HDFS-16586. Purge FsDatasetAsyncDiskService threadgroup; it causes BP… (

#4338)

* HDFS-16586. Purge FsDatasetAsyncDiskService threadgroup; it causes BPServiceActor IllegalThreadStateException 'fatal exception and exit'

Remove the ThreadGroup used by executor factories; they are unused
and ThreadGroups auto-destroy when their Thread-member count goes to zero.
This behavior is incompatible with the configuration we have on the per-volume
executor which is set to let all threads die if no use inside the
keepalive time.

Signed-off-by: Hexiaoqiao <hexiaoqiao@apache.org>

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/datanode/fsdataset/impl/FsDatasetAsyncDiskService.java

@@ -73,7 +73,6 @@ class FsDatasetAsyncDiskService {
 
   private final DataNode datanode;
   private final FsDatasetImpl fsdatasetImpl;
-  private final ThreadGroup threadGroup;
   private Map<String, ThreadPoolExecutor> executors
       = new HashMap<String, ThreadPoolExecutor>();
   private Map<String, Set<Long>> deletedBlockIds 
@@ -91,7 +90,6 @@ class FsDatasetAsyncDiskService {
   FsDatasetAsyncDiskService(DataNode datanode, FsDatasetImpl fsdatasetImpl) {
     this.datanode = datanode;
     this.fsdatasetImpl = fsdatasetImpl;
-    this.threadGroup = new ThreadGroup(getClass().getSimpleName());
     maxNumThreadsPerVolume = datanode.getConf().getInt(
       DFSConfigKeys.DFS_DATANODE_FSDATASETASYNCDISK_MAX_THREADS_PER_VOLUME_KEY,
           DFSConfigKeys.DFS_DATANODE_FSDATASETASYNCDISK_MAX_THREADS_PER_VOLUME_DEFAULT);
@@ -110,7 +108,7 @@ public Thread newThread(Runnable r) {
         synchronized (this) {
           thisIndex = counter++;
         }
-        Thread t = new Thread(threadGroup, r);
+        Thread t = new Thread(r);
         t.setName("Async disk worker #" + thisIndex +
             " for volume " + volume);
         return t;