HBASE-3843 splitLogWorker starts too early

git-svn-id: https://svn.apache.org/repos/asf/hbase/trunk@1098696 13f79535-47bb-0310-9956-ffa450edef68

CHANGES.txt

@@ -92,6 +92,7 @@ Release 0.91.0 - Unreleased
    HBASE-3827  hbase-1502, removing heartbeats, broke master joining a running
                cluster and was returning master hostname for rs to use
    HBASE-3829  TestMasterFailover failures in jenkins
+   HBASE-3843  splitLogWorker starts too early (Prakash Khemani)
 
   IMPROVEMENTS
    HBASE-3290  Max Compaction Size (Nicolas Spiegelberg via Stack)  

src/main/java/org/apache/hadoop/hbase/regionserver/HRegionServer.java

@@ -537,11 +537,6 @@ public class HRegionServer implements HRegionInterface, HBaseRPCErrorHandler,
     this.catalogTracker = new CatalogTracker(this.zooKeeper, this.connection,
       this, this.conf.getInt("hbase.regionserver.catalog.timeout", Integer.MAX_VALUE));
     catalogTracker.start();
-
-    // Create the log splitting worker and start it
-    this.splitLogWorker = new SplitLogWorker(this.zooKeeper,
-        this.getConfiguration(), this.getServerName().toString());
-    splitLogWorker.start();
   }
 
   /**
@@ -1274,6 +1269,11 @@ public class HRegionServer implements HRegionInterface, HBaseRPCErrorHandler,
     // Start Server.  This service is like leases in that it internally runs
     // a thread.
     this.rpcServer.start();
+
+    // Create the log splitting worker and start it
+    this.splitLogWorker = new SplitLogWorker(this.zooKeeper,
+        this.getConfiguration(), this.getServerName().toString());
+    splitLogWorker.start();
   }
 
   /**