From ca0349b87ab1b2d0d2b9dc93de7806d26713165c Mon Sep 17 00:00:00 2001
From: Jian He <jianhe@apache.org>
Date: Fri, 26 Dec 2014 17:59:58 -0800
Subject: [PATCH] YARN-2992. ZKRMStateStore crashes due to session expiry.
 Contributed by Karthik Kambatla (cherry picked from commit
 1454efe5d4fe4214ec5ef9142d55dbeca7dab953)

---
 hadoop-yarn-project/CHANGES.txt                                | 3 +++
 .../yarn/server/resourcemanager/recovery/ZKRMStateStore.java   | 3 +++
 2 files changed, 6 insertions(+)

diff --git a/hadoop-yarn-project/CHANGES.txt b/hadoop-yarn-project/CHANGES.txt
index 1098eaca93d..1c8c23718b3 100644
--- a/hadoop-yarn-project/CHANGES.txt
+++ b/hadoop-yarn-project/CHANGES.txt
@@ -257,6 +257,9 @@ Release 2.7.0 - UNRELEASED
 
     YARN-2988. Graph#save() may leak file descriptors. (Ted Yu via ozawa)
 
+    YARN-2992. ZKRMStateStore crashes due to session expiry. (Karthik Kambatla
+    via jianhe)
+
 Release 2.6.0 - 2014-11-18
 
   INCOMPATIBLE CHANGES
diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/ZKRMStateStore.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/ZKRMStateStore.java
index e559e4c41e1..285f3d0be54 100644
--- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/ZKRMStateStore.java
+++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/ZKRMStateStore.java
@@ -1056,6 +1056,8 @@ public class ZKRMStateStore extends RMStateStore {
       switch (code) {
         case CONNECTIONLOSS:
         case OPERATIONTIMEOUT:
+        case SESSIONEXPIRED:
+        case SESSIONMOVED:
           return true;
         default:
           break;
@@ -1084,6 +1086,7 @@ public class ZKRMStateStore extends RMStateStore {
           if (shouldRetry(ke.code()) && ++retry < numRetries) {
             LOG.info("Retrying operation on ZK. Retry no. " + retry);
             Thread.sleep(zkRetryInterval);
+            createConnection();
             continue;
           }
           LOG.info("Maxed out ZK retries. Giving up!");