work correctly or not. Why don't you simply use 100BaseFX SPF modules in the 2960's?

Andy

 

[wizkidnono]

I will test this.

The 2960 i have do not have a fiber port.
Also the fibers are connected in a room which I do not have access to, and i am not allowed to store "user" equipment there aka. switches.
So another option would be to buy a switch with fiber ports and drag fiber to the place the switch is.
But this seems like an expensive overkill.
Then it would be better to figure out if it is my fiber converters that do not work in Link loss the way they should and replace with another brand.
Should link loss "reactivate" when i reesablish a link or is this a manual task after a breakage ?

Wizkid

 

[zinkann]

UDLD is very useful when it comes to fiber. too bad you don't have sfp ports. Try running UDLD PORT AGGRESIVE. The reason the other switch doesn't know is because it still has a link with the MC. UDLD will know that the link is down because it communicates with its neighbors via hello packets. trust me, we use this with STP.

 

[zinkann]

http://www.cisco.com/en/US/docs/swi...se/12.1_14_ax/configuration/guide/swudld.html

 

[wizkidnono]

Thanks.

Does udld require me to manualy remove the port from "bad" linkstate after an outage has been discovered.
Or will it reenable the port when the problem is fixed ?
From the cisco doc it seems that this is a manuel task.
At least this makes the etherchannel setup still work even if the MC is behaving badly and not dropping link.

Many many thanks

Wizkidno

 

[zinkann]

You have to manually enable it or use the command ,udld reset, which will reset all ports affected by UDLD. Depending on your IOS you can use the command, errdisable recovery cause udld, this enables a timer that will take the port out of err disable. Hope that helps.

 

[wizkidnono]

Thank you for all the help.
This works

I just have to be carfull when re enabeling the port on the etherchannel.

Great support

Wizkidno