[GitHub] [hadoop] cnauroth commented on a diff in pull request #4248: MAPREDUCE-7370. Parallelize MultipleOutputs#close call

[GitBox <gi...@apache.org>]

cnauroth commented on code in PR #4248:
URL: https://github.com/apache/hadoop/pull/4248#discussion_r955436533


##########
hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-jobclient/src/test/java/org/apache/hadoop/mapred/lib/TestMultipleOutputs.java:
##########
@@ -70,6 +76,20 @@ public void testWithCounters() throws Exception {
     _testMOWithJavaSerialization(true);
   }
 
+  @Test(expected=IOException.class)
+  public void testParallelClose() throws IOException, InterruptedException {

Review Comment:
   I suggest naming this `testParallelCloseIOException` to make it clear that we are testing an error case.



##########
hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-core/src/main/java/org/apache/hadoop/mapred/lib/MultipleOutputs.java:
##########
@@ -527,9 +558,41 @@ public void collect(Object key, Object value) throws IOException {
    * @throws java.io.IOException thrown if any of the MultipleOutput files
    *                             could not be closed properly.
    */
-  public void close() throws IOException {
+  public void close() throws IOException, InterruptedException {
+    int nThreads = conf.getInt(MRConfig.MULTIPLE_OUTPUTS_CLOSE_THREAD_COUNT,
+        MRConfig.DEFAULT_MULTIPLE_OUTPUTS_CLOSE_THREAD_COUNT);
+    AtomicBoolean encounteredException = new AtomicBoolean(false);
+    ThreadFactory threadFactory = new ThreadFactoryBuilder().setNameFormat("MultipleOutputs-close")
+        .setUncaughtExceptionHandler(

Review Comment:
   `IOException` is now being propagated back to the caller of `close()`, but any unexpected (unchecked) exceptions would be swallowed in this uncaught exception handler. This is different from the existing code, where the caller would receive the unchecked exception.
   
   I think the best we can do here is to set `encounteredException` from within the uncaught exception handler, resulting in throwing the `IOException` at the bottom of the method. The message would need to be generalized to "encountered exception during close", not mentioning `IOException`, because it might also have been some other unchecked exception.



##########
hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-core/src/main/java/org/apache/hadoop/mapred/lib/MultipleOutputs.java:
##########
@@ -528,8 +545,33 @@ public void collect(Object key, Object value) throws IOException {
    *                             could not be closed properly.
    */
   public void close() throws IOException {
+    int nThreads = 10;
+    AtomicReference<IOException> ioException = new AtomicReference<>();
+    ExecutorService executorService = Executors.newFixedThreadPool(nThreads);
+
+    List<Callable<Object>> callableList = new ArrayList<>();
+
     for (RecordWriter writer : recordWriters.values()) {
-      writer.close(null);
+      callableList.add(() -> {
+        try {
+          writer.close(null);
+          throw new IOException();
+        } catch (IOException e) {
+          ioException.set(e);
+        }
+        return null;
+      });
+    }
+    try {
+      executorService.invokeAll(callableList);
+    } catch (InterruptedException e) {
+      Thread.currentThread().interrupt();
+    } finally {
+      executorService.shutdown();

Review Comment:
   Sorry, I think I gave some bad advice here. I see now that you're using `invokeAll`, and that method only returns after all invocations complete. Therefore, we know the work is all done, and we can proceed to `shutdown`.
   
   Calling `awaitTermination` opens up a new problem: how to decide on the timeout, here arbitrarily chosen as 50 seconds. Since we don't need really need `awaitTermination`, we might as well remove it and avoid that problem.



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: common-issues-unsubscribe@hadoop.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: common-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: common-issues-help@hadoop.apache.org