network

package
v1.5.5 Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: Mar 22, 2017 License: Apache-2.0 Imports: 14 Imported by: 0

Documentation

Index

Constants

View Source
const DefaultInterfaceName = "eth0"

TODO: Consider making this value configurable.

View Source
const DefaultPluginName = "kubernetes.io/no-op"
View Source
const (
	// Indicates the plugin handles Kubernetes bandwidth shaping annotations internally
	NET_PLUGIN_CAPABILITY_SHAPING int = 1
)

Plugin capabilities

View Source
const NET_PLUGIN_EVENT_POD_CIDR_CHANGE = "pod-cidr-change"

Called when the node's Pod CIDR is known when using the controller manager's --allocate-node-cidrs=true option

View Source
const NET_PLUGIN_EVENT_POD_CIDR_CHANGE_DETAIL_CIDR = "pod-cidr"
View Source
const UseDefaultMTU = 0

UseDefaultMTU is a marker value that indicates the plugin should determine its own MTU It is the zero value, so a non-initialized value will mean "UseDefault"

Variables

This section is empty.

Functions

func GetPodIP added in v1.3.0

func GetPodIP(execer utilexec.Interface, nsenterPath, netnsPath, interfaceName string) (net.IP, error)

GetPodIP gets the IP of the pod by inspecting the network info inside the pod's network namespace.

func UnescapePluginName

func UnescapePluginName(in string) string

Types

type Host

type Host interface {
	// NamespaceGetter is a getter for sandbox namespace information.
	// It's the only part of this interface that isn't currently deprecated.
	NamespaceGetter

	// LegacyHost contains methods that trap back into the Kubelet. Dependence
	// *do not* add more dependencies in this interface. In a post-cri world,
	// network plugins will be invoked by the runtime shim, and should only
	// require NamespaceGetter.
	LegacyHost
}

Host is an interface that plugins can use to access the kubelet. TODO(#35457): get rid of this backchannel to the kubelet. The scope of the back channel is restricted to host-ports/testing, and restricted to kubenet. No other network plugin wrapper needs it. Other plugins only require a way to access namespace information, which they can do directly through the embedded NamespaceGetter.

type LegacyHost added in v1.5.0

type LegacyHost interface {
	// Get the pod structure by its name, namespace
	// Only used for hostport management and bw shaping
	GetPodByName(namespace, name string) (*api.Pod, bool)

	// GetKubeClient returns a client interface
	// Only used in testing
	GetKubeClient() clientset.Interface

	// GetContainerRuntime returns the container runtime that implements the containers (e.g. docker/rkt)
	// Only used for hostport management
	GetRuntime() kubecontainer.Runtime

	// SupportsLegacyFeaturs returns true if this host can support hostports
	// and bandwidth shaping. Both will either get added to CNI or dropped,
	// so differnt implementations can choose to ignore them.
	SupportsLegacyFeatures() bool
}

LegacyHost implements the methods required by network plugins that were directly invoked by the kubelet. Implementations of this interface that do not wish to support these features can simply return false to SupportsLegacyFeatures.

type NamespaceGetter added in v1.5.0

type NamespaceGetter interface {
	// GetNetNS returns network namespace information for the given containerID.
	GetNetNS(containerID string) (string, error)
}

NamespaceGetter is an interface to retrieve namespace information for a given sandboxID. Typically implemented by runtime shims that are closely coupled to CNI plugin wrappers like kubenet.

type NetworkPlugin

type NetworkPlugin interface {
	// Init initializes the plugin.  This will be called exactly once
	// before any other methods are called.
	Init(host Host, hairpinMode componentconfig.HairpinMode, nonMasqueradeCIDR string, mtu int) error

	// Called on various events like:
	// NET_PLUGIN_EVENT_POD_CIDR_CHANGE
	Event(name string, details map[string]interface{})

	// Name returns the plugin's name. This will be used when searching
	// for a plugin by name, e.g.
	Name() string

	// Returns a set of NET_PLUGIN_CAPABILITY_*
	Capabilities() utilsets.Int

	// SetUpPod is the method called after the infra container of
	// the pod has been created but before the other containers of the
	// pod are launched.
	// TODO: rename podInfraContainerID to sandboxID
	SetUpPod(namespace string, name string, podInfraContainerID kubecontainer.ContainerID) error

	// TearDownPod is the method called before a pod's infra container will be deleted
	// TODO: rename podInfraContainerID to sandboxID
	TearDownPod(namespace string, name string, podInfraContainerID kubecontainer.ContainerID) error

	// Status is the method called to obtain the ipv4 or ipv6 addresses of the container
	// TODO: rename podInfraContainerID to sandboxID
	GetPodNetworkStatus(namespace string, name string, podInfraContainerID kubecontainer.ContainerID) (*PodNetworkStatus, error)

	// NetworkStatus returns error if the network plugin is in error state
	Status() error
}

Plugin is an interface to network plugins for the kubelet

func InitNetworkPlugin

func InitNetworkPlugin(plugins []NetworkPlugin, networkPluginName string, host Host, hairpinMode componentconfig.HairpinMode, nonMasqueradeCIDR string, mtu int) (NetworkPlugin, error)

InitNetworkPlugin inits the plugin that matches networkPluginName. Plugins must have unique names.

type NoopNetworkPlugin added in v1.3.0

type NoopNetworkPlugin struct {
}

func (*NoopNetworkPlugin) Capabilities added in v1.3.0

func (plugin *NoopNetworkPlugin) Capabilities() utilsets.Int

func (*NoopNetworkPlugin) Event added in v1.3.0

func (plugin *NoopNetworkPlugin) Event(name string, details map[string]interface{})

func (*NoopNetworkPlugin) GetPodNetworkStatus added in v1.3.0

func (plugin *NoopNetworkPlugin) GetPodNetworkStatus(namespace string, name string, id kubecontainer.ContainerID) (*PodNetworkStatus, error)

func (*NoopNetworkPlugin) Init added in v1.3.0

func (plugin *NoopNetworkPlugin) Init(host Host, hairpinMode componentconfig.HairpinMode, nonMasqueradeCIDR string, mtu int) error

func (*NoopNetworkPlugin) Name added in v1.3.0

func (plugin *NoopNetworkPlugin) Name() string

func (*NoopNetworkPlugin) SetUpPod added in v1.3.0

func (plugin *NoopNetworkPlugin) SetUpPod(namespace string, name string, id kubecontainer.ContainerID) error

func (*NoopNetworkPlugin) Status added in v1.3.0

func (plugin *NoopNetworkPlugin) Status() error

func (*NoopNetworkPlugin) TearDownPod added in v1.3.0

func (plugin *NoopNetworkPlugin) TearDownPod(namespace string, name string, id kubecontainer.ContainerID) error

type PodNetworkStatus added in v1.1.0

type PodNetworkStatus struct {
	unversioned.TypeMeta `json:",inline"`

	// IP is the primary ipv4/ipv6 address of the pod. Among other things it is the address that -
	//   - kube expects to be reachable across the cluster
	//   - service endpoints are constructed with
	//   - will be reported in the PodStatus.PodIP field (will override the IP reported by docker)
	IP net.IP `json:"ip" description:"Primary IP address of the pod"`
}

PodNetworkStatus stores the network status of a pod (currently just the primary IP address) This struct represents version "v1beta1"

Directories

Path Synopsis
cni
testing
mock_cni is a mock of the `libcni.CNI` interface.
mock_cni is a mock of the `libcni.CNI` interface.
Package exec scans and loads networking plugins that are installed under /usr/libexec/kubernetes/kubelet-plugins/net/exec/ The layout convention for a plugin is: plugin-name/ (plugins have to be directories first) plugin-name/plugin-name (executable that will be called out, see Vendoring Note for more nuances) plugin-name/<other-files> where, 'executable' has the following requirements: - should have exec permissions - should give non-zero exit code on failure, and zero on success - the arguments will be <action> <pod_namespace> <pod_name> <docker_id_of_infra_container> whereupon, <action> will be one of: - init, called when the kubelet loads the plugin - setup, called after the infra container of a pod is created, but before other containers of the pod are created - teardown, called before the pod infra container is killed - status, called at regular intervals and is supposed to return a json formatted output indicating the pod's IPAddress(v4/v6).
Package exec scans and loads networking plugins that are installed under /usr/libexec/kubernetes/kubelet-plugins/net/exec/ The layout convention for a plugin is: plugin-name/ (plugins have to be directories first) plugin-name/plugin-name (executable that will be called out, see Vendoring Note for more nuances) plugin-name/<other-files> where, 'executable' has the following requirements: - should have exec permissions - should give non-zero exit code on failure, and zero on success - the arguments will be <action> <pod_namespace> <pod_name> <docker_id_of_infra_container> whereupon, <action> will be one of: - init, called when the kubelet loads the plugin - setup, called after the infra container of a pod is created, but before other containers of the pod are created - teardown, called before the pod infra container is killed - status, called at regular intervals and is supposed to return a json formatted output indicating the pod's IPAddress(v4/v6).

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL