This plugin allows users to define Kubernetes networks on top of existing host interfaces. By using the macvtap plugin, the user is able to directly connect the pod to a host interface and consume it through a tap device.
The main use cases are virtualization workloads inside the pod driven by Kubevirt but it can also be used directly with QEMU/libvirt and it might be suitable combined with other virtualization backends.
macvtap CNI includes a device plugin to properly expose the macvtap interfaces to the pods. A metaplugin such as Multus gets the name of the interface allocated by the device plugin and is responsible to invoke the cni plugin with that name as deviceID.
The device plugin is configured through environment variable DP_MACVTAP_CONF
.
The value is a json array and each element of the array is a separate resource
to be made available:
name
(string, required) the name of the resourcelowerDevice
(string, required) the name of the macvtap lower linkmode
(string, optional, default=bridge) the macvtap operating modecapacity
(uint, optional, default=100) the capacity of the resource
In the default deployment, this configuration shall be provided through a config map, for example:
kind: ConfigMap
apiVersion: v1
metadata:
name: macvtap-deviceplugin-config
data:
DP_MACVTAP_CONF: |
[ {
"name" : "dataplane",
"lowerDevice" : "eth0",
"mode": "bridge",
"capacity" : 50
} ]
$ kubectl apply -f https://raw.githubusercontent.com/kubevirt/macvtap-cni/main/examples/macvtap-deviceplugin-config.yaml
configmap "macvtap-deviceplugin-config" created
This configuration will result in up to 50 macvtap interfaces being offered for
consumption, using eth0 as the lower device, in bridge mode, and under
resource name macvtap.network.kubevirt.io/dataplane
.
A configuration consisting of an empty json array, as proposed in the default
example, causes the device
plugin to expose one resource for every physical link or bond on each node. For
example, if a node has a physical link called eth0, a resourced named
macvtap.network.kubevirt.io/eth0
would be made available to use macvtap
interfaces with eth0 as the lower device
The macvtap CNI can be deployed using the proposed daemon set:
$ kubectl apply -f https://raw.githubusercontent.com/kubevirt/macvtap-cni/main/manifests/macvtap.yaml
daemonset "macvtap-cni" created
$ kubectl get pods
NAME READY STATUS RESTARTS AGE
macvtap-cni-745x4 1/1 Running 0 5m
This will result in the CNI being installed and device plugin running on all nodes.
There is also a template available to parameterize the deployment with different configuration options.
macvtap CNI is best used with Multus by defining a NetworkAttachmentDefinition:
kind: NetworkAttachmentDefinition
apiVersion: k8s.cni.cncf.io/v1
metadata:
name: dataplane
annotations:
k8s.v1.cni.cncf.io/resourceName: macvtap.network.kubevirt.io/dataplane
spec:
config: '{
"cniVersion": "0.3.1",
"name": "dataplane",
"type": "macvtap",
"mtu": 1500
}'
The CNI config json allows the following parameters:
name
(string, required): the name of the network. Optional when used within a NetworkAttachmentDefinition, as Multus provides the name in that case.type
(string, required): "macvtap".mac
(string, optional): mac address to assign to the macvtap interface.mtu
(integer, optional): mtu to set in the macvtap interface.deviceID
(string, required): name of an existing macvtap host interface, which will be moved to the correct net namespace and configured. Optional when used within a NetworkAttachmentDefinition, as Multus provides the deviceID in that case.
A pod can be attached to that network which would result in the pod having the corresponding macvtap interface:
apiVersion: v1
kind: Pod
metadata:
name: pod
annotations:
k8s.v1.cni.cncf.io/networks: dataplane
spec:
containers:
- name: busybox
image: busybox
command: ["/bin/sleep", "1800"]
resources:
limits:
macvtap.network.kubevirt.io/dataplane: 1
A mac can also be assigned to the interface through the network annotation:
apiVersion: v1
kind: Pod
metadata:
name: pod-with-mac
annotations:
k8s.v1.cni.cncf.io/networks: |
[
{
"name":"dataplane",
"mac": "02:23:45:67:89:01"
}
]
spec:
containers:
- name: busybox
image: busybox
command: ["/bin/sleep", "1800"]
resources:
limits:
macvtap.network.kubevirt.io/dataplane: 1
Note: The resource limit can be ommited from the pod definition if network-resources-injector is deployed in the cluster.
The device plugin can potentially be used by itself in case you only need the tap device in the pod and not the interface:
apiVersion: v1
kind: Pod
metadata:
name: macvtap-consumer
spec:
containers:
- name: busybox
image: busybox
command: ["/bin/sleep", "123"]
resources:
limits:
macvtap.network.kubevirt.io/dataplane: 1