storage-plugins.go

/*
   Copyright @ 2021 bocloud <fushaosong@beyondcent.com>.

   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 localstorage

import (
	"context"
	"errors"
	"fmt"
	"github.com/bocloud/carina/scheduler/configuration"
	"github.com/bocloud/carina/scheduler/utils"
	v1 "k8s.io/api/core/v1"
	"k8s.io/apimachinery/pkg/runtime"
	lcorev1 "k8s.io/client-go/listers/core/v1"
	lstoragev1 "k8s.io/client-go/listers/storage/v1"
	"k8s.io/klog/v2"
	"k8s.io/kubernetes/pkg/scheduler/framework"
	"sort"
	"strconv"
	"strings"
)

// 插件名称
const Name = "local-storage"
const undefined = "undefined"

type LocalStorage struct {
	handle    framework.Handle
	scLister  lstoragev1.StorageClassLister
	pvcLister lcorev1.PersistentVolumeClaimLister
	pvLister  lcorev1.PersistentVolumeLister
}

var _ framework.FilterPlugin = &LocalStorage{}
var _ framework.ScorePlugin = &LocalStorage{}

//type PluginFactory = func(configuration *runtime.Unknown, f FrameworkHandle) (Plugin, error)
func New(_ runtime.Object, handle framework.Handle) (framework.Plugin, error) {
	scLister := handle.SharedInformerFactory().Storage().V1().StorageClasses().Lister()
	pvcLister := handle.SharedInformerFactory().Core().V1().PersistentVolumeClaims().Lister()
	pvLister := handle.SharedInformerFactory().Core().V1().PersistentVolumes().Lister()
	return &LocalStorage{
		handle:    handle,
		pvcLister: pvcLister,
		scLister:  scLister,
		pvLister:  pvLister,
	}, nil
}

func (ls *LocalStorage) Name() string {
	return Name
}

// 过滤掉不符合当前 Pod 运行条件的Node（相当于旧版本的 predicate）
func (ls *LocalStorage) Filter(ctx context.Context, cycleState *framework.CycleState, pod *v1.Pod, node *framework.NodeInfo) *framework.Status {

	klog.V(3).Infof("filter pod: %v, node: %v", pod.Name, node.Node().Name)

	pvcMap, nodeName, cacheDeviceRequest, err := ls.getLocalStoragePvc(pod)
	if err != nil {
		klog.V(3).ErrorS(err, "get pvc sc failed pod: %v, node: %v", pod.Name, node.Node().Name)
		return framework.NewStatus(framework.Error, "get pv/sc resource error")
	}
	if nodeName != "" && nodeName != node.Node().Name {
		klog.V(3).Infof("mismatch pod: %v, node: %v", pod.Name, node.Node().Name)
		return framework.NewStatus(framework.UnschedulableAndUnresolvable, "pv node mismatch")
	}
	if len(pvcMap) == 0 {
		return framework.NewStatus(framework.Success, "")
	}

	capacityMap := map[string]int64{}
	total := int64(0)
	for key, v := range node.Node().Status.Allocatable {
		if strings.HasPrefix(string(key), utils.DeviceCapacityKeyPrefix) {
			capacityMap[string(key)] = v.Value()
			total += v.Value()
		}
	}

	// 检查节点容量是否充足
	for key, pvs := range pvcMap {
		sort.Slice(pvs, func(i, j int) bool {
			return pvs[i].Spec.Resources.Requests.Storage().Value() > pvs[j].Spec.Resources.Requests.Storage().Value()
		})

		// 对于sc中未设置Device组处理比较复杂,需要判断在多个Device组的情况下，pv是否能够分配
		// 如carina-vg-hdd 20G carina-vg-ssd 40G, pv1.reques
		// t30 pv2.request.15 pv3.request 6G
		// 我们这里不能采取最优分配算法，应该采用贪婪算法，因为我们CSI控制器对PV的创建是逐个进行的，它没有全局视图
		// 即便如此，由于创建PV是由csi-provisioner发起的，请求顺序不确有可能导致pv不合理分配，所以建议sc设置Device组
		// 正因为如此，按照最小满足开始过滤.
		if key == undefined {
			capacityList := []int64{}
			for _, c := range capacityMap {
				capacityList = append(capacityList, c)
			}
			for _, pv := range pvs {
				requestBytes := pv.Spec.Resources.Requests.Storage().Value()
				requestGb := (requestBytes-1)>>30 + 1
				capacityList = minimumValueMinus(capacityList, requestGb)
				if len(capacityList) == 0 {
					klog.V(3).Infof("mismatch pod: %v, node: %v", pod.Name, node.Node().Name)
					return framework.NewStatus(framework.UnschedulableAndUnresolvable, "node storage resource insufficient")
				}
			}
		} else {
			requestTotalBytes := int64(0)
			for _, pv := range pvs {
				requestTotalBytes += pv.Spec.Resources.Requests.Storage().Value()
			}
			requestTotalGb := (requestTotalBytes-1)>>30 + 1
			// add cache device request
			if value, ok := cacheDeviceRequest[key]; ok {
				requestTotalGb += (value-1)>>30 + 1
			}
			if requestTotalGb > capacityMap[key] {
				klog.V(3).Infof("mismatch pod: %v, node: %v, request: %d, capacity: %d", pod.Name, node.Node().Name, requestTotalGb, capacityMap[key])
				return framework.NewStatus(framework.UnschedulableAndUnresolvable, "node storage resource insufficient")
			}
		}
	}

	// check cache device request
	for key, value := range cacheDeviceRequest {
		requestTotalGb := (value-1)>>30 + 1
		if requestTotalGb > capacityMap[key] {
			klog.V(3).Infof("mismatch pod: %v, node: %v, request: %d, capacity: %d", pod.Name, node.Node().Name, requestTotalGb, capacityMap[key])
			return framework.NewStatus(framework.UnschedulableAndUnresolvable, "node cache storage resource insufficient")
		}
	}

	klog.V(3).Infof("filter success pod: %v, node: %v", pod.Name, node.Node().Name)
	return framework.NewStatus(framework.Success, "")
}

// 对节点进行打分（相当于旧版本的 priorities）
func (ls *LocalStorage) Score(ctx context.Context, state *framework.CycleState, pod *v1.Pod, nodeName string) (int64, *framework.Status) {
	klog.V(3).Infof("score pod: %v, node: %v", pod.Name, nodeName)
	pvcMap, node, _, _ := ls.getLocalStoragePvc(pod)
	if node == nodeName {
		return 10, framework.NewStatus(framework.Success)
	}

	if len(pvcMap) == 0 {
		return 5, framework.NewStatus(framework.Success, "")
	}

	// Get Node Info
	// 节点信息快照在执行调度时创建，并在在整个调度周期内不变
	nodeInfo, err := ls.handle.SnapshotSharedLister().NodeInfos().Get(nodeName)
	if err != nil {
		return 0, framework.NewStatus(framework.Error, fmt.Sprintf("getting node %q from Snapshot: %v", nodeName, err))
	}

	capacityMap := map[string]int64{}
	total := int64(0)
	for key, v := range nodeInfo.Node().Status.Allocatable {
		if strings.HasPrefix(string(key), utils.DeviceCapacityKeyPrefix) {
			capacityMap[string(key)] = v.Value()
			total += v.Value()
		}
	}
	var score int64
	// 计算节点分数
	// 影响磁盘分数的有磁盘容量,磁盘上现有pv数量,磁盘IO
	// 在此我们以磁盘容量作为标准，同时配合配置文件中磁盘选择策略
	for key, pvs := range pvcMap {
		sort.Slice(pvs, func(i, j int) bool {
			return pvs[i].Spec.Resources.Requests.Storage().Value() > pvs[j].Spec.Resources.Requests.Storage().Value()
		})
		if key == undefined {
			capacityList := []int64{}
			for _, c := range capacityMap {
				capacityList = append(capacityList, c)
			}
			for _, pv := range pvs {
				requestBytes := pv.Spec.Resources.Requests.Storage().Value()
				requestGb := (requestBytes-1)>>30 + 1
				capacityList = minimumValueMinus(capacityList, requestGb)
				if len(capacityList) > 0 {
					score += 1
				}
			}
		} else {
			requestTotalBytes := int64(0)
			for _, pv := range pvs {
				requestTotalBytes += pv.Spec.Resources.Requests.Storage().Value()
			}
			requestTotalGb := (requestTotalBytes-1)>>30 + 1
			ratio := int64(capacityMap[key] / requestTotalGb)

			if configuration.SchedulerStrategy() == configuration.SchedulerSpradout {
				score = reasonableScore(ratio)
			}
			if configuration.SchedulerStrategy() == configuration.SchedulerBinpack {
				score = 6 - reasonableScore(ratio)
			}
		}
	}
	klog.V(3).Infof("score pod: %v, node: %v score %v", pod.Name, nodeName, score)
	return score, framework.NewStatus(framework.Success)
}

// ScoreExtensions of the Score plugin.
func (ls *LocalStorage) ScoreExtensions() framework.ScoreExtensions {
	return nil
}

func (ls *LocalStorage) getLocalStoragePvc(pod *v1.Pod) (map[string][]*v1.PersistentVolumeClaim, string, map[string]int64, error) {
	nodeName := ""
	localPvc := map[string][]*v1.PersistentVolumeClaim{}
	cacheDeviceRequest := map[string]int64{}
	for _, vol := range pod.Spec.Volumes {
		if vol.PersistentVolumeClaim == nil {
			continue
		}
		pvcName := vol.PersistentVolumeClaim.ClaimName

		pvc, err := ls.pvcLister.PersistentVolumeClaims(pod.Namespace).Get(pvcName)
		if err != nil {
			return localPvc, nodeName, cacheDeviceRequest, err
		}

		if pvc.Spec.StorageClassName == nil {
			continue
		}

		sc, err := ls.scLister.Get(*pvc.Spec.StorageClassName)
		if err != nil {
			return localPvc, nodeName, cacheDeviceRequest, err
		}
		if sc.Provisioner != utils.CSIPluginName {
			continue
		}

		// 如果存在某个pv已经绑定到某个节点，则新创建对pv只能在该节点创建
		if pvc.Status.Phase == v1.ClaimBound {
			pv, err := ls.pvLister.Get(pvc.Spec.VolumeName)
			if err != nil {
				return localPvc, nodeName, cacheDeviceRequest, err
			}
			if nodeName == "" {
				nodeName = pv.Spec.CSI.VolumeAttributes[utils.VolumeDeviceNode]
			} else if nodeName != pv.Spec.CSI.VolumeAttributes[utils.VolumeDeviceNode] {
				return localPvc, nodeName, cacheDeviceRequest, errors.New("pvc node clash")
			}
			continue
		}

		deviceGroup := sc.Parameters[utils.DeviceDiskKey]
		// if bcache device
		if deviceGroup == "" {
			deviceGroup = sc.Parameters[utils.VolumeBackendDiskType]
		}

		cacheGroup := sc.Parameters[utils.VolumeCacheDiskType]
		if cacheGroup != "" {
			cacheGroup = utils.DeviceCapacityKeyPrefix + "carina-vg-" + cacheGroup
			cacheDiskRatio := sc.Parameters[utils.VolumeCacheDiskRatio]
			ratio, err := strconv.ParseInt(cacheDiskRatio, 10, 64)
			if err != nil {
				return localPvc, nodeName, cacheDeviceRequest, errors.New("carina.storage.io/cache-disk-ratio, Should be in 1-100")
			}
			if ratio < 1 || ratio >= 100 {
				return localPvc, nodeName, cacheDeviceRequest, errors.New("carina.storage.io/cache-disk-ratio, Should be in 1-100")
			}
			cacheRequestBytes := pvc.Spec.Resources.Requests.Storage().Value() * ratio / 100
			cacheDeviceRequest[cacheGroup] += cacheRequestBytes
		}

		if deviceGroup == "" {
			// sc中未设置device group
			deviceGroup = undefined
		} else {
			deviceGroup = utils.DeviceCapacityKeyPrefix + "carina-vg-" + deviceGroup
		}
		localPvc[deviceGroup] = append(localPvc[deviceGroup], pvc)
	}
	return localPvc, nodeName, cacheDeviceRequest, nil
}

// 在所有容量列表中，找到最低满足的值，并减去请求容量
// 循环便能判断该节点是否可满足所有pvc请求容量
func minimumValueMinus(array []int64, value int64) []int64 {
	sort.Slice(array, func(i, j int) bool {
		return array[i] < array[j]
	})
	index := -1
	for i, a := range array {
		if a >= value {
			index = i
			break
		}
	}
	if index < 0 {
		return []int64{}
	}
	array[index] = array[index] - value
	return array
}

// 分值范围为0-10，在此降低pv分值比例限制为1-5分
// 考虑到扩容以及提高资源利用率方面，进行中性的评分
// 对于申请用量与现存容量差距巨大，则配置文件中选节点策略可以忽略
func reasonableScore(ratio int64) int64 {
	if ratio > 10 {
		return 5
	}
	if ratio < 2 {
		return 1
	}
	return ratio / 2
}