volsnapshot

Documentation

Index

• func ResetStatus(crd *cosmosalpha.ScheduledVolumeSnapshot)
• type Candidate
• type Client
• type FullNodeControl
  • func NewFullNodeControl(statusClient StatusSyncer, client client.Reader) *FullNodeControl
  • func (control FullNodeControl) ConfirmPodDeletion(ctx context.Context, crd *cosmosalpha.ScheduledVolumeSnapshot) error
  • func (control FullNodeControl) ConfirmPodRestoration(ctx context.Context, crd *cosmosalpha.ScheduledVolumeSnapshot) error
  • func (control FullNodeControl) SignalPodDeletion(ctx context.Context, crd *cosmosalpha.ScheduledVolumeSnapshot) error
  • func (control FullNodeControl) SignalPodRestoration(ctx context.Context, crd *cosmosalpha.ScheduledVolumeSnapshot) error
• type Getter
• type PodFilter
• type Scheduler
  • func NewScheduler(getter Getter) *Scheduler
  • func (s Scheduler) CalcNext(crd *cosmosalpha.ScheduledVolumeSnapshot) (time.Duration, error)
  • func (s Scheduler) IsSnapshotReady(ctx context.Context, crd *cosmosalpha.ScheduledVolumeSnapshot) (bool, error)
• type StatusSyncer
• type VolumeSnapshotControl
  • func NewVolumeSnapshotControl(client Client, filter PodFilter) *VolumeSnapshotControl
  • func (control VolumeSnapshotControl) CreateSnapshot(ctx context.Context, crd *cosmosalpha.ScheduledVolumeSnapshot, ...) error
  • func (control VolumeSnapshotControl) DeleteOldSnapshots(ctx context.Context, log logr.Logger, crd *cosmosalpha.ScheduledVolumeSnapshot) error
  • func (control VolumeSnapshotControl) FindCandidate(ctx context.Context, crd *cosmosalpha.ScheduledVolumeSnapshot) (Candidate, error)

Functions

func ResetStatus

func ResetStatus(crd *cosmosalpha.ScheduledVolumeSnapshot)

ResetStatus resets the CRD's status to appropriate values for the start of a reconcile loop.

Types

type Candidate

type Candidate = cosmosalpha.SnapshotCandidate

type Client

type Client interface {
	Create(ctx context.Context, obj client.Object, opts ...client.CreateOption) error
	List(ctx context.Context, list client.ObjectList, opts ...client.ListOption) error
	Delete(ctx context.Context, obj client.Object, opts ...client.DeleteOption) error
}

Client is a subset of client.Client.

type FullNodeControl

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

FullNodeControl manages a ScheduledVolumeSnapshot's spec.fullNodeRef.

func NewFullNodeControl

func NewFullNodeControl(statusClient StatusSyncer, client client.Reader) *FullNodeControl

func (FullNodeControl) ConfirmPodDeletion

func (control FullNodeControl) ConfirmPodDeletion(ctx context.Context, crd *cosmosalpha.ScheduledVolumeSnapshot) error

ConfirmPodDeletion returns a nil error if the pod is deleted. Any non-nil error is transient, including if the pod has not been deleted yet. Assumes crd's status.candidate is set, otherwise this method panics.

func (FullNodeControl) ConfirmPodRestoration

func (control FullNodeControl) ConfirmPodRestoration(ctx context.Context, crd *cosmosalpha.ScheduledVolumeSnapshot) error

ConfirmPodRestoration verifies the pod has been restored.

func (FullNodeControl) SignalPodDeletion

func (control FullNodeControl) SignalPodDeletion(ctx context.Context, crd *cosmosalpha.ScheduledVolumeSnapshot) error

SignalPodDeletion updates the LocalFullNodeRef's status to indicate it should delete the pod candidate. The pod is gracefully removed to ensure the highest data integrity while taking a VolumeSnapshot. Assumes crd's status.candidate is set, otherwise this method panics. Any error returned can be treated as transient and retried.

func (FullNodeControl) SignalPodRestoration

func (control FullNodeControl) SignalPodRestoration(ctx context.Context, crd *cosmosalpha.ScheduledVolumeSnapshot) error

SignalPodRestoration updates the LocalFullNodeRef's status to indicate it should recreate the pod candidate. Any error returned can be treated as transient and retried.

type Getter

type Getter interface {
	Get(ctx context.Context, key client.ObjectKey, obj client.Object, opts ...client.GetOption) error
}

Getter is a subset of client.Client.

type PodFilter

type PodFilter interface {
	SyncedPods(ctx context.Context, controller client.ObjectKey) []*corev1.Pod
}

type Scheduler

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

Scheduler calculates schedules using crontabs and currently running VolumeSnapshots.

func NewScheduler

func NewScheduler(getter Getter) *Scheduler

func (Scheduler) CalcNext

func (s Scheduler) CalcNext(crd *cosmosalpha.ScheduledVolumeSnapshot) (time.Duration, error)

CalcNext the duration until it's time to take the next snapshot. A zero duration without an error indicates a VolumeSnapshot should be created. Updates crd.status with the last VolumeSnapshot status.

func (Scheduler) IsSnapshotReady

func (s Scheduler) IsSnapshotReady(ctx context.Context, crd *cosmosalpha.ScheduledVolumeSnapshot) (bool, error)

IsSnapshotReady returns true if the status.LastSnapshot is ready for use and updates the crd.status.lastSnapshot. A non-nil error can be treated as transient. If VolumeSnapshot is not found, this indicates a rare case where something deleted the VolumeSnapshot before detecting if it's ready. In that case, this method returns that the snapshot is ready.

type StatusSyncer

type StatusSyncer interface {
	SyncUpdate(ctx context.Context, key client.ObjectKey, update func(status *cosmosv1.FullNodeStatus)) error
}

type VolumeSnapshotControl

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

VolumeSnapshotControl manages VolumeSnapshots

func NewVolumeSnapshotControl

func NewVolumeSnapshotControl(client Client, filter PodFilter) *VolumeSnapshotControl

func (VolumeSnapshotControl) CreateSnapshot

func (control VolumeSnapshotControl) CreateSnapshot(ctx context.Context, crd *cosmosalpha.ScheduledVolumeSnapshot, candidate Candidate) error

CreateSnapshot creates VolumeSnapshot from the Candidate.PVCName and updates crd.status to reflect the created VolumeSnapshot. CreateSnapshot does not set an owner reference to avoid garbage collection if the CRD is deleted. Any error returned is considered transient and can be retried.

func (VolumeSnapshotControl) DeleteOldSnapshots

func (control VolumeSnapshotControl) DeleteOldSnapshots(ctx context.Context, log logr.Logger, crd *cosmosalpha.ScheduledVolumeSnapshot) error

DeleteOldSnapshots deletes old VolumeSnapshots given crd's spec.limit. If limit not set, defaults to keeping the 3 most recent.

func (VolumeSnapshotControl) FindCandidate

func (control VolumeSnapshotControl) FindCandidate(ctx context.Context, crd *cosmosalpha.ScheduledVolumeSnapshot) (Candidate, error)

FindCandidate finds a suitable candidate for creating a volume snapshot. Only selects a pod that is in-sync. Any errors returned can be treated as transient; worth a retry.