autoscaler/cluster-autoscaler/simulator/cluster.go

/*
Copyright 2016 The Kubernetes Authors All rights reserved.

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/

package simulator

import (
	"fmt"
	"math"

	kube_api "k8s.io/kubernetes/pkg/api"
	"k8s.io/kubernetes/pkg/api/resource"
	kube_client "k8s.io/kubernetes/pkg/client/unversioned"
	cmd "k8s.io/kubernetes/pkg/kubectl/cmd"
	"k8s.io/kubernetes/plugin/pkg/scheduler/schedulercache"

	"github.com/golang/glog"
)

const (
	// Nodes with utilization below this are considered unused and may be subject to scale down.
	unusedThreshold = float64(0.5)
)

// FindNodeToRemove finds a node that can be removed.
func FindNodeToRemove(nodes []*kube_api.Node, pods []*kube_api.Pod, client *kube_client.Client) (*kube_api.Node, error) {
	nodeNameToNodeInfo := schedulercache.CreateNodeNameToInfoMap(pods)

	//TODO: Interate over underutulized nodes first.
	for _, node := range nodes {
		nodeInfo, found := nodeNameToNodeInfo[node.Name]
		if !found {
			glog.Errorf("Node info for %s not found", node.Name)
			continue
		}

		reservation, err := calculateReservation(node, nodeInfo)

		if err != nil {
			glog.Warningf("Failed to calculate reservation for %s: %v", node.Name, err)
		}
		glog.V(4).Infof("Node %s - reservation %f", node.Name, reservation)

		if reservation > unusedThreshold {
			glog.Infof("Node %s is not suitable for removal - reservation to big (%f)", node.Name, reservation)
			continue
		}

		// Let's try to remove this one.
		glog.V(2).Infof("Considering %s for removal", node.Name)

		podsToRemoveList, _, _, err := cmd.GetPodsForDeletionOnNodeDrain(client, node.Name,
			kube_api.Codecs.UniversalDecoder(), false, true)

		if err != nil {
			glog.V(1).Infof("Node %s cannot be removed: %v", node.Name, err)
			continue
		}

		tempNodeNameToNodeInfo := schedulercache.CreateNodeNameToInfoMap(pods)
		delete(tempNodeNameToNodeInfo, node.Name)
		for _, node := range nodes {
			if nodeInfo, found := tempNodeNameToNodeInfo[node.Name]; found {
				nodeInfo.SetNode(node)
			}
		}
		ptrPodsToRemove := make([]*kube_api.Pod, 0, len(podsToRemoveList))
		for i := range podsToRemoveList {
			ptrPodsToRemove = append(ptrPodsToRemove, &podsToRemoveList[i])
		}

		findProblems := findPlaceFor(ptrPodsToRemove, nodes, tempNodeNameToNodeInfo)
		if findProblems == nil {
			return node, nil
		}
		glog.Infof("Node %s is not suitable for removal %v", node.Name, err)
	}
	return nil, nil
}

func calculateReservation(node *kube_api.Node, nodeInfo *schedulercache.NodeInfo) (float64, error) {
	cpu, err := calculateReservationOfResource(node, nodeInfo, kube_api.ResourceCPU)
	if err != nil {
		return 0, err
	}
	mem, err := calculateReservationOfResource(node, nodeInfo, kube_api.ResourceMemory)
	if err != nil {
		return 0, err
	}
	return math.Max(cpu, mem), nil
}

func calculateReservationOfResource(node *kube_api.Node, nodeInfo *schedulercache.NodeInfo, resourceName kube_api.ResourceName) (float64, error) {
	nodeCapacity, found := node.Status.Capacity[resourceName]
	if !found {
		return 0, fmt.Errorf("Failed to get %v from %s", resourceName, node.Name)
	}
	if nodeCapacity.MilliValue() == 0 {
		return 0, fmt.Errorf("%v is 0 at %s", resourceName, node.Name)
	}
	podsRequest := resource.MustParse("0")
	for _, pod := range nodeInfo.Pods() {
		for _, container := range pod.Spec.Containers {
			if resourceValue, found := container.Resources.Requests[resourceName]; found {
				podsRequest.Add(resourceValue)
			}
		}
	}
	return float64(podsRequest.MilliValue()) / float64(nodeCapacity.MilliValue()), nil
}

// TODO: We don't need to pass list of nodes here as they are already available in nodeInfos.
func findPlaceFor(pods []*kube_api.Pod, nodes []*kube_api.Node, nodeInfos map[string]*schedulercache.NodeInfo) error {
	predicateChecker := NewPredicateChecker()
	for _, pod := range pods {
		foundPlace := false
		glog.V(4).Infof("Looking for place for %s/%s", pod.Namespace, pod.Name)

		// TODO: Sort nodes by reservation
	nodeloop:
		for _, node := range nodes {
			node.Status.Allocatable = node.Status.Capacity
			if nodeInfo, found := nodeInfos[node.Name]; found {
				err := predicateChecker.CheckPredicates(pod, nodeInfo)
				glog.V(4).Infof("Evaluation %s for %s/%s -> %v", node.Name, pod.Namespace, pod.Name, err)
				if err == nil {
					foundPlace = true
					// TODO(mwielgus): Optimize it.
					podsOnNode := nodeInfo.Pods()
					podsOnNode = append(podsOnNode, pod)
					nodeInfos[node.Name] = schedulercache.NewNodeInfo(podsOnNode...)
					break nodeloop
				}
			}
		}
		if !foundPlace {
			return fmt.Errorf("failed to find place for %s/%s", pod.Namespace, pod.Name)
		}
	}
	return nil
}