Check several items that may be incorrect on the OpenStack side. We'll start with specific networks:
- Receive the list of network agents with the OpenStack Client:
openstack network agent list --network <network uuid>
- We need the DHCP port IP addresses for the network. This can be collected through the OpenStack horizon GUI or from the OpenStack Client with:
openstack port list --device-owner network:dhcp
- For each DHCP agent we need to perform the following:
- SSH into the relevant controller. The VIO controllers can be gathered from the manager with "viocli get controllers". The SSH command is: "viossh <controller UUID>"
- Check the networking namespaces for qdhcp-<network ID> with: ip a l
- Verify there is a network interface existing with the above command that matches the DHCP ports found in step 2.
- Verify the dnsmasq process for the networks are running with: ps guax | grep dnsmasq
Note: If anything from step #3 is missing, the DHCP agent pod can be deleted. Upon re-creation all network namespaces will be re-created for networks scheduled to the VIO controller.
Once everything is validated OpenStack can safely be ruled out as the issue. The issue exists on the physical network outside of VIO.