pcm-coordinator/api/internal/scheduler/scheduler.go

/*

 Copyright (c) [2023] [pcm]
 [pcm-coordinator] is licensed under Mulan PSL v2.
 You can use this software according to the terms and conditions of the Mulan PSL v2.
 You may obtain a copy of Mulan PSL v2 at:
          http://license.coscl.org.cn/MulanPSL2
 THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND,
 EITHER EXPaRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT,
 MERCHANTABILITY OR FIT FOR A PARTICULAR PURPOSE.
 See the Mulan PSL v2 for more details.

*/

package scheduler

import (
	"encoding/json"
	"github.com/pkg/errors"
	"github.com/zeromicro/go-zero/core/logx"
	"gitlink.org.cn/jcce-pcm/pcm-coordinator/api/internal/scheduler/common"
	"gitlink.org.cn/jcce-pcm/pcm-coordinator/api/internal/scheduler/database"
	"gitlink.org.cn/jcce-pcm/pcm-coordinator/api/internal/scheduler/service/collector"
	"gitlink.org.cn/jcce-pcm/pcm-coordinator/api/internal/scheduler/service/executor"
	"gitlink.org.cn/jcce-pcm/pcm-coordinator/api/pkg/response"
	"gitlink.org.cn/jcce-pcm/pcm-coordinator/rpc/client/participantservice"
	"gorm.io/gorm"
	"sigs.k8s.io/yaml"
	"strings"
)

type Scheduler struct {
	task              *response.TaskInfo
	participantIds    []int64
	subSchedule       common.SubSchedule
	dbEngin           *gorm.DB
	result            []string //pID:子任务yamlstring 键值对
	participantRpc    participantservice.ParticipantService
	ResourceCollector *map[string]collector.ResourceCollector
	Storages          database.Storage
	AiExecutor        *map[string]executor.AiExecutor
}

func NewScheduler(subSchedule common.SubSchedule, val string, dbEngin *gorm.DB, participantRpc participantservice.ParticipantService) (*Scheduler, error) {
	var task *response.TaskInfo
	err := json.Unmarshal([]byte(val), &task)
	if err != nil {
		return nil, errors.New("create scheduler failed : " + err.Error())
	}
	return &Scheduler{task: task, subSchedule: subSchedule, dbEngin: dbEngin, participantRpc: participantRpc}, nil
}

func NewScheduler2(resourceCollector *map[string]collector.ResourceCollector, storages database.Storage, aiExecutor *map[string]executor.AiExecutor) *Scheduler {
	return &Scheduler{ResourceCollector: resourceCollector, Storages: storages, AiExecutor: aiExecutor}
}

func (s *Scheduler) SpecifyClusters() {
	// 如果已指定集群名，通过数据库查询后返回p端ip列表
	if len(s.task.Clusters) != 0 {
		s.dbEngin.Raw("select id from sc_participant_phy_info where `name` in (?)", s.task.Clusters).Scan(&s.participantIds)
		return
	}
}

func (s *Scheduler) SpecifyNsID() {
	// 未指定集群名，只指定nsID
	if len(s.task.Clusters) == 0 {
		if len(s.task.NsID) != 0 {
			var clusters string
			s.dbEngin.Raw("select clusters from sc_tenant_info where `tenant_name` = ?", s.task.NsID).Scan(&clusters)

			clusterArr := strings.Split(clusters, ",")

			s.dbEngin.Raw("select id from sc_participant_phy_info where `name` in (?)", clusterArr).Scan(&s.participantIds)
		}
	} else {
		return
	}
}

func (s *Scheduler) MatchLabels() {

	var ids []int64
	count := 0
	// 集群和nsID都未指定，则通过标签匹配
	if len(s.task.Clusters) == 0 && len(s.task.NsID) == 0 {
		//如果集群列表或nsID均未指定
		for key := range s.task.MatchLabels {
			var participantIds []int64
			s.dbEngin.Raw("select participant_id from sc_participant_label_info where `key` = ? and value = ?", key, s.task.MatchLabels[key]).Scan(&participantIds)
			if count == 0 {
				ids = participantIds
			}
			ids = common.Intersect(ids, participantIds)
			count++
		}
		s.participantIds = ids
	} else {
		return
	}
}

// TempAssign todo 屏蔽原调度算法
func (s *Scheduler) TempAssign() error {

	//需要判断task中的资源类型，针对metadata中的多个kind做不同处理
	//输入副本数和集群列表，最终结果输出为pID对应副本数量列表，针对多个kind需要做拆分和重新拼接组合
	var meData []string
	for _, yamlString := range s.task.Metadata {
		var data map[string]interface{}
		err := yaml.Unmarshal([]byte(yamlString), &data)
		if err != nil {
		}

		jsonData, err := json.Marshal(data)
		if err != nil {
		}
		meData = append(meData, string(jsonData))
	}
	s.task.Metadata = meData
	return nil
}

func (s *Scheduler) AssignAndSchedule(ss common.SubSchedule) error {
	//// 已指定 ParticipantId
	//if s.task.ParticipantId != 0 {
	//	return nil
	//}
	//// 标签匹配以及后，未找到ParticipantIds
	//if len(s.participantIds) == 0 {
	//	return errors.New("未找到匹配的ParticipantIds")
	//}
	//
	//// 指定或者标签匹配的结果只有一个集群，给任务信息指定
	//if len(s.participantIds) == 1 {
	//	s.task.ParticipantId = s.participantIds[0]
	//	//replicas := s.task.Metadata.(map[string]interface{})["spec"].(map[string]interface{})["replicas"].(float64)
	//	//result := make(map[int64]string)
	//	//result[s.participantIds[0]] = strconv.FormatFloat(replicas, 'f', 2, 64)
	//	//s.result = result
	//
	//	return nil
	//}

	strategy, err := ss.PickOptimalStrategy()
	if err != nil {
		return err
	}

	clusters, err := strategy.Schedule()
	if err != nil {
		return err
	}

	//集群数量不满足，指定到标签匹配后第一个集群
	//if len(providerList) < 2 {
	//	s.task.ParticipantId = s.participantIds[0]
	//	return nil
	//}

	err = ss.AssignTask(clusters)
	if err != nil {
		return err
	}

	return nil
}

func (s *Scheduler) SaveToDb() error {

	for _, participantId := range s.participantIds {

		for _, resource := range s.task.Metadata {
			structForDb, err := s.subSchedule.GetNewStructForDb(s.task, resource, participantId)
			if err != nil {
				return err
			}

			tx := s.dbEngin.Create(structForDb)
			if tx.Error != nil {
				logx.Error(tx.Error)
				return tx.Error
			}
		}

	}
	return nil
}