YARN-11588. [Federation] Fix uncleaned threads in yarn router thread pool executor (#6159) Contributed by Jeffrey Chang.

Reviewed-by: Inigo Goiri <inigoiri@apache.org>
Signed-off-by: Shilun Fan <slfan1989@apache.org>

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-router/src/main/java/org/apache/hadoop/yarn/server/router/clientrm/FederationClientInterceptor.java

@@ -230,6 +230,9 @@ public class FederationClientInterceptor
     this.executorService = new ThreadPoolExecutor(numMinThreads, numMaxThreads,
         keepAliveTime, TimeUnit.MILLISECONDS, workQueue, threadFactory);
 
+    // Adding this line so that unused user threads will exit and be cleaned up if idle for too long
+    this.executorService.allowCoreThreadTimeOut(true);
+
     final Configuration conf = this.getConf();
 
     try {