Documentation
¶
Index ¶
- Constants
- func GetNodeClaimNameFromVMName(vmName string) string
- type CloudProvider
- func (c *CloudProvider) Create(ctx context.Context, nodeClaim *karpv1.NodeClaim) (*karpv1.NodeClaim, error)
- func (c *CloudProvider) Delete(ctx context.Context, nodeClaim *karpv1.NodeClaim) error
- func (c *CloudProvider) Get(ctx context.Context, providerID string) (*karpv1.NodeClaim, error)
- func (c *CloudProvider) GetInstanceTypes(ctx context.Context, nodePool *karpv1.NodePool) ([]*cloudprovider.InstanceType, error)
- func (c *CloudProvider) GetSupportedNodeClasses() []status.Object
- func (c *CloudProvider) IsDrifted(ctx context.Context, nodeClaim *karpv1.NodeClaim) (cloudprovider.DriftReason, error)
- func (c *CloudProvider) List(ctx context.Context) ([]*karpv1.NodeClaim, error)
- func (c *CloudProvider) LivenessProbe(req *http.Request) error
- func (c *CloudProvider) Name() string
- func (c *CloudProvider) RepairPolicies() []cloudprovider.RepairPolicy
- func (c *CloudProvider) WaitForInstancePromises()
Constants ¶
const ( NodeClassReadinessUnknownReason = "NodeClassReadinessUnknown" InstanceTypeResolutionFailedReason = "InstanceTypeResolutionFailed" CreateInstanceFailedReason = "CreateInstanceFailed" )
const ( NodeClassDrift cloudprovider.DriftReason = "NodeClassDrift" K8sVersionDrift cloudprovider.DriftReason = "K8sVersionDrift" ImageDrift cloudprovider.DriftReason = "ImageDrift" KubeletIdentityDrift cloudprovider.DriftReason = "KubeletIdentityDrift" ClusterConfigDrift cloudprovider.DriftReason = "ClusterConfigDrift" // This is a catch-all for cluster-level config changes (e.g., from PUT ManagedCluster), where Karpenter does not directly "own" them. // TODO (charliedmcb): Use this const across code and test locations which are signaling/checking for "no drift" NoDrift cloudprovider.DriftReason = "" )
Variables ¶
This section is empty.
Functions ¶
func GetNodeClaimNameFromVMName ¶ added in v1.6.6
Types ¶
type CloudProvider ¶
type CloudProvider struct {
// contains filtered or unexported fields
}
func New ¶
func New( instanceTypeProvider instancetype.Provider, vmInstanceProvider instance.VMProvider, aksMachineInstanceProvider instance.AKSMachineProvider, recorder events.Recorder, kubeClient client.Client, imageProvider imagefamily.NodeImageProvider, store *nodeoverlay.InstanceTypeStore, ) *CloudProvider
func (*CloudProvider) Delete ¶
Delete deletes the underlying node Note: Delete may be called many times while delete is ongoing (blocking) as the core Karpenter termination controller watches and reconciles on all node updates (including node status updates, which happen during deletion), so while one Delete call is blocking more will come in every ~5s due to excess Node events + requeues.
func (*CloudProvider) GetInstanceTypes ¶
func (c *CloudProvider) GetInstanceTypes(ctx context.Context, nodePool *karpv1.NodePool) ([]*cloudprovider.InstanceType, error)
GetInstanceTypes returns all available InstanceTypes May return apimachinery.NotFoundError if NodeClass is not found.
func (*CloudProvider) GetSupportedNodeClasses ¶ added in v0.4.0
func (c *CloudProvider) GetSupportedNodeClasses() []status.Object
func (*CloudProvider) IsDrifted ¶
func (c *CloudProvider) IsDrifted(ctx context.Context, nodeClaim *karpv1.NodeClaim) (cloudprovider.DriftReason, error)
IsDrifted checks if the NodeClaim has drifted from our goal state. Note: During the initial launch and registration of a NodeClaim, core calls IsDrifted quite frequently as it waits for the Node to register and become ready. This is because the core pkg/controllers/nodeclaim/disruption/controller.go watches NodeClaims without a generation filter, so any update to the NodeClaim (including updates to status such as when updating conditions during launch) will trigger a call to IsDrifted. The following things produce a large number of IsDrifted calls:
- The initialization controller pkg/controllers/nodeclaim/lifecycle/initialization.go changes the ConditionTypeInitialized condition a number of times during this process, which triggers disruption/controller.go to call IsDrifted each time.
- Any pod scheduling that happens during this time will trigger the core disruption/controller.go, because it watches pod updates and maps each pod update to a NodeClaim event. This means every time a pod (including a DaemonSet pod) is scheduled to the node we'll get called.
func (*CloudProvider) LivenessProbe ¶
func (c *CloudProvider) LivenessProbe(req *http.Request) error
func (*CloudProvider) Name ¶
func (c *CloudProvider) Name() string
Name returns the CloudProvider implementation name.
func (*CloudProvider) RepairPolicies ¶ added in v1.4.0
func (c *CloudProvider) RepairPolicies() []cloudprovider.RepairPolicy
TODO: review repair policies
func (*CloudProvider) WaitForInstancePromises ¶ added in v1.10.0
func (c *CloudProvider) WaitForInstancePromises()
WaitForInstancePromises blocks until all in-flight async Create goroutines have completed.
Source Files
Directories