Summary:

• We (RHOSO pcinfra team) have beaker nodes in RDU2, but there is no L2 to RDU4.
• It is not supported to have VXLAN between RDU locations (Petr had asked IT Network few months ago), so we can only use DCN topology and virtual-media based provisioning of EDPM hosts with BMO.
• In RHOSO, CEPH is the only storage tested with DCN topology. There is support for other than ceph backends in rhoso, the storage team tracker https://issues.redhat.com/browse/RHOSRFE-268. So we could help them with validating this topology as well

Goal:

• Be able to use an available beaker node in rdu2 to deploy SNO by cifmw for connecting it with EDPM hardware in RDU4, via DCN, to test features in RHOSO-19 (also 18, if needed).  EDPM could be pre-provisioned with beaker as well.
• We might use a standard hybrid OCP setup on a hypervisor on rdu2 as well, unless the node is not usable as VIRTHOST, then we will need cifmw changes to deploy SNO (in this task scope)

Acceptance criteria:

• cifmw can be used to deploy SNO on a target host directly (changing the execution mode of dev-scripts from the hypervisor host to CI controller)
• We can deploy SNO directly on the beaker host in rdu2 with cifmw in downstream zuul testproject
• We can verify that the rdu4 beaker node with GPU HW is ping-able from SNO host in rdu2

Deliverables/Outcomes:

• We know how to to test features in RDU4, and can do that
• We know how to set up a control plane to connect to the GPU hardware in RDU4, and can do that