svn commit: r462911 - in /lucene/hadoop/trunk: CHANGES.txt src/java/org/apache/hadoop/mapred/Task.java

[cu...@apache.org]

Author: cutting
Date: Wed Oct 11 12:32:47 2006
New Revision: 462911

URL: http://svn.apache.org/viewvc?view=rev&rev=462911
Log:
HADOOP-598.  Fix tasks to retry when reporting completion.  Contributed by Owen.

Modified:
    lucene/hadoop/trunk/CHANGES.txt
    lucene/hadoop/trunk/src/java/org/apache/hadoop/mapred/Task.java

Modified: lucene/hadoop/trunk/CHANGES.txt
URL: http://svn.apache.org/viewvc/lucene/hadoop/trunk/CHANGES.txt?view=diff&rev=462911&r1=462910&r2=462911
==============================================================================
--- lucene/hadoop/trunk/CHANGES.txt (original)
+++ lucene/hadoop/trunk/CHANGES.txt Wed Oct 11 12:32:47 2006
@@ -13,6 +13,9 @@
     .999, so that nearly all blocks must be reported before filesystem
     modifications are permitted.  (Konstantin Shvachko via cutting)
 
+ 4. HADOOP-598.  Fix tasks to retry when reporting completion, so that
+    a single RPC timeout won't fail a task.  (omalley via cutting)
+
 
 Release 0.7.0 - 2006-10-06
 

Modified: lucene/hadoop/trunk/src/java/org/apache/hadoop/mapred/Task.java
URL: http://svn.apache.org/viewvc/lucene/hadoop/trunk/src/java/org/apache/hadoop/mapred/Task.java?view=diff&rev=462911&r1=462910&r2=462911
==============================================================================
--- lucene/hadoop/trunk/src/java/org/apache/hadoop/mapred/Task.java (original)
+++ lucene/hadoop/trunk/src/java/org/apache/hadoop/mapred/Task.java Wed Oct 11 12:32:47 2006
@@ -176,10 +176,26 @@
     }
   }
 
-  public void done(TaskUmbilicalProtocol umbilical)
-    throws IOException {
-    umbilical.progress(getTaskId(),               // send a final status report
-                       taskProgress.get(), taskProgress.toString(), phase);
-    umbilical.done(getTaskId());
+  public void done(TaskUmbilicalProtocol umbilical) throws IOException {
+    int retries = 10;
+    boolean needProgress = true;
+    while (true) {
+      try {
+        if (needProgress) {
+          // send a final status report
+          umbilical.progress(getTaskId(), taskProgress.get(), 
+                             taskProgress.toString(), phase);
+          needProgress = false;
+        }
+        umbilical.done(getTaskId());
+        return;
+      } catch (IOException ie) {
+        LOG.warn("Failure signalling completion: " + 
+                 StringUtils.stringifyException(ie));
+        if (--retries == 0) {
+          throw ie;
+        }
+      }
+    }
   }
 }