Compute requirements

In order for LLDP topology discovery from the fabric to function correctly, all computes must have LLDP enabled on their data interfaces. Only when LLDP is enabled on all computes can workloads be automatically scheduled on the fabric.

Both the OpenStack OSPD/CBIS and Vmware integration enable this for you. Manual intervention is needed when not using those integrations.

• On a Linux-based compute, enable LLDP:

  NIC_DIR="/sys/class/net"
  for itf in $(ls $NIC_DIR | grep -E 'eno|enp|ens') 
  do
      if [ -d "${NIC_DIR}/${itf}/device" -a ! -L "${NIC_DIR}/${itf}/device/physfn" ]
      then
          lldptool set-lldp -i $itf adminStatus=Tx
          lldptool -T -i $itf -V sysName enableTx=yes
          lldptool -T -i $itf -V portID subtype=PORT_ID_INTERFACE_NAME
      fi
  done
  

• On VMware-based systems, ensure that LLDP is enabled to both send and receive LLDP in the Discovery protocol settings of the distributed vSwitch.
• For NICs that support hardware based LLDP (in-nic LLDP), make sure to disable this capability as it may send out conflicting LLDP information compared to what Linux or VMware would send out.
• The hostname returned by hostnamectl should correspond to hostname in output of openstack host list.

  If these are not the same, use the command hostnamectl set-hostname to align them.