Nexus 2Ks and an HA pair of MX105 are having issues

[AP3000]

I have deployed MX105 HA, the LAN side is connected to Nexus 9Ks and there are absolutely no issues. The WAN ports are connected to the service provider in the colo datacenter and they are using Nexus 2Ks for the WAN handoffs. For some unknown reason the WAN port of the primary MX105 does down because on the nexus 2K it does into error disabled state. I hard coded the port on Meraki MX 105 to 1GBps full. Still no luck. Once the service provider does a shut and no shut on their 2K port everything comes back up for 12 hours and then the same error disabled state happens again. No one else is experiencing this issue. 

[alemabrahao]

Can you share the logs?

[Brash]

You'd have to look at the logs from the n2k (or rather whatever Nexus the n2k is attached to)

[PhilipDAth]

Ask the service provider the reason given by the N2K for the port being placed in err-disabled state.

[BryW]

 As the Nexuus 2K are fex switches they are not typical switches. They are essentially external line cards. FEX's can't be connected to another switch, just end user devices. Technically this is because if it receives a BPDU the port will go into an error disabled state.
I am fairly sure this is what is happening, however, the service provider will be able to give you the cause as BPDUerror.

I imagine the Meraki MX is passing through a BPDU received on the LAN ports out of the WAN ports which is triggering the error.
If so I can only think of 3solutions (not saying these are best)

1) Confirm, it is a BPDU and then confirm via a packet trace that the MX is forwarding it out the WAN port.

If so raise a case with meraki as I don't believe the MX should be doing this?

2) Convert it to a routed port if possible as a routed port won't send BPDUs. This may not work with your service provider network.

3) Ask your service provider to allow bpdu to be sent on the port. They won't like that though and it's not best practice.

[AP3000]

Asked the service provided to hardcode the port to 1000 full and that solved the problem. Looks like negotiation issues.