[SPARK-3410] The priority of shutdownhook for ApplicationMaster should not be integer literal

I think, it need to keep the priority of shutdown hook for ApplicationMaster than the priority of shutdown hook for o.a.h.FileSystem depending on changing the priority for FileSystem.

Author: Kousuke Saruta <sarutak@oss.nttdata.co.jp>

Closes #2283 from sarutak/SPARK-3410 and squashes the following commits:

1d44fef [Kousuke Saruta] Merge branch 'master' of git://git.apache.org/spark into SPARK-3410
bd6cc53 [Kousuke Saruta] Modified style
ee6f1aa [Kousuke Saruta] Added constant "SHUTDOWN_HOOK_PRIORITY" to ApplicationMaster
54eb68f [Kousuke Saruta] Changed Shutdown hook priority to 20
2f0aee3 [Kousuke Saruta] Merge branch 'master' of git://git.apache.org/spark into SPARK-3410
4c5cb93 [Kousuke Saruta] Modified the priority for AM's shutdown hook
217d1a4 [Kousuke Saruta] Removed unused import statements
717aba2 [Kousuke Saruta] Modified ApplicationMaster to make to keep the priority of shutdown hook for ApplicationMaster higher than the priority of shutdown hook for HDFS

yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala

@@ -21,12 +21,8 @@ import java.io.IOException
 import java.net.Socket
 import java.util.concurrent.atomic.AtomicReference
 
-import scala.collection.JavaConversions._
-import scala.util.Try
-
 import akka.actor._
 import akka.remote._
-import org.apache.hadoop.conf.Configuration
 import org.apache.hadoop.fs.{FileSystem, Path}
 import org.apache.hadoop.util.ShutdownHookManager
 import org.apache.hadoop.yarn.api._
@@ -107,8 +103,11 @@ private[spark] class ApplicationMaster(args: ApplicationMasterArguments,
         }
       }
     }
-    // Use priority 30 as it's higher than HDFS. It's the same priority MapReduce is using.
-    ShutdownHookManager.get().addShutdownHook(cleanupHook, 30)
+
+    // Use higher priority than FileSystem.
+    assert(ApplicationMaster.SHUTDOWN_HOOK_PRIORITY > FileSystem.SHUTDOWN_HOOK_PRIORITY)
+    ShutdownHookManager
+      .get().addShutdownHook(cleanupHook, ApplicationMaster.SHUTDOWN_HOOK_PRIORITY)
 
     // Call this to force generation of secret so it gets populated into the
     // Hadoop UGI. This has to happen before the startUserClass which does a
@@ -407,6 +406,8 @@ private[spark] class ApplicationMaster(args: ApplicationMasterArguments,
 
 object ApplicationMaster extends Logging {
 
+  val SHUTDOWN_HOOK_PRIORITY: Int = 30
+
   private var master: ApplicationMaster = _
 
   def main(args: Array[String]) = {