svn commit: r1520653 - /hbase/branches/0.89-fb/src/main/java/org/apache/hadoop/hbase/master/HMaster.java

[li...@apache.org]

Author: liyin
Date: Fri Sep  6 18:18:50 2013
New Revision: 1520653

URL: http://svn.apache.org/r1520653
Log:
[master]Let HMaster not close DFS client when there is transient failure

Author: fan

Summary: Currently when DFS client has problem, HMaster will close it but still loops on checking its availability. So when name node comes back, HMaster does not recover.

Test Plan: Push new jar to master of a shadow cluster, then shut down name node for a couple minutes. When name node is brought back, HMaster should recover soon.

Reviewers: aaiyer

Reviewed By: aaiyer

CC: hbase-eng@, manukranthk

Differential Revision: https://phabricator.fb.com/D956599

Modified:
    hbase/branches/0.89-fb/src/main/java/org/apache/hadoop/hbase/master/HMaster.java

Modified: hbase/branches/0.89-fb/src/main/java/org/apache/hadoop/hbase/master/HMaster.java
URL: http://svn.apache.org/viewvc/hbase/branches/0.89-fb/src/main/java/org/apache/hadoop/hbase/master/HMaster.java?rev=1520653&r1=1520652&r2=1520653&view=diff
==============================================================================
--- hbase/branches/0.89-fb/src/main/java/org/apache/hadoop/hbase/master/HMaster.java (original)
+++ hbase/branches/0.89-fb/src/main/java/org/apache/hadoop/hbase/master/HMaster.java Fri Sep  6 18:18:50 2013
@@ -660,7 +660,9 @@ public class HMaster extends HasThread i
   protected boolean checkFileSystem(boolean shutdownClusterOnFail) {
     if (this.fsOk) {
       try {
-        FSUtils.checkFileSystemAvailable(this.fs);
+        // Do not close dfs client when check fails. It's OK to reuse the
+        //  same dfs client when namenode comes back.
+        FSUtils.checkFileSystemAvailable(this.fs, false);
       } catch (IOException e) {
         if (shutdownClusterOnFail) {
           LOG.fatal("Shutting down HBase cluster: file system not available", e);