Has Anybody seen this VPC error on the 5k?

[justin.allen]

Ok, my VPC is up and working for all my devices except my connection to the uplink switch. It is a sinle 3750 with port channles consisting of 4 interfaces. It is connected to two 5010s. I show the following error when I show VPC

"vpc port channel mis-config due to vpc links in the 2 switches connected to differnt partners"

Google does not have much on this error and this is the first time I have seen this. Any idea on what might casue this?

Thanks for any help.

[allan.thomas]

Hi Justin,

I have not specifically come across this error before, but it seems to suggest that both your vpc peer devices do not recognise the 3750 as a single switch, but two separate entities, thus perceive the links as separate.  Is the 3750 a stack of switches where you have distributed the etherchannel uplinks across seperate switches in the stack, commonly referred to as cross-stack Etherchannel?

Ensure that all four uplinks on the 3750s are configured within the same port-channel group using LACP (mode active), similarly that the N5K uplinks are configured or associated with the same port-channel and vpc identifier.  Could you post the configuration of your uplink from the 3750(s), and N5K uplinks to the 3750 and peer-link and keep-alive link.

If you do a show interface trunk on the N5K, do any vlans show not forwarding and that are err-disabled, this suggest that the incorrect VLANs are not consistent across the uplinks.

Hope this helps.

Allan.

[kiliasov]

I had similat error with N7K.

Every time I reload stack master it would shutdown the vPC to 3750 stack.

It's the way 3750 handles MAC address for master switch.There's no "virtual MAC" for the whole stack. It's always the MAC of stack master used for identification. When it goes down, new master is elected and it's MAC is used.

So from the vPC point of view all of a sudden you have 2 LACP id's for one port-channel. vPC thinks it's connected to 2 diferent devices, suspends vPC link.

To avoid this try:

stack-mac persistent timer 0

[Network Technologies Team · Accepted solution]

Essentially to fix this, remove the vpc from the port-channel on both cores and re-add it back in. if you do a show lacp interface for that port you can see the details of the MAC addresses it's using for the 3750 stack. We had this problem on a 3750 stack and going to n7k

[Darkmatter]

I too had this issue with a back to back vPC between N7K and N3K.

show vpc 55

The vPC on N7K showed being up and success passed consistency checks.

The vPC on N3K side consistency checks failed and reason was "vpc port channel mis-config due to vpc links in the 2 switches connected to different partners"

Carefully investigating the config of every port involved in the vPC on both Nexus switches N7K and N3K did not show any configuration error, as the config was correct.

The hint came when i ran the command "show lacp interface ethernet x/y" on the N3K and N7K.

On N7K, the Neighbor stats for LACP for one of the interfaces downstream, to one of the N3K, had basically no info and neighbor MAC was all 00-00-00-00-00-00 (checking this for every interface in the port-channel going downstream)

This way i could narrow it down to one specific switch that posed to be the problem.

Seemed like it was one of the N3K switches that did not properly sent out LACP negotiations to the upstream N7K.

Looks like you can trigger this behaviour if you reconfigure interfaces that were already part of a vPC, but now decided to use these interfaces for another vPC.

Bouncing the port-channel on this switch (N3K-B)did the trick! (so shut -> no shut)

From that point on, port-channel and vPC came up directly on this switch and the issue was resolved.

I'm aware this is a very old post, still i hope people who will find it will be supported in their troubleshooting when they come across this issue.