Free cookie consent management tool by TermsFeed Policy Generator

source: branches/gteufl/HeuristicLab.Services.Hive/3.3/Manager/EventManager.cs @ 10786

Last change on this file since 10786 was 9665, checked in by ascheibe, 11 years ago

#2030 merged hive performance branch back into trunk

File size: 5.3 KB
Line 
1#region License Information
2/* HeuristicLab
3 * Copyright (C) 2002-2013 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.DataAccess;
26using DT = HeuristicLab.Services.Hive.DataTransfer;
27
28
29namespace HeuristicLab.Services.Hive {
30  /// <summary>
31  /// This class offers methods for cleaning up offline slaves and task
32  /// </summary>
33  public class EventManager : IEventManager {
34    private IHiveDao dao {
35      get { return ServiceLocator.Instance.HiveDao; }
36    }
37    private IAuthorizationManager auth {
38      get { return ServiceLocator.Instance.AuthorizationManager; }
39    }
40    private ILogger log {
41      get { return LogFactory.GetLogger(this.GetType().Namespace); }
42    }
43    private DataAccess.ITransactionManager trans {
44      get { return ServiceLocator.Instance.TransactionManager; }
45    }
46
47    public void Cleanup() {
48      trans.UseTransaction(() => {
49        SetTimeoutSlavesOffline();
50        SetTimeoutTasksWaiting();
51        DeleteObsoleteSlaves();
52      });
53
54      trans.UseTransaction(() => {
55        FinishParentTasks();
56        UpdateStatistics();
57      });
58    }
59
60    private void UpdateStatistics() {
61      var slaves = dao.GetSlaves(x => x.SlaveState == SlaveState.Calculating || x.SlaveState == SlaveState.Idle);
62
63      var stats = new DataTransfer.Statistics();
64      stats.TimeStamp = DateTime.Now;
65      var slaveStats = new List<DT.SlaveStatistics>();
66      foreach (var slave in slaves) {
67        slaveStats.Add(new DT.SlaveStatistics() {
68          SlaveId = slave.Id,
69          Cores = slave.Cores.HasValue ? slave.Cores.Value : 0,
70          FreeCores = slave.FreeCores.HasValue ? slave.FreeCores.Value : 0,
71          Memory = slave.Memory.HasValue ? slave.Memory.Value : 0,
72          FreeMemory = slave.FreeMemory.HasValue ? slave.FreeMemory.Value : 0,
73          CpuUtilization = slave.CpuUtilization
74        });
75      }
76      stats.SlaveStatistics = slaveStats;
77      //collecting user statistics slows down the db and results in timeouts.
78      //we have to find another way to deal with this. 
79      //until then the next line is commented out...
80      //stats.UserStatistics = dtoDao.GetUserStatistics();
81      dao.AddStatistics(stats);
82    }
83
84    /// <summary>
85    /// Searches for slaves which are timed out, puts them and their task offline
86    /// </summary>
87    private void SetTimeoutSlavesOffline() {
88      var slaves = dao.GetSlaves(x => x.SlaveState != SlaveState.Offline);
89      foreach (DT.Slave slave in slaves) {
90        if (!slave.LastHeartbeat.HasValue || (DateTime.Now - slave.LastHeartbeat.Value) > HeuristicLab.Services.Hive.Properties.Settings.Default.SlaveHeartbeatTimeout) {
91          slave.SlaveState = DT.SlaveState.Offline;
92          dao.UpdateSlave(slave);
93        }
94      }
95    }
96
97    /// <summary>
98    /// Looks for parent tasks which have FinishWhenChildJobsFinished and set their state to finished
99    /// </summary>
100    private void FinishParentTasks() {
101      var parentTasksToFinish = dao.GetParentTasks(dao.GetResources(x => true).Select(x => x.Id), 0, true);
102      foreach (var task in parentTasksToFinish) {
103        dao.UpdateTaskState(task.Id, TaskState.Finished, null, null, string.Empty);
104      }
105    }
106
107    /// <summary>
108    /// Looks for task which have not sent heartbeats for some time and reschedules them for calculation
109    /// </summary>
110    private void SetTimeoutTasksWaiting() {
111      var tasks = dao.GetTasks(x => (x.State == TaskState.Calculating && (DateTime.Now - x.LastHeartbeat) > HeuristicLab.Services.Hive.Properties.Settings.Default.CalculatingJobHeartbeatTimeout)
112                               || (x.State == TaskState.Transferring && (DateTime.Now - x.LastHeartbeat) > HeuristicLab.Services.Hive.Properties.Settings.Default.TransferringJobHeartbeatTimeout));
113      foreach (var j in tasks) {
114        DT.Task task = dao.UpdateTaskState(j.Id, TaskState.Waiting, null, null, "Slave timed out.");
115        task.Command = null;
116        dao.UpdateTask(task);
117      }
118    }
119
120    /// <summary>
121    /// Searches for slaves that are disposable and deletes them if they were offline for too long
122    /// </summary>
123    private void DeleteObsoleteSlaves() {
124      var slaves = dao.GetSlaves(x => x.IsDisposable.GetValueOrDefault() && x.SlaveState == SlaveState.Offline && (DateTime.Now - x.LastHeartbeat) > HeuristicLab.Services.Hive.Properties.Settings.Default.SweepInterval);
125      foreach (DT.Slave slave in slaves) dao.DeleteSlave(slave.Id);
126    }
127  }
128}
Note: See TracBrowser for help on using the repository browser.