HDFS-15210. EC : File write hanged when DN is shutdown by admin command. Contributed by Surendra Singh Lilhore.

hadoop-hdfs-project/hadoop-hdfs-client/src/main/java/org/apache/hadoop/hdfs/DFSStripedOutputStream.java

@@ -283,6 +283,7 @@ private void flipDataBuffers() {
   private ExecutorService flushAllExecutor;
   private CompletionService<Void> flushAllExecutorCompletionService;
   private int blockGroupIndex;
+  private long datanodeRestartTimeout;
 
   /** Construct a new output stream for creating a file. */
   DFSStripedOutputStream(DFSClient dfsClient, String src, HdfsFileStatus stat,
@@ -322,6 +323,7 @@ private void flipDataBuffers() {
       streamers.add(streamer);
     }
     currentPackets = new DFSPacket[streamers.size()];
+    datanodeRestartTimeout = dfsClient.getConf().getDatanodeRestartTimeout();
     setCurrentStreamer(0);
   }
 
@@ -637,6 +639,11 @@ private Set<StripedDataStreamer> markExternalErrorOnStreamers() {
             "streamer: " + streamer);
         streamer.setExternalError();
         healthySet.add(streamer);
+      } else if (!streamer.streamerClosed()
+          && streamer.getErrorState().hasDatanodeError()
+          && streamer.getErrorState().doWaitForRestart()) {
+        healthySet.add(streamer);
+        failedStreamers.remove(streamer);
       }
     }
     return healthySet;
@@ -701,6 +708,14 @@ private void checkStreamerFailures(boolean isNeedFlushAllPackets)
       for (int i = 0; i < numAllBlocks; i++) {
         coordinator.offerStreamerUpdateResult(i, newFailed.size() == 0);
       }
+      //wait for get notify to failed stream
+      if (newFailed.size() != 0) {
+        try {
+          Thread.sleep(datanodeRestartTimeout);
+        } catch (InterruptedException e) {
+          // Do nothing
+        }
+      }
     }
   }
 

hadoop-hdfs-project/hadoop-hdfs-client/src/main/java/org/apache/hadoop/hdfs/StripedDataStreamer.java

@@ -143,7 +143,8 @@ protected void setupPipelineInternal(DatanodeInfo[] nodes,
 
       // set up the pipeline again with the remaining nodes. when a striped
       // data streamer comes here, it must be in external error state.
-      assert getErrorState().hasExternalError();
+      assert getErrorState().hasExternalError()
+          || getErrorState().doWaitForRestart();
       success = createBlockOutputStream(nodes, nodeStorageTypes,
           nodeStorageIDs, newGS, true);