recommender

Documentation

Index

• func ScaleDownOrderedByDescendingCost(ctx context.Context, vca scalesim.VirtualClusterAccess, w http.ResponseWriter, ...) ([]string, error)
• type Recommendation
• type Recommender
  • func NewRecommender(engine scalesim.Engine, scenarioName, podOrder string, shoot *v1beta1.Shoot, ...) *Recommender
  • func (r *Recommender) Run(ctx context.Context, unscheduledPods []corev1.Pod) ([]Recommendation, error)
• type StrategyWeights

Functions

func ScaleDownOrderedByDescendingCost

func ScaleDownOrderedByDescendingCost(ctx context.Context, vca scalesim.VirtualClusterAccess, w http.ResponseWriter, nodes []corev1.Node) ([]string, error)

ScaleDownOrderedByDescendingCost scales down the nodes in the cluster ordered by descending cost. It does the following

0. Order all existing nodes by their cost.
1. Iterate over all existing ordered nodes, for each node: 1.1 Taint the node with NoSchedule. 1.2 Get the pods on the node and deploy a copy of the pods with new names in the cluster. 1.3 Check whether there are un-scheduled pods. If len(unscheduledPods) == 0 { then delete the node from virtual cluster and record as recommendation else { mark as essential (not to be removed). Delete the newly deployed pods. Un-taint the node if this node is essential. }

Types

type Recommendation

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

type Recommender

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

func NewRecommender

func NewRecommender(engine scalesim.Engine, scenarioName, podOrder string, shoot *v1beta1.Shoot, instanceTypeCostRatios map[string]float64, strategyWeights StrategyWeights, logWriter http.ResponseWriter) *Recommender

func (*Recommender) Run

func (r *Recommender) Run(ctx context.Context, unscheduledPods []corev1.Pod) ([]Recommendation, error)

type StrategyWeights

type StrategyWeights struct {
	LeastWaste float64
	LeastCost  float64
}