github.com/matrixorigin/matrixone@v0.7.0/pkg/hakeeper/task/task_scheduler.go (about)

     1  // Copyright 2022 Matrix Origin
     2  //
     3  // Licensed under the Apache License, Version 2.0 (the "License");
     4  // you may not use this file except in compliance with the License.
     5  // You may obtain a copy of the License at
     6  //
     7  //      http://www.apache.org/licenses/LICENSE-2.0
     8  //
     9  // Unless required by applicable law or agreed to in writing, software
    10  // distributed under the License is distributed on an "AS IS" BASIS,
    11  // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
    12  // See the License for the specific language governing permissions and
    13  // limitations under the License.
    14  
    15  package task
    16  
    17  import (
    18  	"context"
    19  	"time"
    20  
    21  	"github.com/matrixorigin/matrixone/pkg/common/moerr"
    22  	"github.com/matrixorigin/matrixone/pkg/common/runtime"
    23  	"github.com/matrixorigin/matrixone/pkg/hakeeper"
    24  	"github.com/matrixorigin/matrixone/pkg/pb/logservice"
    25  	"github.com/matrixorigin/matrixone/pkg/pb/task"
    26  	"github.com/matrixorigin/matrixone/pkg/taskservice"
    27  	"go.uber.org/zap"
    28  )
    29  
    30  const (
    31  	taskSchedulerDefaultTimeout = 2 * time.Second
    32  )
    33  
    34  type scheduler struct {
    35  	cfg               hakeeper.Config
    36  	taskServiceGetter func() taskservice.TaskService
    37  }
    38  
    39  var _ hakeeper.TaskScheduler = (*scheduler)(nil)
    40  
    41  func NewScheduler(taskServiceGetter func() taskservice.TaskService, cfg hakeeper.Config) hakeeper.TaskScheduler {
    42  	cfg.Fill()
    43  	s := &scheduler{
    44  		taskServiceGetter: taskServiceGetter,
    45  		cfg:               cfg,
    46  	}
    47  	return s
    48  }
    49  
    50  func (s *scheduler) Schedule(cnState logservice.CNState, currentTick uint64) {
    51  	workingCN, expiredCN := parseCNStores(s.cfg, cnState, currentTick)
    52  
    53  	runningTasks := s.queryTasks(task.TaskStatus_Running)
    54  	createdTasks := s.queryTasks(task.TaskStatus_Created)
    55  	tasks := append(runningTasks, createdTasks...)
    56  	for _, task := range tasks {
    57  		if task.IsInitTask() {
    58  			runtime.ProcessLevelRuntime().
    59  				SubLogger(runtime.SystemInit).Debug(
    60  				"task schedule query init task",
    61  				zap.String("task", task.Metadata.String()))
    62  		}
    63  	}
    64  
    65  	runtime.ProcessLevelRuntime().Logger().Debug("task schedule query tasks", zap.Int("created", len(createdTasks)),
    66  		zap.Int("running", len(runningTasks)))
    67  	if len(tasks) == 0 {
    68  		return
    69  	}
    70  	orderedCN := getCNOrdered(runningTasks, workingCN)
    71  
    72  	s.allocateTasks(createdTasks, orderedCN)
    73  
    74  	expiredTasks := getExpiredTasks(runningTasks, expiredCN)
    75  	s.allocateTasks(expiredTasks, orderedCN)
    76  }
    77  
    78  func (s *scheduler) Create(ctx context.Context, tasks []task.TaskMetadata) error {
    79  	ts := s.taskServiceGetter()
    80  	if ts == nil {
    81  		return moerr.NewInternalError(ctx, "failed to get task service")
    82  	}
    83  	if err := ts.CreateBatch(ctx, tasks); err != nil {
    84  		runtime.ProcessLevelRuntime().Logger().Error("failed to create new tasks", zap.Error(err))
    85  		return err
    86  	}
    87  	runtime.ProcessLevelRuntime().Logger().Debug("new tasks created", zap.Int("created", len(tasks)))
    88  	v, err := ts.QueryTask(ctx)
    89  	if len(v) == 0 && err == nil {
    90  		panic("cannot read created tasks")
    91  	}
    92  	runtime.ProcessLevelRuntime().Logger().Debug("new tasks created, query", zap.Int("created", len(v)), zap.Error(err))
    93  	return nil
    94  }
    95  
    96  func (s *scheduler) StartScheduleCronTask() {
    97  	if ts := s.taskServiceGetter(); ts != nil {
    98  		ts.StartScheduleCronTask()
    99  	}
   100  }
   101  
   102  func (s *scheduler) StopScheduleCronTask() {
   103  	if ts := s.taskServiceGetter(); ts != nil {
   104  		ts.StopScheduleCronTask()
   105  	}
   106  }
   107  
   108  func (s *scheduler) queryTasks(status task.TaskStatus) []task.Task {
   109  	ts := s.taskServiceGetter()
   110  	if ts == nil {
   111  		return nil
   112  	}
   113  	ctx, cancel := context.WithTimeout(context.Background(), taskSchedulerDefaultTimeout)
   114  	defer cancel()
   115  
   116  	tasks, err := ts.QueryTask(ctx, taskservice.WithTaskStatusCond(taskservice.EQ, status))
   117  	if err != nil {
   118  		runtime.ProcessLevelRuntime().Logger().Error("failed to query tasks",
   119  			zap.String("status", status.String()),
   120  			zap.Error(err))
   121  		return nil
   122  	}
   123  	return tasks
   124  }
   125  
   126  func (s *scheduler) allocateTasks(tasks []task.Task, orderedCN *cnMap) {
   127  	ts := s.taskServiceGetter()
   128  	if ts == nil {
   129  		return
   130  	}
   131  
   132  	for _, t := range tasks {
   133  		s.allocateTask(ts, t, orderedCN)
   134  	}
   135  }
   136  
   137  func (s *scheduler) allocateTask(ts taskservice.TaskService, t task.Task, orderedCN *cnMap) {
   138  	runner := orderedCN.min()
   139  	if runner == "" {
   140  		runtime.ProcessLevelRuntime().Logger().Warn("no CN available")
   141  		return
   142  	}
   143  	ctx, cancel := context.WithTimeout(context.Background(), taskSchedulerDefaultTimeout)
   144  	defer cancel()
   145  
   146  	if err := ts.Allocate(ctx, t, runner); err != nil {
   147  		runtime.ProcessLevelRuntime().Logger().Error("failed to allocate task",
   148  			zap.Uint64("task-id", t.ID),
   149  			zap.String("task-metadata-id", t.Metadata.ID),
   150  			zap.String("task-runner", runner),
   151  			zap.Error(err))
   152  		return
   153  	}
   154  	orderedCN.inc(t.TaskRunner)
   155  }
   156  
   157  func getExpiredTasks(tasks []task.Task, expiredCN []string) (expired []task.Task) {
   158  	for _, t := range tasks {
   159  		if contains(expiredCN, t.TaskRunner) {
   160  			expired = append(expired, t)
   161  		}
   162  	}
   163  	return
   164  }
   165  
   166  func getCNOrdered(tasks []task.Task, workingCN []string) *cnMap {
   167  	orderedMap := newOrderedMap(workingCN)
   168  	for _, t := range tasks {
   169  		if contains(workingCN, t.TaskRunner) {
   170  			orderedMap.inc(t.TaskRunner)
   171  		}
   172  	}
   173  
   174  	return orderedMap
   175  }