
This patch makes the necessary charm changes in order to expose and attach SR-IOV VFs to Openstack instances. VFs that support hardware offloading will be handled by ovn, the others will be processed by the Neutron SR-IOV nic agent. neutron-k8s charm ----------------- * enable sriovnicswitch ml2 mechanism driver nova-k8s charm -------------- * add pci-aliases setting * defines aliases for PCI devices, which can be requested through flavor extra specs * passed to openstack-hypervisor through the relation data * enable PciPassthroughFilter and NUMATopologyFilter n-sch filters * enable "filter_scheduler.pci_in_placement" openstack-hypervisor charm -------------------------- * add pci-device-specs setting * allows whitelisting PCI devices (including SR-IOV PF/VFs) Change-Id: Ie118ac84b975275df84af5a378a95bbe49bfeea2 Signed-off-by: Lucian Petrut <lpetrut@cloudbasesolutions.com>
openstack-hypervisor
This charm deploys the openstack-hypervisor snap.
It is expected to be related to the control plane via cross model relations. To achieve this assuming the control plane is in a model called k8s.
juju offer k8s.rabbitmq-k8s:amqp
juju offer k8s.keystone:identity-credentials
juju offer k8s.certificate-authority:certificates
juju offer k8s.ovn-relay:ovsdb-cms-relay
juju relate -m hypervisor openstack-hypervisor admin/k8s.rabbitmq-k8s
juju relate -m hypervisor openstack-hypervisor admin/k8s.keystone
juju relate -m hypervisor openstack-hypervisor admin/k8s.certificate-authority
juju relate -m hypervisor openstack-hypervisor admin/k8s.ovn-relay