Changeset 6100

Timestamp:

05/02/11 17:56:52 (14 years ago)

Author:

ascheibe

Message:

#1233

• Executor now sends all exceptions to the ExperimentManager as NetNamedPipe communication won't be possible in a Sandbox due to security constraints
• count stopped and aborted jobs correctly
• send correct status when a job is stopped by the ExperimentManager
• try to log unhandled exceptions to gui if no EventLog is available
• don't crash if job is sent more than once by server

Location:

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

Files:

• Core.cs (modified) (7 diffs)
• Executor.cs (modified) (6 diffs)

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

@@ -96 +96 @@
           catch (Exception) { }
         } else {
-          throw ex;
+          //try to log with clientCom. if this works the user sees at least a message, 
+          //else an exception will be thrown anyways.
+          clientCom.LogMessage("Error on startup: " + ex.ToString() +
+            Environment.NewLine + "Core is going to shutdown.");
         }
       }
@@ -187 +190 @@
             break;
           case MessageContainer.MessageType.AbortJob:
+            SlaveStatusInfo.JobsAborted++;
             KillAppDomain(container.JobId);
             break;
@@ -222 +226 @@
 
           try {
+            if (engines[job.Id].CurrentException != string.Empty) {
+              wcfService.UpdateJobState(job.Id, JobState.Failed, engines[job.Id].CurrentException);
+              SlaveStatusInfo.JobsAborted++;
+            } else {
+              SlaveStatusInfo.JobsProcessed++;
+            }
             clientCom.LogMessage("Sending the paused job with id: " + job.Id);
             wcfService.UpdateJobData(job, sJob, ConfigManager.Instance.GetClientInfo().Id, JobState.Paused);
@@ -246 +256 @@
           job.ExecutionTime = engines[job.Id].ExecutionTime;
 
+
           try {
+            if (engines[job.Id].CurrentException != string.Empty) {
+              wcfService.UpdateJobState(job.Id, JobState.Failed, engines[job.Id].CurrentException);
+            }
+            SlaveStatusInfo.JobsAborted++;
+
             clientCom.LogMessage("Sending the stoppped job with id: " + job.Id);
-            wcfService.UpdateJobData(job, sJob, ConfigManager.Instance.GetClientInfo().Id, JobState.Finished);
+            wcfService.UpdateJobData(job, sJob, ConfigManager.Instance.GetClientInfo().Id, JobState.Aborted);
           }
           catch (Exception e) {
@@ -440 +456 @@
 
       lock (locker) {
-        if (engines.ContainsKey(myJob.Id))
-          throw new Exception("Job with key " + myJob.Id + " already exists");
+        if (engines.ContainsKey(myJob.Id)) {
+          clientCom.LogMessage("Job with key " + myJob.Id + " already exists. Job will be ignored.");
+          return;
+        }
 
         String pluginDir = Path.Combine(PluginCache.Instance.PluginTempBaseDir, myJob.Id.ToString());
@@ -476 +494 @@
               engine.Start(jobData.Data);
             }
+
           }
           catch (Exception exception) {
@@ -544 +563 @@
                 repeat--;
                 if (repeat == 0) {
+                  clientCom.LogMessage("Could not unload AppDomain, shutting down core...");
                   throw; // rethrow and let app crash
                 }

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

@@ -98 +98 @@
       SendHeartbeatForExecutor = false;
       if (Job == null) {
-        SlaveClientCom.Instance.ClientCom.LogMessage("Pausing job: Job is null");
+        currentException = new Exception("Pausing job " + this.JobId + ": Job is null");
         Core.EnqueueExecutorMessage(Core.KillAppDomain, JobId);
       }
@@ -109 +109 @@
         }
         catch (Exception ex) {
-          SlaveClientCom.Instance.ClientCom.LogMessage("Error pausing job:" + ex.ToString());
+          currentException = new Exception("Error pausing job " + this.JobId + ": " + ex.ToString());
         }
       }
@@ -117 +117 @@
       SendHeartbeatForExecutor = false;
       if (Job == null) {
-        SlaveClientCom.Instance.ClientCom.LogMessage("Stopping job: Job is null");
+        currentException = new Exception("Stopping job " + this.JobId + ": Job is null");
         Core.EnqueueExecutorMessage(Core.KillAppDomain, JobId);
       }
@@ -128 +128 @@
         }
         catch (Exception ex) {
-          SlaveClientCom.Instance.ClientCom.LogMessage("Error stopping job:" + ex.ToString());
+          currentException = new Exception("Error stopping job " + this.JobId + ": " + ex.ToString());
         }
       }
@@ -207 +207 @@
     public JobData GetFinishedJob() {
       if (Job == null) {
-        SlaveClientCom.Instance.ClientCom.LogMessage("Getting finished job: Job is null");
+        if (currentException == null) {
+          currentException = new Exception("Getting finished job " + this.JobId + ": Job is null");
+        }
         Core.EnqueueExecutorMessage(Core.KillAppDomain, JobId);
       }
@@ -218 +220 @@
         }
         catch (Exception ex) {
-          SlaveClientCom.Instance.ClientCom.LogMessage("Error stopping job:" + ex.ToString());
+          currentException = new Exception("Error getting finished job " + this.JobId + ": " + ex.ToString());
         }
       }