From 60af8793b45b4057101a22e4248d7ca022b52d79 Mon Sep 17 00:00:00 2001
From: Abhishek Modi <abmodi@apache.org>
Date: Mon, 9 Sep 2019 14:26:00 +0530
Subject: [PATCH] YARN-9821. NM hangs at serviceStop when ATSV2 Backend Hbase
 is Down. Contributed by Prabhu Joseph.

---
 .../storage/HBaseTimelineWriterImpl.java      | 74 +++++++++++--------
 1 file changed, 42 insertions(+), 32 deletions(-)

diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-timelineservice-hbase/hadoop-yarn-server-timelineservice-hbase-client/src/main/java/org/apache/hadoop/yarn/server/timelineservice/storage/HBaseTimelineWriterImpl.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-timelineservice-hbase/hadoop-yarn-server-timelineservice-hbase-client/src/main/java/org/apache/hadoop/yarn/server/timelineservice/storage/HBaseTimelineWriterImpl.java
index a398febccc..dda004d7ec 100644
--- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-timelineservice-hbase/hadoop-yarn-server-timelineservice-hbase-client/src/main/java/org/apache/hadoop/yarn/server/timelineservice/storage/HBaseTimelineWriterImpl.java
+++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-timelineservice-hbase/hadoop-yarn-server-timelineservice-hbase-client/src/main/java/org/apache/hadoop/yarn/server/timelineservice/storage/HBaseTimelineWriterImpl.java
@@ -630,39 +630,49 @@ public void flush() throws IOException {
    */
   @Override
   protected void serviceStop() throws Exception {
-    if (entityTable != null) {
-      LOG.info("closing the entity table");
-      // The close API performs flushing and releases any resources held
-      entityTable.close();
+    boolean isStorageUp = true;
+    try {
+      storageMonitor.checkStorageIsUp();
+    } catch (IOException e) {
+      LOG.warn("Failed to close the timeline tables as Hbase is down", e);
+      isStorageUp = false;
     }
-    if (appToFlowTable != null) {
-      LOG.info("closing the app_flow table");
-      // The close API performs flushing and releases any resources held
-      appToFlowTable.close();
-    }
-    if (applicationTable != null) {
-      LOG.info("closing the application table");
-      applicationTable.close();
-    }
-    if (flowRunTable != null) {
-      LOG.info("closing the flow run table");
-      // The close API performs flushing and releases any resources held
-      flowRunTable.close();
-    }
-    if (flowActivityTable != null) {
-      LOG.info("closing the flowActivityTable table");
-      // The close API performs flushing and releases any resources held
-      flowActivityTable.close();
-    }
-    if (subApplicationTable != null) {
-      subApplicationTable.close();
-    }
-    if (domainTable != null) {
-      domainTable.close();
-    }
-    if (conn != null) {
-      LOG.info("closing the hbase Connection");
-      conn.close();
+
+    if (isStorageUp) {
+      if (entityTable != null) {
+        LOG.info("closing the entity table");
+        // The close API performs flushing and releases any resources held
+        entityTable.close();
+      }
+      if (appToFlowTable != null) {
+        LOG.info("closing the app_flow table");
+        // The close API performs flushing and releases any resources held
+        appToFlowTable.close();
+      }
+      if (applicationTable != null) {
+        LOG.info("closing the application table");
+        applicationTable.close();
+      }
+      if (flowRunTable != null) {
+        LOG.info("closing the flow run table");
+        // The close API performs flushing and releases any resources held
+        flowRunTable.close();
+      }
+      if (flowActivityTable != null) {
+        LOG.info("closing the flowActivityTable table");
+        // The close API performs flushing and releases any resources held
+        flowActivityTable.close();
+      }
+      if (subApplicationTable != null) {
+        subApplicationTable.close();
+      }
+      if (domainTable != null) {
+        domainTable.close();
+      }
+      if (conn != null) {
+        LOG.info("closing the hbase Connection");
+        conn.close();
+      }
     }
     storageMonitor.stop();
     super.serviceStop();