Free cookie consent management tool by TermsFeed Policy Generator

source: branches/GP-MoveOperators/HeuristicLab.Services.Hive/3.3/Manager/HeartbeatManager.cs @ 12075

Last change on this file since 12075 was 7723, checked in by ascheibe, 13 years ago

#1722 minor improvement of the messages the hive server logs

File size: 7.1 KB
Line 
1#region License Information
2/* HeuristicLab
3 * Copyright (C) 2002-2012 Heuristic and Evolutionary Algorithms Laboratory (HEAL)
4 *
5 * This file is part of HeuristicLab.
6 *
7 * HeuristicLab is free software: you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation, either version 3 of the License, or
10 * (at your option) any later version.
11 *
12 * HeuristicLab is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with HeuristicLab. If not, see <http://www.gnu.org/licenses/>.
19 */
20#endregion
21
22using System;
23using System.Collections.Generic;
24using System.Linq;
25using HeuristicLab.Services.Hive.DataTransfer;
26using DA = HeuristicLab.Services.Hive.DataAccess;
27
28namespace HeuristicLab.Services.Hive {
29  public class HeartbeatManager {
30    private IHiveDao dao {
31      get { return ServiceLocator.Instance.HiveDao; }
32    }
33    private IAuthorizationManager auth {
34      get { return ServiceLocator.Instance.AuthorizationManager; }
35    }
36
37    /// <summary>
38    /// This method will be called every time a slave sends a heartbeat (-> very often; concurrency is important!)
39    /// </summary>
40    /// <returns>a list of actions the slave should do</returns>
41    public List<MessageContainer> ProcessHeartbeat(Heartbeat heartbeat) {
42      List<MessageContainer> actions = new List<MessageContainer>();
43      Slave slave = dao.GetSlave(heartbeat.SlaveId);
44      if (slave == null) {
45        actions.Add(new MessageContainer(MessageContainer.MessageType.SayHello));
46      } else {
47        if (heartbeat.HbInterval != slave.HbInterval) {
48          actions.Add(new MessageContainer(MessageContainer.MessageType.NewHBInterval));
49        }
50
51        // update slave data
52        slave.FreeCores = heartbeat.FreeCores;
53        slave.FreeMemory = heartbeat.FreeMemory;
54        slave.CpuUtilization = heartbeat.CpuUtilization;
55        slave.IsAllowedToCalculate = SlaveIsAllowedToCalculate(slave.Id);
56        slave.SlaveState = (heartbeat.JobProgress != null && heartbeat.JobProgress.Count > 0) ? SlaveState.Calculating : SlaveState.Idle;
57        slave.LastHeartbeat = DateTime.Now;
58        dao.UpdateSlave(slave);
59
60        // update task data
61        actions.AddRange(UpdateTasks(heartbeat, slave.IsAllowedToCalculate));
62
63        // assign new task
64        if (heartbeat.AssignJob && slave.IsAllowedToCalculate && heartbeat.FreeCores > 0) {
65          var availableJobs = dao.GetWaitingTasks(slave, 1);
66          if (availableJobs.Count() > 0) {
67            var job = availableJobs.First();
68            if (AssignJob(slave, job))
69              actions.Add(new MessageContainer(MessageContainer.MessageType.CalculateTask, job.Id));
70          }
71        }
72      }
73      return actions;
74    }
75
76    // returns true if assignment was successful
77    private bool AssignJob(Slave slave, Task task) {
78      // load task again and check if it is still available (this is an attempt to reduce the race condition which causes multiple heartbeats to get the same task assigned)
79      if (dao.GetTask(task.Id).State != TaskState.Waiting) return false;
80
81      task = dao.UpdateTaskState(task.Id, DataAccess.TaskState.Transferring, slave.Id, null, null);
82
83      // from now on the task has some time to send the next heartbeat (ApplicationConstants.TransferringJobHeartbeatTimeout)
84      task.LastHeartbeat = DateTime.Now;
85      dao.UpdateTask(task);
86      return true;
87    }
88
89    /// <summary>
90    /// Update the progress of each task
91    /// Checks if all the task sent by heartbeat are supposed to be calculated by this slave
92    /// </summary>
93    private IEnumerable<MessageContainer> UpdateTasks(Heartbeat heartbeat, bool IsAllowedToCalculate) {
94      List<MessageContainer> actions = new List<MessageContainer>();
95
96      if (heartbeat.JobProgress == null)
97        return actions;
98
99      if (!IsAllowedToCalculate && heartbeat.JobProgress.Count != 0) {
100        actions.Add(new MessageContainer(MessageContainer.MessageType.PauseAll));
101      } else {
102        // process the jobProgresses
103        foreach (var jobProgress in heartbeat.JobProgress) {
104          Task curTask = dao.GetTask(jobProgress.Key);
105          if (curTask == null) {
106            // task does not exist in db
107            actions.Add(new MessageContainer(MessageContainer.MessageType.AbortTask, jobProgress.Key));
108            DA.LogFactory.GetLogger(this.GetType().Namespace).Log("Task on slave " + heartbeat.SlaveId + " does not exist in DB: " + jobProgress.Key);
109          } else {
110            if (curTask.CurrentStateLog.SlaveId == Guid.Empty || curTask.CurrentStateLog.SlaveId != heartbeat.SlaveId) {
111              // assigned slave does not match heartbeat
112              actions.Add(new MessageContainer(MessageContainer.MessageType.AbortTask, curTask.Id));
113              DA.LogFactory.GetLogger(this.GetType().Namespace).Log("The slave " + heartbeat.SlaveId + " is not supposed to calculate task: " + curTask);
114            } else if (!TaskIsAllowedToBeCalculatedBySlave(heartbeat.SlaveId, curTask)) {
115              // assigned resources ids of task do not match with slaveId (and parent resourceGroupIds); this might happen when slave is moved to different group
116              actions.Add(new MessageContainer(MessageContainer.MessageType.PauseTask, curTask.Id));
117            } else {
118              // save task execution time
119              curTask.ExecutionTime = jobProgress.Value;
120              curTask.LastHeartbeat = DateTime.Now;
121
122              switch (curTask.Command) {
123                case Command.Stop:
124                  actions.Add(new MessageContainer(MessageContainer.MessageType.StopTask, curTask.Id));
125                  break;
126                case Command.Pause:
127                  actions.Add(new MessageContainer(MessageContainer.MessageType.PauseTask, curTask.Id));
128                  break;
129                case Command.Abort:
130                  actions.Add(new MessageContainer(MessageContainer.MessageType.AbortTask, curTask.Id));
131                  break;
132              }
133              dao.UpdateTask(curTask);
134            }
135          }
136        }
137      }
138      return actions;
139    }
140
141    private bool TaskIsAllowedToBeCalculatedBySlave(Guid slaveId, Task curTask) {
142      var assignedResourceIds = dao.GetAssignedResources(curTask.Id).Select(x => x.Id);
143      var slaveResourceIds = dao.GetParentResources(slaveId).Select(x => x.Id);
144      return assignedResourceIds.Any(x => slaveResourceIds.Contains(x));
145    }
146
147    private bool SlaveIsAllowedToCalculate(Guid slaveId) {
148      // the slave may only calculate if there is no downtime right now. this needs to be checked for every parent resource also
149      return dao.GetParentResources(slaveId).All(r => dao.GetDowntimes(x => x.ResourceId == r.Id && (DateTime.Now >= x.StartDate) && (DateTime.Now <= x.EndDate)).Count() == 0);
150    }
151  }
152}
Note: See TracBrowser for help on using the repository browser.