Changeset 5793 for branches/HeuristicLab.Hive-3.4/sources/HeuristicLab.Clients.Hive.Slave/3.4

Timestamp:

03/22/11 15:54:58 (13 years ago)

Author:

cneumuel

Message:

#1233

• implemented correct downloading of paused jobs. its now also possible to change parameters and resume a algorithm
• removed Prepare() calls in ExperimentManager and in slave, as it prevents corrent resuming of paused jobs
• made events in ItemTreeView be invoked in the correct thread
• reduced log output in ExperimentManager

Location:

branches/HeuristicLab.Hive-3.4/sources/HeuristicLab.Clients.Hive.Slave/3.4

Files:

• Core.cs (modified) (2 diffs)
• Executor.cs (modified) (1 diff)

branches/HeuristicLab.Hive-3.4/sources/HeuristicLab.Clients.Hive.Slave/3.4/Core.cs

@@ -159 +159 @@
         switch (container.Message) {
           case MessageContainer.MessageType.CalculateJob:
-            clientCom.LogMessage("Task.StartNew[0]: jobId: " + container.JobId);
             Task.Factory.StartNew((jobIdObj) => {
               Guid jobId = (Guid)jobIdObj;
-              clientCom.LogMessage("Task.StartNew[1]: jobId: " + jobId);
               Job job = wcfService.GetJob(jobId);
               if (job == null) throw new JobNotFoundException(jobId);
@@ -217 +215 @@
 
     private void DoPauseJob(Guid jobId) {
-      Job job = Jobs[jobId];
-
-      if (job != null) {
-        engines[job.Id].Pause();
-        JobData sJob = engines[job.Id].GetPausedJob();
-        // job.Exception = engines[job.Id].CurrentException; // can there be an exception if a job is paused
-        job.ExecutionTime = engines[job.Id].ExecutionTime;
-
-        try {
-          clientCom.LogMessage("Sending the paused job with id: " + job.Id);
-          wcfService.UpdateJobData(job, sJob, ConfigManager.Instance.GetClientInfo().Id, JobState.Paused);
-          SlaveStatusInfo.JobsProcessed++;    //TODO: count or not count, thats the question
-        }
-        catch (Exception e) {
-          clientCom.LogMessage("Transmitting to server failed. Storing the paused job with id: " + job.Id + " to hdd (" + e.ToString() + ")");
-        }
-        finally {
-          KillAppDomain(job.Id); // kill app-domain in every case          
-        }
-      }
-    }
-
-    private void DoStopJob(Guid guid) {
-      Job job = Jobs[guid];
-
-      if (job != null) {
-        engines[job.Id].Stop();
-        JobData sJob = engines[job.Id].GetFinishedJob();
-        // job.Exception = engines[job.Id].CurrentException; // can there be an exception if a job is stopped regularly
-        job.ExecutionTime = engines[job.Id].ExecutionTime;
-
-        try {
-          clientCom.LogMessage("Sending the stoppped job with id: " + job.Id);
-          wcfService.UpdateJobData(job, sJob, ConfigManager.Instance.GetClientInfo().Id, JobState.Paused);
-          SlaveStatusInfo.JobsProcessed++;    //TODO: count or not count, thats the question
-        }
-        catch (Exception e) {
-          clientCom.LogMessage("Transmitting to server failed. Storing the paused job with id: " + job.Id + " to hdd (" + e.ToString() + ")");
-        }
-        finally {
-          KillAppDomain(job.Id); // kill app-domain in every case          
+      if (!Jobs.ContainsKey(jobId)) {
+        clientCom.LogMessage("DoPauseJob: Can't find job with id " + jobId);
+      } else {
+        Job job = Jobs[jobId];
+
+        if (job != null) {
+          engines[job.Id].Pause();
+          JobData sJob = engines[job.Id].GetPausedJob();
+          job.ExecutionTime = engines[job.Id].ExecutionTime;
+
+          try {
+            clientCom.LogMessage("Sending the paused job with id: " + job.Id);
+            wcfService.UpdateJobData(job, sJob, ConfigManager.Instance.GetClientInfo().Id, JobState.Paused);
+            SlaveStatusInfo.JobsProcessed++;    //TODO: count or not count, thats the question
+          }
+          catch (Exception e) {
+            clientCom.LogMessage("Transmitting to server failed. Storing the paused job with id: " + job.Id + " to hdd (" + e.ToString() + ")");
+          }
+          finally {
+            KillAppDomain(job.Id); // kill app-domain in every case          
+          }
+        }
+      }
+    }
+
+    private void DoStopJob(Guid jobId) {
+      if (!Jobs.ContainsKey(jobId)) {
+        clientCom.LogMessage("DoStopJob: Can't find job with id " + jobId);
+      } else {
+        Job job = Jobs[jobId];
+
+        if (job != null) {
+          engines[job.Id].Stop();
+          JobData sJob = engines[job.Id].GetFinishedJob();
+          job.ExecutionTime = engines[job.Id].ExecutionTime;
+
+          try {
+            clientCom.LogMessage("Sending the stoppped job with id: " + job.Id);
+            wcfService.UpdateJobData(job, sJob, ConfigManager.Instance.GetClientInfo().Id, JobState.Finished);
+            SlaveStatusInfo.JobsProcessed++;    //TODO: count or not count, thats the question
+          }
+          catch (Exception e) {
+            clientCom.LogMessage("Transmitting to server failed. Storing the paused job with id: " + job.Id + " to hdd (" + e.ToString() + ")");
+          }
+          finally {
+            KillAppDomain(job.Id); // kill app-domain in every case          
+          }
         }
       }

branches/HeuristicLab.Hive-3.4/sources/HeuristicLab.Clients.Hive.Slave/3.4/Executor.cs

@@ -76 +76 @@
           Job.Resume(childjobs.Select(j => PersistenceUtil.Deserialize<IJob>(j.Data)));
         } else {
-          Job.Prepare();
+          // Job.Prepare(); // do NOT prepare here, otherwise paused jobs get restarted
           Job.Start();
         }