Retry for transient exceptions while doing cleanup for Hadoop Jobs

indexing-hadoop/src/main/java/io/druid/indexer/JobHelper.java

@@ -774,4 +774,26 @@ public void stopSection(String section)
       }
     };
   }
+
+  public static boolean deleteWithRetry(final FileSystem fs, final Path path, final boolean recursive)
+  {
+    try {
+      return RetryUtils.retry(
+          new Callable<Boolean>()
+          {
+            @Override
+            public Boolean call() throws Exception
+            {
+              return fs.delete(path, recursive);
+            }
+          },
+          shouldRetryPredicate(),
+          NUM_RETRIES
+      );
+    }
+    catch (Exception e) {
+      log.error(e, "Failed to cleanup path[%s]", path);
+      throw Throwables.propagate(e);
+    }
+  }
 }

indexing-hadoop/src/main/java/io/druid/indexer/updater/HadoopConverterJob.java

@@ -147,8 +147,26 @@ public static void cleanup(Job job) throws IOException
   {
     final Path jobDir = getJobPath(job.getJobID(), job.getWorkingDirectory());
     final FileSystem fs = jobDir.getFileSystem(job.getConfiguration());
-    fs.delete(jobDir, true);
-    fs.delete(getJobClassPathDir(job.getJobName(), job.getWorkingDirectory()), true);
+    RuntimeException e = null;
+    try {
+      JobHelper.deleteWithRetry(fs, jobDir, true);
+    }
+    catch (RuntimeException ex) {
+      e = ex;
+    }
+    try {
+      JobHelper.deleteWithRetry(fs, getJobClassPathDir(job.getJobName(), job.getWorkingDirectory()), true);
+    }
+    catch (RuntimeException ex) {
+      if (e == null) {
+        e = ex;
+      } else {
+        e.addSuppressed(ex);
+      }
+    }
+    if (e != null) {
+      throw e;
+    }
   }