Documentation ¶
Overview ¶
Package master contains code for setting up and running a Kubernetes cluster master.
Index ¶
- func NewEtcdHelper(client tools.EtcdGetSet, version string) (helper tools.EtcdHelper, err error)
- func NewHandlerContainer(mux *http.ServeMux) *restful.Container
- func NewIPCache(cp cloudprovider.Interface, clock util.Clock, ttl time.Duration) *ipCache
- func RESTStorageToNodes(storage apiserver.RESTStorage) client.NodesInterface
- type Config
- type IPGetter
- type Master
- type PodCache
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func NewEtcdHelper ¶
func NewEtcdHelper(client tools.EtcdGetSet, version string) (helper tools.EtcdHelper, err error)
NewEtcdHelper returns an EtcdHelper for the provided arguments or an error if the version is incorrect.
func NewHandlerContainer ¶
func NewIPCache ¶
NewIPCache makes a new ip caching layer, which will get IP addresses from cp, and use clock for deciding when to re-get an IP address. Thread-safe.
TODO: when we switch to go1.4, this class would be a good candidate for something that could be produced from a template and a type via `go generate`.
func RESTStorageToNodes ¶
func RESTStorageToNodes(storage apiserver.RESTStorage) client.NodesInterface
RESTStorageToNodes will take a RESTStorage object and return a client interface which will work for any use expecting a client.Nodes() interface. The advantage of doing this in server code is that it doesn't require an actual trip through the network.
TODO: considering that the only difference between the various client types and RESTStorage type is the type of the arguments, maybe use "go generate" to write a specialized adaptor for every client type?
TODO: this also means that pod and node API endpoints have to be colocated in the same process
Types ¶
type Config ¶
type Config struct { Client *client.Client Cloud cloudprovider.Interface EtcdHelper tools.EtcdHelper HealthCheckMinions bool EventTTL time.Duration MinionRegexp string KubeletClient client.KubeletClient PortalNet *net.IPNet EnableLogsSupport bool EnableUISupport bool EnableSwaggerSupport bool EnableV1Beta3 bool APIPrefix string CorsAllowedOriginList util.StringList Authenticator authenticator.Request Authorizer authorizer.Authorizer AdmissionControl admission.Interface MasterServiceNamespace string // If specified, all web services will be registered into this container RestfulContainer *restful.Container // Number of masters running; all masters must be started with the // same value for this field. (Numbers > 1 currently untested.) MasterCount int // The port on PublicAddress where a read-only server will be installed. // Defaults to 7080 if not set. ReadOnlyPort int // The port on PublicAddress where a read-write server will be installed. // Defaults to 443 if not set. ReadWritePort int // If empty, the first result from net.InterfaceAddrs will be used. PublicAddress string }
Config is a structure used to configure a Master.
type Master ¶
type Master struct { // "Outputs" Handler http.Handler InsecureHandler http.Handler // contains filtered or unexported fields }
Master contains state for a Kubernetes cluster master/api server.
func New ¶
New returns a new instance of Master from the given config. Certain config fields will be set to a default value if unset, including:
PortalNet MasterCount ReadOnlyPort ReadWritePort PublicAddress
Certain config fields must be specified, including:
KubeletClient
Public fields:
Handler -- The returned master has a field TopHandler which is an http.Handler which handles all the endpoints provided by the master, including the API, the UI, and miscelaneous debugging endpoints. All these are subject to authorization and authentication. InsecureHandler -- an http.Handler which handles all the same endpoints as Handler, but no authorization and authentication is done.
Public methods:
HandleWithAuth -- Allows caller to add an http.Handler for an endpoint that uses the same authentication and authorization (if any is configured) as the master's built-in endpoints. If the caller wants to add additional endpoints not using the master's auth, then the caller should create a handler for those endpoints, which delegates the any unhandled paths to "Handler".
func (*Master) HandleFuncWithAuth ¶
func (m *Master) HandleFuncWithAuth(pattern string, handler func(http.ResponseWriter, *http.Request))
HandleFuncWithAuth adds an http.Handler for pattern to an http.ServeMux Applies the same authentication and authorization (if any is configured) to the request is used for the master's built-in endpoints.
func (*Master) HandleWithAuth ¶
HandleWithAuth adds an http.Handler for pattern to an http.ServeMux Applies the same authentication and authorization (if any is configured) to the request is used for the master's built-in endpoints.
func (*Master) InstallSwaggerAPI ¶
func (m *Master) InstallSwaggerAPI()
InstallSwaggerAPI installs the /swaggerapi/ endpoint to allow schema discovery and traversal. It is optional to allow consumers of the Kubernetes master to register their own web services into the Kubernetes mux prior to initialization of swagger, so that other resource types show up in the documentation.
type PodCache ¶
type PodCache struct {
// contains filtered or unexported fields
}
PodCache contains both a cache of container information, as well as the mechanism for keeping that cache up to date.
func NewPodCache ¶
func NewPodCache(ipCache IPGetter, info client.PodInfoGetter, nodes client.NodeInterface, pods pod.Registry) *PodCache
NewPodCache returns a new PodCache which watches container information registered in the given PodRegistry. TODO(lavalamp): pods should be a client.PodInterface.
func (*PodCache) GetPodStatus ¶
GetPodStatus gets the stored pod status.
func (*PodCache) UpdateAllContainers ¶
func (p *PodCache) UpdateAllContainers()
UpdateAllContainers updates information about all containers. Callers should let one call to UpdateAllContainers finish before calling again, or risk having new info getting clobbered by delayed old info.