Unsolved
1 Rookie
•
6 Posts
0
65
LACP uplinks to Nexus vPC pair
I'm curious, has anyone has successfully set up their VxRail hosts to a Nexus vPC pair, using vPC and not simply trunk links?
The previous regime set up a 6 node vSAN cluster, it's been up for years, uplinks from each host are 2 NIC to a primary Nexus and 2 NIC to a secondary Nexus. The Nexus is a vPC pair. The links were all trunks. We lost the secondary. And the entire VxRail infrastructure was inaccessible, there was no fault tolerance. As a Cisco guy my argument is your distribution layer is a Nexus pair, these links need to be vPC, and is a best practice.
Finding support for this theory on the Dell/EMC VxRail/VMware side is proving to be a challenge.
The fix at the time was a hardware replacement with the exact config and once that new secondary was brought up everything just worked again, so we are still sitting with trunk links to these hosts. And yes without vPC and an outage like this it's like there was a massive spanning tree issue, but that was not the case.
I'm going in circles reading stuff online about how to set up LACP on the VxRail side, and appears at least I'll need to create a new dvSwitch for starters with LACP turned on...and jumbo frames enabled since it was never was :( but my physical NICs are all in use, so not sure how I migrate or what to do about port groups or vmkernel adapters and so on....
In fact when professional services set this up they created a single VMware HCIA Distributed Switch with all of our port groups. In my older vmware infrastructure we have many more dvswitches created separating business units and types of traffic, not to mention vmotion and management..