[6371] | 1 | #region License Information
|
---|
| 2 | /* HeuristicLab
|
---|
| 3 | * Copyright (C) 2002-2011 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 |
|
---|
[6203] | 22 | using System;
|
---|
| 23 | using System.IO;
|
---|
| 24 | using System.Threading;
|
---|
[6456] | 25 | using HeuristicLab.Clients.Hive.SlaveCore.Properties;
|
---|
[6357] | 26 | using HeuristicLab.Common;
|
---|
| 27 | using HeuristicLab.Core;
|
---|
| 28 | using HeuristicLab.PluginInfrastructure.Sandboxing;
|
---|
[6203] | 29 |
|
---|
| 30 | namespace HeuristicLab.Clients.Hive.SlaveCore {
|
---|
| 31 |
|
---|
[6371] | 32 | /// <summary>
|
---|
| 33 | /// Manages a single job and it's appdomain.
|
---|
| 34 | /// </summary>
|
---|
[6203] | 35 | public class SlaveJob : MarshalByRefObject {
|
---|
| 36 | private Executor executor;
|
---|
| 37 | private AppDomain appDomain;
|
---|
| 38 | private Semaphore waitForStartBeforeKillSem;
|
---|
| 39 | private bool executorMonitoringRun;
|
---|
| 40 | private Thread executorMonitoringThread;
|
---|
[6357] | 41 | private PluginManager pluginManager;
|
---|
| 42 | private ILog log;
|
---|
| 43 | public Guid JobId { get; private set; }
|
---|
| 44 | public bool IsPrepared { get; private set; }
|
---|
[6203] | 45 |
|
---|
[6357] | 46 | private int coresNeeded;
|
---|
| 47 | public int CoresNeeded {
|
---|
| 48 | get { return coresNeeded; }
|
---|
| 49 | set { this.coresNeeded = value; }
|
---|
| 50 | }
|
---|
[6203] | 51 |
|
---|
[6357] | 52 | public TimeSpan ExecutionTime {
|
---|
| 53 | get { return executor != null ? executor.ExecutionTime : TimeSpan.Zero; }
|
---|
| 54 | }
|
---|
[6203] | 55 |
|
---|
[6357] | 56 | public SlaveJob(PluginManager pluginManager, int coresNeeded, ILog log) {
|
---|
| 57 | this.pluginManager = pluginManager;
|
---|
| 58 | this.coresNeeded = coresNeeded;
|
---|
| 59 | this.log = log;
|
---|
[6203] | 60 | waitForStartBeforeKillSem = new Semaphore(0, 1);
|
---|
| 61 | executorMonitoringRun = true;
|
---|
[6357] | 62 | IsPrepared = false;
|
---|
[6203] | 63 | }
|
---|
| 64 |
|
---|
[6357] | 65 | public void StartJobAsync(Job job, JobData jobData) {
|
---|
| 66 | try {
|
---|
| 67 | this.JobId = job.Id;
|
---|
| 68 | Prepare(job);
|
---|
| 69 | StartJobInAppDomain(jobData);
|
---|
| 70 | }
|
---|
| 71 | catch (Exception) {
|
---|
[6371] | 72 | // make sure to clean up if something went wrong
|
---|
| 73 | DisposeAppDomain();
|
---|
[6357] | 74 | throw;
|
---|
| 75 | }
|
---|
[6216] | 76 | }
|
---|
[6203] | 77 |
|
---|
[6357] | 78 | public void PauseJob() {
|
---|
| 79 | if (!IsPrepared) throw new AppDomainNotCreatedException();
|
---|
| 80 | if (!executor.IsPausing && !executor.IsStopping) executor.Pause();
|
---|
[6203] | 81 | }
|
---|
| 82 |
|
---|
| 83 | public void StopJob() {
|
---|
[6357] | 84 | if (!IsPrepared) throw new AppDomainNotCreatedException();
|
---|
| 85 | if (!executor.IsPausing && !executor.IsStopping) executor.Stop();
|
---|
[6203] | 86 | }
|
---|
| 87 |
|
---|
[6357] | 88 | private void Prepare(Job job) {
|
---|
| 89 | string pluginDir = Path.Combine(pluginManager.PluginTempBaseDir, job.Id.ToString());
|
---|
| 90 | string configFileName;
|
---|
| 91 | pluginManager.PreparePlugins(job, out configFileName);
|
---|
| 92 | appDomain = CreateAppDomain(job, pluginDir, configFileName);
|
---|
| 93 | IsPrepared = true;
|
---|
[6203] | 94 | }
|
---|
| 95 |
|
---|
[6357] | 96 | private AppDomain CreateAppDomain(Job job, String pluginDir, string configFileName) {
|
---|
[6371] | 97 | if (job.IsPrivileged) {
|
---|
| 98 | appDomain = SandboxManager.CreateAndInitPrivilegedSandbox(job.Id.ToString(), pluginDir, Path.Combine(pluginDir, configFileName));
|
---|
| 99 | } else {
|
---|
| 100 | appDomain = SandboxManager.CreateAndInitSandbox(job.Id.ToString(), pluginDir, Path.Combine(pluginDir, configFileName));
|
---|
| 101 | }
|
---|
[6357] | 102 | appDomain.UnhandledException += new UnhandledExceptionEventHandler(AppDomain_UnhandledException);
|
---|
[6203] | 103 |
|
---|
[6357] | 104 | log.LogMessage("Creating AppDomain");
|
---|
| 105 | executor = (Executor)appDomain.CreateInstanceAndUnwrap(typeof(Executor).Assembly.GetName().Name, typeof(Executor).FullName);
|
---|
[6203] | 106 |
|
---|
[6357] | 107 | executor.JobId = job.Id;
|
---|
| 108 | executor.CoresNeeded = job.CoresNeeded;
|
---|
| 109 | executor.MemoryNeeded = job.MemoryNeeded;
|
---|
| 110 | return appDomain;
|
---|
[6203] | 111 | }
|
---|
| 112 |
|
---|
[6357] | 113 | private void StartJobInAppDomain(JobData jobData) {
|
---|
| 114 | executor.Start(jobData.Data);
|
---|
| 115 | waitForStartBeforeKillSem.Release();
|
---|
| 116 | StartExecutorMonitoringThread();
|
---|
[6203] | 117 | }
|
---|
| 118 |
|
---|
[6357] | 119 | public void DisposeAppDomain() {
|
---|
| 120 | log.LogMessage(string.Format("Shutting down Appdomain for Job {0}", JobId));
|
---|
| 121 | StopExecutorMonitoringThread();
|
---|
[6203] | 122 |
|
---|
[6357] | 123 | if (executor != null) {
|
---|
| 124 | executor.Dispose();
|
---|
| 125 | }
|
---|
[6203] | 126 |
|
---|
[6357] | 127 | if (appDomain != null) {
|
---|
| 128 | appDomain.UnhandledException -= new UnhandledExceptionEventHandler(AppDomain_UnhandledException);
|
---|
[6456] | 129 | int repeat = Settings.Default.PluginDeletionRetries;
|
---|
[6357] | 130 | while (repeat > 0) {
|
---|
| 131 | try {
|
---|
| 132 | waitForStartBeforeKillSem.WaitOne();
|
---|
| 133 | AppDomain.Unload(appDomain);
|
---|
| 134 | waitForStartBeforeKillSem.Dispose();
|
---|
| 135 | repeat = 0;
|
---|
| 136 | }
|
---|
| 137 | catch (CannotUnloadAppDomainException) {
|
---|
| 138 | log.LogMessage("Could not unload AppDomain, will try again in 1 sec.");
|
---|
[6456] | 139 | Thread.Sleep(Settings.Default.PluginDeletionTimeout);
|
---|
[6357] | 140 | repeat--;
|
---|
| 141 | if (repeat == 0) {
|
---|
| 142 | throw; // rethrow and let app crash
|
---|
[6203] | 143 | }
|
---|
| 144 | }
|
---|
| 145 | }
|
---|
| 146 | }
|
---|
[6357] | 147 | pluginManager.DeletePluginsForJob(JobId);
|
---|
[6203] | 148 | GC.Collect();
|
---|
| 149 | }
|
---|
| 150 |
|
---|
| 151 | private void AppDomain_UnhandledException(object sender, UnhandledExceptionEventArgs e) {
|
---|
[6357] | 152 | DisposeAppDomain();
|
---|
[6371] | 153 | OnExceptionOccured(new Exception("Unhandled exception: " + e.ExceptionObject.ToString()));
|
---|
[6203] | 154 | }
|
---|
| 155 |
|
---|
[6357] | 156 | public JobData GetJobData() {
|
---|
| 157 | return executor.GetJobData();
|
---|
| 158 | }
|
---|
[6203] | 159 |
|
---|
| 160 | #region ExecutorMonitorThread
|
---|
| 161 | private void StartExecutorMonitoringThread() {
|
---|
| 162 | executorMonitoringThread = new Thread(MonitorExecutor);
|
---|
| 163 | executorMonitoringThread.Start();
|
---|
| 164 | }
|
---|
| 165 |
|
---|
| 166 | private void StopExecutorMonitoringThread() {
|
---|
| 167 | if (executorMonitoringThread != null) {
|
---|
| 168 | if (executorMonitoringRun) {
|
---|
| 169 | executorMonitoringRun = false;
|
---|
[6371] | 170 | executor.ExecutorCommandQueue.AddMessage(ExecutorMessageType.StopExecutorMonitoringThread);
|
---|
[6203] | 171 | }
|
---|
| 172 | }
|
---|
| 173 | }
|
---|
| 174 |
|
---|
| 175 | /// <summary>
|
---|
| 176 | /// Because the executor is in an appdomain and is not able to call back
|
---|
| 177 | /// (because of security -> lease time for marshall-by-ref object is 5 min),
|
---|
| 178 | /// we have to poll the executor for events we have to react to (e.g. job finished...)
|
---|
| 179 | /// </summary>
|
---|
| 180 | private void MonitorExecutor() {
|
---|
| 181 | while (executorMonitoringRun) {
|
---|
| 182 | //this blocks through the appdomain border, that's why the lease gets renewed
|
---|
[6371] | 183 | ExecutorMessage message = executor.ExecutorCommandQueue.GetMessage();
|
---|
[6203] | 184 |
|
---|
| 185 | switch (message.MessageType) {
|
---|
[6357] | 186 | case ExecutorMessageType.JobStarted:
|
---|
| 187 | OnJobStarted();
|
---|
| 188 | break;
|
---|
| 189 |
|
---|
| 190 | case ExecutorMessageType.JobPaused:
|
---|
| 191 | executorMonitoringRun = false;
|
---|
| 192 | OnJobPaused();
|
---|
| 193 | DisposeAppDomain();
|
---|
| 194 | break;
|
---|
| 195 |
|
---|
[6203] | 196 | case ExecutorMessageType.JobStopped:
|
---|
| 197 | executorMonitoringRun = false;
|
---|
[6357] | 198 | OnJobStopped();
|
---|
| 199 | DisposeAppDomain();
|
---|
[6203] | 200 | break;
|
---|
| 201 |
|
---|
| 202 | case ExecutorMessageType.JobFailed:
|
---|
| 203 | executorMonitoringRun = false;
|
---|
[6357] | 204 | OnJobFailed(new JobFailedException(executor.CurrentException));
|
---|
| 205 | DisposeAppDomain();
|
---|
[6203] | 206 | break;
|
---|
| 207 |
|
---|
| 208 | case ExecutorMessageType.StopExecutorMonitoringThread:
|
---|
| 209 | executorMonitoringRun = false;
|
---|
| 210 | return;
|
---|
| 211 | }
|
---|
| 212 | }
|
---|
| 213 | }
|
---|
| 214 | #endregion
|
---|
[6357] | 215 |
|
---|
| 216 | public event EventHandler<EventArgs<Guid>> JobStarted;
|
---|
| 217 | private void OnJobStarted() {
|
---|
| 218 | var handler = JobStarted;
|
---|
| 219 | if (handler != null) handler(this, new EventArgs<Guid>(this.JobId));
|
---|
| 220 | }
|
---|
| 221 |
|
---|
| 222 | public event EventHandler<EventArgs<Guid>> JobStopped;
|
---|
| 223 | private void OnJobStopped() {
|
---|
| 224 | var handler = JobStopped;
|
---|
| 225 | if (handler != null) handler(this, new EventArgs<Guid>(this.JobId));
|
---|
| 226 | }
|
---|
| 227 |
|
---|
| 228 | public event EventHandler<EventArgs<Guid>> JobPaused;
|
---|
| 229 | private void OnJobPaused() {
|
---|
| 230 | var handler = JobPaused;
|
---|
| 231 | if (handler != null) handler(this, new EventArgs<Guid>(this.JobId));
|
---|
| 232 | }
|
---|
| 233 |
|
---|
| 234 | public event EventHandler<EventArgs<Guid>> JobAborted;
|
---|
| 235 | private void OnJobAborted() {
|
---|
| 236 | var handler = JobAborted;
|
---|
| 237 | if (handler != null) handler(this, new EventArgs<Guid>(this.JobId));
|
---|
| 238 | }
|
---|
| 239 |
|
---|
| 240 | public event EventHandler<EventArgs<Guid, Exception>> JobFailed;
|
---|
| 241 | private void OnJobFailed(Exception exception) {
|
---|
| 242 | var handler = JobFailed;
|
---|
| 243 | if (handler != null) handler(this, new EventArgs<Guid, Exception>(this.JobId, exception));
|
---|
| 244 | }
|
---|
| 245 |
|
---|
| 246 | public event EventHandler<EventArgs<Guid, Exception>> ExceptionOccured;
|
---|
| 247 | private void OnExceptionOccured(Exception exception) {
|
---|
| 248 | var handler = ExceptionOccured;
|
---|
| 249 | if (handler != null) handler(this, new EventArgs<Guid, Exception>(this.JobId, exception));
|
---|
| 250 | }
|
---|
[6203] | 251 | }
|
---|
| 252 | }
|
---|