Fix issue azkaban#1559: Error writing out logs for job on a retry att…

…empt
juhoautio · Nov 14, 2017 · ae43931 · ae43931
1 parent 089a755
commit ae43931
Show file tree

Hide file tree

Showing 2 changed files with 13 additions and 7 deletions.
diff --git a/azkaban-common/src/main/java/azkaban/executor/ExecutableNode.java b/azkaban-common/src/main/java/azkaban/executor/ExecutableNode.java
@@ -63,7 +63,7 @@ public class ExecutableNode {
   private Set<String> outNodes = new HashSet<>();
   private Props inputProps;
   private Props outputProps;
-  private int attempt = 0;
+  private volatile int attempt = 0;
   private long delayExecution = 0;
   private ArrayList<ExecutionAttempt> pastAttempts = null;
 
@@ -235,7 +235,9 @@ public void setAttempt(final int attempt) {
 
   public void resetForRetry() {
     final ExecutionAttempt pastAttempt = new ExecutionAttempt(this.attempt, this);
-    this.attempt++;
+    synchronized (this) {
+      this.attempt++;
+    }
 
     synchronized (this) {
       if (this.pastAttempts == null) {

diff --git a/azkaban-exec-server/src/main/java/azkaban/execapp/JobRunner.java b/azkaban-exec-server/src/main/java/azkaban/execapp/JobRunner.java
@@ -609,11 +609,15 @@ private void doRun() {
         "Finishing job " + this.jobId + getNodeRetryLog() + " at " + this.node.getEndTime()
             + " with status " + this.node.getStatus());
 
-    fireEvent(Event.create(this, EventType.JOB_FINISHED,
-        new EventData(finalStatus, this.node.getNestedId())), false);
-    finalizeLogFile(this.node.getAttempt());
-    finalizeAttachmentFile();
-    writeStatus();
+    try {
+      finalizeLogFile(this.node.getAttempt());
+      finalizeAttachmentFile();
+      writeStatus();
+    } finally {
+      // note that FlowRunner thread does node.attempt++ when it receives the JOB_FINISHED event
+      fireEvent(Event.create(this, EventType.JOB_FINISHED,
+          new EventData(finalStatus, this.node.getNestedId())), false);
+    }
   }
 
   private String getNodeRetryLog() {