Connection between Dell Force10 and Cisco Catalyst showing abnormal behavior

[Rafael Santos]

Hello,

I have 2 Dell Force10 MXL10 / 40 switches that form a VLT domain. From each of these 2 Dell Force 10 switches, one port-channel configured in vlt peer-lag mode comes out and these two ports arrive on a Catalyst 4500 forming a single Port-channel on Catalyst.

I created 1 more port-channel on each of the Dell Force10, to arrive with a single port-channel, now on a Catalyst 3750. When activating the port-channel between Force10 and Catalyst 3750 the port-channel is UP, but the Port-channel interfaces between the Dell Force10 and Catalyst 4500 are in "err-disable" status on the Catalyst 4500 side.

When I disable Port-channel between Force10 and Catalyst 3750, and apply the shutdown and no shutdown command on the Catalyst 4500 interfaces that connect to Force10, the Port-channel between Force10 and Catalyst 4500 is up again.

The catalyst 3750 I added to the scenario is completely isolated from the network so there are no loop conditions.

Attached current topology and the topology that generates the problem.

Thanks.

[Haydn Andrews]

this has been posted to the wireless forum, might be better suited in the Network Architecture forum

[Rafael Santos]

Hello Haydn,

Sorry for the mistake. I'll be more careful next time.

[patoberli]

Without investigating to much, can you please show a "show logging" from the 4500 when the issue happens? It should log why it's error-disabling the ports.

[Rafael Santos]

Hello Patoberli,

Following are the logs from Catalyst 4500 that put Force10-connected interfaces in err-disable when I enabled port-channel between Force10 and Catalyst 3750:

089645: Oct 8 22:11:52.812 BRA: %PM-4-ERR_DISABLE: udld error detected on Te2/10, putting Te2/10 in err-disable state
089646: Oct 8 22:11:52.815 BRA: %EC-5-UNBUNDLE: Interface Te2/10 left the port-channel Po6
089647: 000257: Oct 8 22:11:52.822 BRA: %EC-5-UNBUNDLE: STANDBY:Interface Te2/10 left the port-channel Po6
089648: 000258: Oct 8 22:11:52.911 BRA: %PM-4-ERR_DISABLE: STANDBY:udld error detected on Te2/10, putting Te2/10 in err-disable state
089649: Oct 8 22:11:55.913 BRA: %UDLD-4-UDLD_PORT_DISABLED: UDLD disabled interface Te2/9, neighbor mismatch detected
089650: Oct 8 22:11:55.913 BRA: %PM-4-ERR_DISABLE: udld error detected on Te2/9, putting Te2/9 in err-disable state
089651: Oct 8 22:11:55.915 BRA: %EC-5-UNBUNDLE: Interface Te2/9 left the port-channel Po6
089652: 000259: Oct 8 22:11:55.919 BRA: %EC-5-UNBUNDLE: STANDBY:Interface Te2/9 left the port-channel Po6
089653: 000260: Oct 8 22:11:56.016 BRA: %PM-4-ERR_DISABLE: STANDBY:udld error detected on Te2/9, putting Te2/9 in err-disable state

[patoberli]

Have a look at your drawing and the error message.

There seems to be a Po6, which isn't in your drawing, which is causing the error. Have a look at the Te2/9 and Te2/10 configuration, something is wrong there.
Or you might need to disable UDLD on the ports on the 4500, not sure if that is supported with two Dell Force 10 as directly attached devices, my gut feeling says no.

[Rafael Santos]

Hello Patoberli,

You are right. In the outline of the topology I submitted, the log Po6 corresponds to the topology Po10. Excuse me for the mistake in the drawing.
Your UDLD suspicion is fully valid, I searched some sites and really this feature is not supported by Dell Force 10.
The UDLD configuration is configured globally on Catalyst 4500. I will keep the global configuration and schedule a maintenance window to disable UDLD for Catalyst 4500 interfaces that connects with Dell Force 10.

The only point that is still unclear is this, because Catalyst 4500 interfaces (Te2 / 9 and Te2 / 10) that connect to the Dell Force 10 go err-disable only when Port-channel 15 and Catalyst are enabled. the Dell Force 10?
I reinforce that the Catalyst 3750 is completely isolated from the network, so there is no scenario for loop condition.

[patoberli · Accepted solution]

I assume that the Dell forwards and the Cat 3750 responds to the UDLD packets. Because the port numbers don't match, it will shut the port.

[Rafael Santos]

Hello Patoberli,

I am awaiting release for a new maintenance window and trying to apply the scenario again.

This time I will disable the UDL feature of Catalyst 4500 and 3750 interfaces that connect with Dell Force 10. I hope the err-disable status will stop with this action.

Once you apply the settings I'll be back here to give feedback if it worked or not.

Thank you very much.

[Rafael Santos]

Hello people,

Just to give feedback, I tested by turning off the UDLD feature on the Cisco Catalyst 4500 switches and also on the 3750.

When port-channel 15 was raised between the DELL switch and Catalyst 3750, Po 15 was UP and no err-disable status occurred on the physical interfaces of the Po that connects with Catalyst 4500, so the UDLD feature was what it was. triggering the problem that knocked Po out between Catalyst 4500 and DELL. But I came across a new problem, when the Po between 15 between the DELL switch and the Catalyst 3750 goes UP, some networks that pass this trunk get pinged while others work normally.

The scenario is very strange, as Catalyst 3750 is isolated from the network at first, so there is no loop scenario in this condition.

The condition only normalizes when I disable Po 15 between the DELL switch and Catalyst 3750 and also need to shutdown and shutdown command on Catalyst 4500 Po 10 that connects with the Dell Switch. Once this is done, the intermittences that occur on some networks normalize and no longer failures.

I checked the logs of all switches involved and no STP loop and convergence logs are logged.

[patoberli · Accepted solution]

Do they all use the same native VLAN?
There should be STP changes when a link goes down and up, did you look at the correct counters?
Which STP variant do you have enabled?

[Rafael Santos]

Dear,

I tested between Cisco and Dell Force 10 and the STP scenario worked perfectly.
The DELL switch preferred to go out of the box with the Catalyst 4500 which has 40G bandwidth and kept the alternate path that has 20G block bandwidth.
When I deactivated the interfaces between the Dell and Catalyst 4500, the block ports assumed forwarding status after the convergence time, and the scenario worked perfectly.

Thank you all for your support and attention.