VPC: epmLearn errdisable on one switch

[Nicholas Zerwig]

We have had an HP 7000 Chassis with two vConnect modules configured for LACP.  There are 4 connections for each vConnect module, two to each switch.  This has worked great up until a few weeks ago.  We migrated a vlan gateway into the fabric that a few of the blades were on and half of the vpc goes down on the chassis.  For VC-A the ports to switch 2 are down eth1/1 and eth1/2 and for VC2 the ports to switch 1 are down eth1/3 and eth1/4. 

We have worked with TAC and corrected some loadbalancing methods on the blade servers,  however, the epmLearn fault is still there.   Furthermore,  we have physically unplugged all four of the errdisable connections and the errdisable is still showing up on the vpc port channel.

"Port is down, reason:epmLearn errDisable(connected), used by:EPG"

Where else should we be looking for this fault if the ports that were effected aren't even connected any more?

Note:  ACI version 2.2 (3J)  

[aleccham]

Hello Nick,

What is the server side NIC teaming configurations?

[Nicholas Zerwig]

We have been instructed to have the physical blades set up with Active Standby  connections to the vConnect modules with Address hash as the LB method.  

The strange thing is that the ports have been physically uncabled and the fault still persist. 

[aleccham]

I think that should be good.

Reading over your attempts to remediate this issue, I can only see that you unplugged and replugged the affected ports. Did you attempt to bring the ports up and down. If not, please give that a try as well.

One other thing you could try, would be to disable and re-enable the EP loop protection policy. This would cause ACI to stop disabling ports if there was EP movement (this is what is causing your ports to become disabled now). Once re-enabled and the ports back up, we can check to see if the ports become disabled again.

[Nicholas Zerwig]

We have attempted to disable and re-enable the ports and the port-channel.  It errdisabled immediately when attempting that. Again,  The ports are STILL physically disconnected while doing this and the error remains. We would like to get the fault cleared before we plug the connections back in.

I can try bouncing the loop prevention policy.

[aleccham]

Hmmm, there is really no need to disconnect the ports in this scenario. A simple up/down should have resolved. I don't believe I have even ran into a customer that disconnected when trying to clear this fault out, so I cannot speak to the behavior when its configured in this manner.

[Nicholas Zerwig]

ok,  We will continue working with TAC to identify the root cause of the epmlearn fault.  I was just hoping someone has ran into this before and knows where to look.

[obi_nwamarah]

Solution I found was to disable and re-enable the VPC interface not the individual ports

[hethakur]

Leaf101# show int po7
port-channel7 is down (epm-learn-err-disable)

disable and enable port-channel 7 on leaf, solved the issue. 

Leaf101# show interface port-channel 7 status
----------------------------------------------------------------------------------------------
Port Name Status Vlan Duplex Speed Type
----------------------------------------------------------------------------------------------
Po7  LB connected trunk full 10G --