snapshot

Documentation

Index

• type ResourceClaimId
  • func GetClaimId(claim *resourceapi.ResourceClaim) ResourceClaimId
• type Snapshot
  • func NewSnapshot(claims map[ResourceClaimId]*resourceapi.ResourceClaim, ...) Snapshot
  • func (s Snapshot) AddClaims(newClaims []*resourceapi.ResourceClaim) error
  • func (s Snapshot) AddNodeResourceSlices(nodeName string, slices []*resourceapi.ResourceSlice) error
  • func (s Snapshot) Clone() Snapshot
  • func (s Snapshot) DeviceClasses() schedulerframework.DeviceClassLister
  • func (s Snapshot) NodeResourceSlices(nodeName string) ([]*resourceapi.ResourceSlice, bool)
  • func (s Snapshot) PodClaims(pod *apiv1.Pod) ([]*resourceapi.ResourceClaim, error)
  • func (s Snapshot) RemoveNodeResourceSlices(nodeName string)
  • func (s Snapshot) RemovePodOwnedClaims(pod *apiv1.Pod)
  • func (s Snapshot) ReservePodClaims(pod *apiv1.Pod) error
  • func (s Snapshot) ResourceClaims() schedulerframework.ResourceClaimTracker
  • func (s Snapshot) ResourceSlices() schedulerframework.ResourceSliceLister
  • func (s Snapshot) UnreservePodClaims(pod *apiv1.Pod) error
  • func (s Snapshot) WrapSchedulerNodeInfo(schedNodeInfo *schedulerframework.NodeInfo) (*framework.NodeInfo, error)

Types

type ResourceClaimId

type ResourceClaimId struct {
	Name      string
	Namespace string
}

ResourceClaimId is a unique identifier for a ResourceClaim.

func GetClaimId

func GetClaimId(claim *resourceapi.ResourceClaim) ResourceClaimId

GetClaimId returns the unique identifier for a ResourceClaim.

type Snapshot

type Snapshot struct {
	// contains filtered or unexported fields
}

Snapshot contains a snapshot of all DRA objects taken at a ~single point in time. The Snapshot should be obtained via the Provider. Then, it can be modified using the exposed methods, to simulate scheduling actions in the cluster.

func NewSnapshot

func NewSnapshot(claims map[ResourceClaimId]*resourceapi.ResourceClaim, nodeLocalSlices map[string][]*resourceapi.ResourceSlice, globalSlices []*resourceapi.ResourceSlice, deviceClasses map[string]*resourceapi.DeviceClass) Snapshot

NewSnapshot returns a Snapshot created from the provided data.

func (Snapshot) AddClaims

func (s Snapshot) AddClaims(newClaims []*resourceapi.ResourceClaim) error

AddClaims adds additional ResourceClaims to the Snapshot. It can be used e.g. if we need to duplicate a Pod that owns ResourceClaims. Returns an error if any of the claims is already tracked in the snapshot.

func (Snapshot) AddNodeResourceSlices

func (s Snapshot) AddNodeResourceSlices(nodeName string, slices []*resourceapi.ResourceSlice) error

AddNodeResourceSlices adds additional node-local ResourceSlices to the Snapshot. This should be used whenever a Node with node-local ResourceSlices is duplicated in the cluster snapshot.

func (Snapshot) Clone

func (s Snapshot) Clone() Snapshot

Clone returns a copy of this Snapshot that can be independently modified without affecting this Snapshot. The only mutable objects in the Snapshot are ResourceClaims, so they are deep-copied. The rest is only a shallow copy.

func (Snapshot) DeviceClasses

func (s Snapshot) DeviceClasses() schedulerframework.DeviceClassLister

DeviceClasses exposes the Snapshot as schedulerframework.DeviceClassLister, in order to interact with the scheduler framework.

func (Snapshot) NodeResourceSlices

func (s Snapshot) NodeResourceSlices(nodeName string) ([]*resourceapi.ResourceSlice, bool)

NodeResourceSlices returns all node-local ResourceSlices for the given Node.

func (Snapshot) PodClaims

func (s Snapshot) PodClaims(pod *apiv1.Pod) ([]*resourceapi.ResourceClaim, error)

PodClaims returns ResourceClaims objects for all claims referenced by the Pod. If any of the referenced claims isn't tracked in the Snapshot, an error is returned.

func (Snapshot) RemoveNodeResourceSlices

func (s Snapshot) RemoveNodeResourceSlices(nodeName string)

RemoveNodeResourceSlices removes all node-local ResourceSlices for the Node with the given nodeName. It's a no-op if there aren't any slices to remove.

func (Snapshot) RemovePodOwnedClaims

func (s Snapshot) RemovePodOwnedClaims(pod *apiv1.Pod)

RemovePodOwnedClaims iterates over all the claims referenced by the Pod, and removes the ones owned by the Pod from the Snapshot. Claims referenced by the Pod but not owned by it are not removed, but the Pod's reservation is removed from them. This method removes all relevant claims that are in the snapshot, and doesn't error out if any of the claims are missing.

func (Snapshot) ReservePodClaims

func (s Snapshot) ReservePodClaims(pod *apiv1.Pod) error

ReservePodClaims adds a reservation for the provided Pod to all the claims it references. If any of the referenced claims isn't tracked in the Snapshot, or if any of the claims are already at maximum reservation count, an error is returned.

func (Snapshot) ResourceClaims

func (s Snapshot) ResourceClaims() schedulerframework.ResourceClaimTracker

ResourceClaims exposes the Snapshot as schedulerframework.ResourceClaimTracker, in order to interact with the scheduler framework.

func (Snapshot) ResourceSlices

func (s Snapshot) ResourceSlices() schedulerframework.ResourceSliceLister

ResourceSlices exposes the Snapshot as schedulerframework.ResourceSliceLister, in order to interact with the scheduler framework.

func (Snapshot) UnreservePodClaims

func (s Snapshot) UnreservePodClaims(pod *apiv1.Pod) error

UnreservePodClaims removes reservations for the provided Pod from all the claims it references. If any of the referenced claims isn't tracked in the Snapshot, an error is returned. If a claim is owned by the pod, or if the claim has no more reservations, its allocation is cleared.

func (Snapshot) WrapSchedulerNodeInfo

func (s Snapshot) WrapSchedulerNodeInfo(schedNodeInfo *schedulerframework.NodeInfo) (*framework.NodeInfo, error)

WrapSchedulerNodeInfo wraps the provided *schedulerframework.NodeInfo into an internal *framework.NodeInfo, adding dra information. Node-local ResourceSlices are added to the NodeInfo, and all ResourceClaims referenced by each Pod are added to each PodInfo. Returns an error if any of the Pods is missing a ResourceClaim.