Wednesday, September 12, 2018

Does LACP between Cisco and Linux (Ubuntu 18.04.1) simply not work?

My Google-fu is failing me and this is incredibly frustrating.

I'm trying to get an LACP port-channel setup between a Nexus 7706 and an Ubuntu 18.04.1 host. Of course, the Nexus side went into "suspended" for each of the two links, because no LACP PDUs are getting through from the Linux host. Apparently this is a common problem.

The common wisdom however, is to simply do "no lacp suspend-individual" and call it a day. I'm not ok with that. If LACP PDU frames arent getting through at all then why even have LACP? I found a Cisco doc that says on a UCS Server the solution is to change the native VLAN of the server's ports, because the PDUs are not getting through to the switch on VLAN 0, due to VLAN 0 being invalid on the switch. (https://www.cisco.com/c/en/us/support/docs/switches/nexus-9000-series-switches/118851-technote-lacp-00.html) No problem...wait...it seems that's not possible to do on Ubuntu 18.04.1, and I don't even see anyone explaining how to do it on RHEL, either. What gives? Is Cisco's implementation of 802.3ad simply not compatible with the Linux implementation? Is this simply accepted and people run link aggregation sans LACP when connecting to Linux hosts? It doesn't seem to be a useful workaround to simply force the port to come up despite no LACP frames being received by the switch.

Anyone faced this before? Any tips that I can do on the Nexus to get it to accept the PDUs from the Linux host?

It looks like the netplan YAML file doesn't have may options available, so I don't think there's much to be gained from the Ubuntu side, unless I can force the LACP PDUs to be on VLAN1 or untagged using iptables or something similar.

plz halp!



No comments:

Post a Comment