hbase-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From git-site-r...@apache.org
Subject [39/51] [partial] hbase-site git commit: Published site at .
Date Wed, 04 Oct 2017 15:13:43 GMT
http://git-wip-us.apache.org/repos/asf/hbase-site/blob/387c1112/apidocs/src-html/org/apache/hadoop/hbase/mapreduce/TableInputFormatBase.html
----------------------------------------------------------------------
diff --git a/apidocs/src-html/org/apache/hadoop/hbase/mapreduce/TableInputFormatBase.html b/apidocs/src-html/org/apache/hadoop/hbase/mapreduce/TableInputFormatBase.html
index 2a28959..898d59a 100644
--- a/apidocs/src-html/org/apache/hadoop/hbase/mapreduce/TableInputFormatBase.html
+++ b/apidocs/src-html/org/apache/hadoop/hbase/mapreduce/TableInputFormatBase.html
@@ -32,23 +32,23 @@
 <span class="sourceLineNo">024</span>import java.net.InetSocketAddress;<a name="line.24"></a>
 <span class="sourceLineNo">025</span>import java.net.UnknownHostException;<a name="line.25"></a>
 <span class="sourceLineNo">026</span>import java.util.ArrayList;<a name="line.26"></a>
-<span class="sourceLineNo">027</span>import java.util.Arrays;<a name="line.27"></a>
-<span class="sourceLineNo">028</span>import java.util.HashMap;<a name="line.28"></a>
-<span class="sourceLineNo">029</span>import java.util.List;<a name="line.29"></a>
+<span class="sourceLineNo">027</span>import java.util.HashMap;<a name="line.27"></a>
+<span class="sourceLineNo">028</span>import java.util.List;<a name="line.28"></a>
+<span class="sourceLineNo">029</span>import java.util.Map;<a name="line.29"></a>
 <span class="sourceLineNo">030</span><a name="line.30"></a>
 <span class="sourceLineNo">031</span>import org.apache.commons.logging.Log;<a name="line.31"></a>
 <span class="sourceLineNo">032</span>import org.apache.commons.logging.LogFactory;<a name="line.32"></a>
-<span class="sourceLineNo">033</span>import org.apache.hadoop.conf.Configuration;<a name="line.33"></a>
-<span class="sourceLineNo">034</span>import org.apache.yetus.audience.InterfaceAudience;<a name="line.34"></a>
-<span class="sourceLineNo">035</span>import org.apache.hadoop.hbase.HConstants;<a name="line.35"></a>
-<span class="sourceLineNo">036</span>import org.apache.hadoop.hbase.HRegionLocation;<a name="line.36"></a>
-<span class="sourceLineNo">037</span>import org.apache.hadoop.hbase.TableName;<a name="line.37"></a>
-<span class="sourceLineNo">038</span>import org.apache.hadoop.hbase.client.Admin;<a name="line.38"></a>
-<span class="sourceLineNo">039</span>import org.apache.hadoop.hbase.client.Connection;<a name="line.39"></a>
-<span class="sourceLineNo">040</span>import org.apache.hadoop.hbase.client.RegionLocator;<a name="line.40"></a>
-<span class="sourceLineNo">041</span>import org.apache.hadoop.hbase.client.Result;<a name="line.41"></a>
-<span class="sourceLineNo">042</span>import org.apache.hadoop.hbase.client.Scan;<a name="line.42"></a>
-<span class="sourceLineNo">043</span>import org.apache.hadoop.hbase.client.Table;<a name="line.43"></a>
+<span class="sourceLineNo">033</span>import org.apache.yetus.audience.InterfaceAudience;<a name="line.33"></a>
+<span class="sourceLineNo">034</span>import org.apache.hadoop.hbase.HConstants;<a name="line.34"></a>
+<span class="sourceLineNo">035</span>import org.apache.hadoop.hbase.HRegionLocation;<a name="line.35"></a>
+<span class="sourceLineNo">036</span>import org.apache.hadoop.hbase.TableName;<a name="line.36"></a>
+<span class="sourceLineNo">037</span>import org.apache.hadoop.hbase.client.Admin;<a name="line.37"></a>
+<span class="sourceLineNo">038</span>import org.apache.hadoop.hbase.client.Connection;<a name="line.38"></a>
+<span class="sourceLineNo">039</span>import org.apache.hadoop.hbase.client.RegionLocator;<a name="line.39"></a>
+<span class="sourceLineNo">040</span>import org.apache.hadoop.hbase.client.Result;<a name="line.40"></a>
+<span class="sourceLineNo">041</span>import org.apache.hadoop.hbase.client.Scan;<a name="line.41"></a>
+<span class="sourceLineNo">042</span>import org.apache.hadoop.hbase.client.Table;<a name="line.42"></a>
+<span class="sourceLineNo">043</span>import org.apache.hadoop.hbase.exceptions.IllegalArgumentIOException;<a name="line.43"></a>
 <span class="sourceLineNo">044</span>import org.apache.hadoop.hbase.io.ImmutableBytesWritable;<a name="line.44"></a>
 <span class="sourceLineNo">045</span>import org.apache.hadoop.hbase.util.Addressing;<a name="line.45"></a>
 <span class="sourceLineNo">046</span>import org.apache.hadoop.hbase.util.Bytes;<a name="line.46"></a>
@@ -101,563 +101,566 @@
 <span class="sourceLineNo">093</span> *     }<a name="line.93"></a>
 <span class="sourceLineNo">094</span> *   }<a name="line.94"></a>
 <span class="sourceLineNo">095</span> * &lt;/pre&gt;<a name="line.95"></a>
-<span class="sourceLineNo">096</span> */<a name="line.96"></a>
-<span class="sourceLineNo">097</span>@InterfaceAudience.Public<a name="line.97"></a>
-<span class="sourceLineNo">098</span>public abstract class TableInputFormatBase<a name="line.98"></a>
-<span class="sourceLineNo">099</span>extends InputFormat&lt;ImmutableBytesWritable, Result&gt; {<a name="line.99"></a>
-<span class="sourceLineNo">100</span><a name="line.100"></a>
-<span class="sourceLineNo">101</span>  /** Specify if we enable auto-balance for input in M/R jobs.*/<a name="line.101"></a>
-<span class="sourceLineNo">102</span>  public static final String MAPREDUCE_INPUT_AUTOBALANCE = "hbase.mapreduce.input.autobalance";<a name="line.102"></a>
-<span class="sourceLineNo">103</span>  /** Specify if ratio for data skew in M/R jobs, it goes well with the enabling hbase.mapreduce<a name="line.103"></a>
-<span class="sourceLineNo">104</span>   * .input.autobalance property.*/<a name="line.104"></a>
-<span class="sourceLineNo">105</span>  public static final String INPUT_AUTOBALANCE_MAXSKEWRATIO = "hbase.mapreduce.input.autobalance" +<a name="line.105"></a>
-<span class="sourceLineNo">106</span>          ".maxskewratio";<a name="line.106"></a>
-<span class="sourceLineNo">107</span>  /** Specify if the row key in table is text (ASCII between 32~126),<a name="line.107"></a>
-<span class="sourceLineNo">108</span>   * default is true. False means the table is using binary row key*/<a name="line.108"></a>
-<span class="sourceLineNo">109</span>  public static final String TABLE_ROW_TEXTKEY = "hbase.table.row.textkey";<a name="line.109"></a>
-<span class="sourceLineNo">110</span><a name="line.110"></a>
-<span class="sourceLineNo">111</span>  private static final Log LOG = LogFactory.getLog(TableInputFormatBase.class);<a name="line.111"></a>
-<span class="sourceLineNo">112</span><a name="line.112"></a>
-<span class="sourceLineNo">113</span>  private static final String NOT_INITIALIZED = "The input format instance has not been properly " +<a name="line.113"></a>
-<span class="sourceLineNo">114</span>      "initialized. Ensure you call initializeTable either in your constructor or initialize " +<a name="line.114"></a>
-<span class="sourceLineNo">115</span>      "method";<a name="line.115"></a>
-<span class="sourceLineNo">116</span>  private static final String INITIALIZATION_ERROR = "Cannot create a record reader because of a" +<a name="line.116"></a>
-<span class="sourceLineNo">117</span>            " previous error. Please look at the previous logs lines from" +<a name="line.117"></a>
-<span class="sourceLineNo">118</span>            " the task's full log for more details.";<a name="line.118"></a>
-<span class="sourceLineNo">119</span><a name="line.119"></a>
-<span class="sourceLineNo">120</span>  /** Holds the details for the internal scanner.<a name="line.120"></a>
-<span class="sourceLineNo">121</span>   *<a name="line.121"></a>
-<span class="sourceLineNo">122</span>   * @see Scan */<a name="line.122"></a>
-<span class="sourceLineNo">123</span>  private Scan scan = null;<a name="line.123"></a>
-<span class="sourceLineNo">124</span>  /** The {@link Admin}. */<a name="line.124"></a>
-<span class="sourceLineNo">125</span>  private Admin admin;<a name="line.125"></a>
-<span class="sourceLineNo">126</span>  /** The {@link Table} to scan. */<a name="line.126"></a>
-<span class="sourceLineNo">127</span>  private Table table;<a name="line.127"></a>
-<span class="sourceLineNo">128</span>  /** The {@link RegionLocator} of the table. */<a name="line.128"></a>
-<span class="sourceLineNo">129</span>  private RegionLocator regionLocator;<a name="line.129"></a>
-<span class="sourceLineNo">130</span>  /** The reader scanning the table, can be a custom one. */<a name="line.130"></a>
-<span class="sourceLineNo">131</span>  private TableRecordReader tableRecordReader = null;<a name="line.131"></a>
-<span class="sourceLineNo">132</span>  /** The underlying {@link Connection} of the table. */<a name="line.132"></a>
-<span class="sourceLineNo">133</span>  private Connection connection;<a name="line.133"></a>
-<span class="sourceLineNo">134</span><a name="line.134"></a>
-<span class="sourceLineNo">135</span><a name="line.135"></a>
-<span class="sourceLineNo">136</span>  /** The reverse DNS lookup cache mapping: IPAddress =&gt; HostName */<a name="line.136"></a>
-<span class="sourceLineNo">137</span>  private HashMap&lt;InetAddress, String&gt; reverseDNSCacheMap = new HashMap&lt;&gt;();<a name="line.137"></a>
-<span class="sourceLineNo">138</span><a name="line.138"></a>
-<span class="sourceLineNo">139</span>  /**<a name="line.139"></a>
-<span class="sourceLineNo">140</span>   * Builds a {@link TableRecordReader}. If no {@link TableRecordReader} was provided, uses<a name="line.140"></a>
-<span class="sourceLineNo">141</span>   * the default.<a name="line.141"></a>
-<span class="sourceLineNo">142</span>   *<a name="line.142"></a>
-<span class="sourceLineNo">143</span>   * @param split  The split to work with.<a name="line.143"></a>
-<span class="sourceLineNo">144</span>   * @param context  The current context.<a name="line.144"></a>
-<span class="sourceLineNo">145</span>   * @return The newly created record reader.<a name="line.145"></a>
-<span class="sourceLineNo">146</span>   * @throws IOException When creating the reader fails.<a name="line.146"></a>
-<span class="sourceLineNo">147</span>   * @see org.apache.hadoop.mapreduce.InputFormat#createRecordReader(<a name="line.147"></a>
-<span class="sourceLineNo">148</span>   *   org.apache.hadoop.mapreduce.InputSplit,<a name="line.148"></a>
-<span class="sourceLineNo">149</span>   *   org.apache.hadoop.mapreduce.TaskAttemptContext)<a name="line.149"></a>
-<span class="sourceLineNo">150</span>   */<a name="line.150"></a>
-<span class="sourceLineNo">151</span>  @Override<a name="line.151"></a>
-<span class="sourceLineNo">152</span>  public RecordReader&lt;ImmutableBytesWritable, Result&gt; createRecordReader(<a name="line.152"></a>
-<span class="sourceLineNo">153</span>      InputSplit split, TaskAttemptContext context)<a name="line.153"></a>
-<span class="sourceLineNo">154</span>  throws IOException {<a name="line.154"></a>
-<span class="sourceLineNo">155</span>    // Just in case a subclass is relying on JobConfigurable magic.<a name="line.155"></a>
-<span class="sourceLineNo">156</span>    if (table == null) {<a name="line.156"></a>
-<span class="sourceLineNo">157</span>      initialize(context);<a name="line.157"></a>
-<span class="sourceLineNo">158</span>    }<a name="line.158"></a>
-<span class="sourceLineNo">159</span>    // null check in case our child overrides getTable to not throw.<a name="line.159"></a>
-<span class="sourceLineNo">160</span>    try {<a name="line.160"></a>
-<span class="sourceLineNo">161</span>      if (getTable() == null) {<a name="line.161"></a>
-<span class="sourceLineNo">162</span>        // initialize() must not have been implemented in the subclass.<a name="line.162"></a>
-<span class="sourceLineNo">163</span>        throw new IOException(INITIALIZATION_ERROR);<a name="line.163"></a>
-<span class="sourceLineNo">164</span>      }<a name="line.164"></a>
-<span class="sourceLineNo">165</span>    } catch (IllegalStateException exception) {<a name="line.165"></a>
-<span class="sourceLineNo">166</span>      throw new IOException(INITIALIZATION_ERROR, exception);<a name="line.166"></a>
-<span class="sourceLineNo">167</span>    }<a name="line.167"></a>
-<span class="sourceLineNo">168</span>    TableSplit tSplit = (TableSplit) split;<a name="line.168"></a>
-<span class="sourceLineNo">169</span>    LOG.info("Input split length: " + StringUtils.humanReadableInt(tSplit.getLength()) + " bytes.");<a name="line.169"></a>
-<span class="sourceLineNo">170</span>    final TableRecordReader trr =<a name="line.170"></a>
-<span class="sourceLineNo">171</span>        this.tableRecordReader != null ? this.tableRecordReader : new TableRecordReader();<a name="line.171"></a>
-<span class="sourceLineNo">172</span>    Scan sc = new Scan(this.scan);<a name="line.172"></a>
-<span class="sourceLineNo">173</span>    sc.setStartRow(tSplit.getStartRow());<a name="line.173"></a>
-<span class="sourceLineNo">174</span>    sc.setStopRow(tSplit.getEndRow());<a name="line.174"></a>
-<span class="sourceLineNo">175</span>    trr.setScan(sc);<a name="line.175"></a>
-<span class="sourceLineNo">176</span>    trr.setTable(getTable());<a name="line.176"></a>
-<span class="sourceLineNo">177</span>    return new RecordReader&lt;ImmutableBytesWritable, Result&gt;() {<a name="line.177"></a>
-<span class="sourceLineNo">178</span><a name="line.178"></a>
-<span class="sourceLineNo">179</span>      @Override<a name="line.179"></a>
-<span class="sourceLineNo">180</span>      public void close() throws IOException {<a name="line.180"></a>
-<span class="sourceLineNo">181</span>        trr.close();<a name="line.181"></a>
-<span class="sourceLineNo">182</span>        closeTable();<a name="line.182"></a>
-<span class="sourceLineNo">183</span>      }<a name="line.183"></a>
-<span class="sourceLineNo">184</span><a name="line.184"></a>
-<span class="sourceLineNo">185</span>      @Override<a name="line.185"></a>
-<span class="sourceLineNo">186</span>      public ImmutableBytesWritable getCurrentKey() throws IOException, InterruptedException {<a name="line.186"></a>
-<span class="sourceLineNo">187</span>        return trr.getCurrentKey();<a name="line.187"></a>
-<span class="sourceLineNo">188</span>      }<a name="line.188"></a>
+<span class="sourceLineNo">096</span> *<a name="line.96"></a>
+<span class="sourceLineNo">097</span> *<a name="line.97"></a>
+<span class="sourceLineNo">098</span> * The number of InputSplits(mappers) match the number of regions in a table by default.<a name="line.98"></a>
+<span class="sourceLineNo">099</span> * Set "hbase.mapreduce.tableinput.mappers.per.region" to specify how many mappers per region, set<a name="line.99"></a>
+<span class="sourceLineNo">100</span> * this property will disable autobalance below.\<a name="line.100"></a>
+<span class="sourceLineNo">101</span> * Set "hbase.mapreduce.tif.input.autobalance" to enable autobalance, hbase will assign mappers<a name="line.101"></a>
+<span class="sourceLineNo">102</span> * based on average region size; For regions, whose size larger than average region size may assigned<a name="line.102"></a>
+<span class="sourceLineNo">103</span> * more mappers, and for smaller one, they may group together to use one mapper. If actual average<a name="line.103"></a>
+<span class="sourceLineNo">104</span> * region size is too big, like 50G, it is not good to only assign 1 mapper for those large regions.<a name="line.104"></a>
+<span class="sourceLineNo">105</span> * Use "hbase.mapreduce.tif.ave.regionsize" to set max average region size when enable "autobalanece",<a name="line.105"></a>
+<span class="sourceLineNo">106</span> * default mas average region size is 8G.<a name="line.106"></a>
+<span class="sourceLineNo">107</span> */<a name="line.107"></a>
+<span class="sourceLineNo">108</span>@InterfaceAudience.Public<a name="line.108"></a>
+<span class="sourceLineNo">109</span>public abstract class TableInputFormatBase<a name="line.109"></a>
+<span class="sourceLineNo">110</span>    extends InputFormat&lt;ImmutableBytesWritable, Result&gt; {<a name="line.110"></a>
+<span class="sourceLineNo">111</span><a name="line.111"></a>
+<span class="sourceLineNo">112</span>  private static final Log LOG = LogFactory.getLog(TableInputFormatBase.class);<a name="line.112"></a>
+<span class="sourceLineNo">113</span><a name="line.113"></a>
+<span class="sourceLineNo">114</span>  private static final String NOT_INITIALIZED = "The input format instance has not been properly " +<a name="line.114"></a>
+<span class="sourceLineNo">115</span>      "initialized. Ensure you call initializeTable either in your constructor or initialize " +<a name="line.115"></a>
+<span class="sourceLineNo">116</span>      "method";<a name="line.116"></a>
+<span class="sourceLineNo">117</span>  private static final String INITIALIZATION_ERROR = "Cannot create a record reader because of a" +<a name="line.117"></a>
+<span class="sourceLineNo">118</span>      " previous error. Please look at the previous logs lines from" +<a name="line.118"></a>
+<span class="sourceLineNo">119</span>      " the task's full log for more details.";<a name="line.119"></a>
+<span class="sourceLineNo">120</span><a name="line.120"></a>
+<span class="sourceLineNo">121</span>  /** Specify if we enable auto-balance to set number of mappers in M/R jobs. */<a name="line.121"></a>
+<span class="sourceLineNo">122</span>  public static final String MAPREDUCE_INPUT_AUTOBALANCE = "hbase.mapreduce.tif.input.autobalance";<a name="line.122"></a>
+<span class="sourceLineNo">123</span>  /** In auto-balance, we split input by ave region size, if calculated region size is too big, we can set it. */<a name="line.123"></a>
+<span class="sourceLineNo">124</span>  public static final String MAX_AVERAGE_REGION_SIZE = "hbase.mapreduce.tif.ave.regionsize";<a name="line.124"></a>
+<span class="sourceLineNo">125</span><a name="line.125"></a>
+<span class="sourceLineNo">126</span>  /** Set the number of Mappers for each region, all regions have same number of Mappers */<a name="line.126"></a>
+<span class="sourceLineNo">127</span>  public static final String NUM_MAPPERS_PER_REGION = "hbase.mapreduce.tableinput.mappers.per.region";<a name="line.127"></a>
+<span class="sourceLineNo">128</span><a name="line.128"></a>
+<span class="sourceLineNo">129</span><a name="line.129"></a>
+<span class="sourceLineNo">130</span>  /** Holds the details for the internal scanner.<a name="line.130"></a>
+<span class="sourceLineNo">131</span>   *<a name="line.131"></a>
+<span class="sourceLineNo">132</span>   * @see Scan */<a name="line.132"></a>
+<span class="sourceLineNo">133</span>  private Scan scan = null;<a name="line.133"></a>
+<span class="sourceLineNo">134</span>  /** The {@link Admin}. */<a name="line.134"></a>
+<span class="sourceLineNo">135</span>  private Admin admin;<a name="line.135"></a>
+<span class="sourceLineNo">136</span>  /** The {@link Table} to scan. */<a name="line.136"></a>
+<span class="sourceLineNo">137</span>  private Table table;<a name="line.137"></a>
+<span class="sourceLineNo">138</span>  /** The {@link RegionLocator} of the table. */<a name="line.138"></a>
+<span class="sourceLineNo">139</span>  private RegionLocator regionLocator;<a name="line.139"></a>
+<span class="sourceLineNo">140</span>  /** The reader scanning the table, can be a custom one. */<a name="line.140"></a>
+<span class="sourceLineNo">141</span>  private TableRecordReader tableRecordReader = null;<a name="line.141"></a>
+<span class="sourceLineNo">142</span>  /** The underlying {@link Connection} of the table. */<a name="line.142"></a>
+<span class="sourceLineNo">143</span>  private Connection connection;<a name="line.143"></a>
+<span class="sourceLineNo">144</span><a name="line.144"></a>
+<span class="sourceLineNo">145</span><a name="line.145"></a>
+<span class="sourceLineNo">146</span>  /** The reverse DNS lookup cache mapping: IPAddress =&gt; HostName */<a name="line.146"></a>
+<span class="sourceLineNo">147</span>  private HashMap&lt;InetAddress, String&gt; reverseDNSCacheMap =<a name="line.147"></a>
+<span class="sourceLineNo">148</span>      new HashMap&lt;&gt;();<a name="line.148"></a>
+<span class="sourceLineNo">149</span><a name="line.149"></a>
+<span class="sourceLineNo">150</span>  /**<a name="line.150"></a>
+<span class="sourceLineNo">151</span>   * Builds a {@link TableRecordReader}. If no {@link TableRecordReader} was provided, uses<a name="line.151"></a>
+<span class="sourceLineNo">152</span>   * the default.<a name="line.152"></a>
+<span class="sourceLineNo">153</span>   *<a name="line.153"></a>
+<span class="sourceLineNo">154</span>   * @param split  The split to work with.<a name="line.154"></a>
+<span class="sourceLineNo">155</span>   * @param context  The current context.<a name="line.155"></a>
+<span class="sourceLineNo">156</span>   * @return The newly created record reader.<a name="line.156"></a>
+<span class="sourceLineNo">157</span>   * @throws IOException When creating the reader fails.<a name="line.157"></a>
+<span class="sourceLineNo">158</span>   * @see org.apache.hadoop.mapreduce.InputFormat#createRecordReader(<a name="line.158"></a>
+<span class="sourceLineNo">159</span>   *   org.apache.hadoop.mapreduce.InputSplit,<a name="line.159"></a>
+<span class="sourceLineNo">160</span>   *   org.apache.hadoop.mapreduce.TaskAttemptContext)<a name="line.160"></a>
+<span class="sourceLineNo">161</span>   */<a name="line.161"></a>
+<span class="sourceLineNo">162</span>  @Override<a name="line.162"></a>
+<span class="sourceLineNo">163</span>  public RecordReader&lt;ImmutableBytesWritable, Result&gt; createRecordReader(<a name="line.163"></a>
+<span class="sourceLineNo">164</span>      InputSplit split, TaskAttemptContext context)<a name="line.164"></a>
+<span class="sourceLineNo">165</span>      throws IOException {<a name="line.165"></a>
+<span class="sourceLineNo">166</span>    // Just in case a subclass is relying on JobConfigurable magic.<a name="line.166"></a>
+<span class="sourceLineNo">167</span>    if (table == null) {<a name="line.167"></a>
+<span class="sourceLineNo">168</span>      initialize(context);<a name="line.168"></a>
+<span class="sourceLineNo">169</span>    }<a name="line.169"></a>
+<span class="sourceLineNo">170</span>    // null check in case our child overrides getTable to not throw.<a name="line.170"></a>
+<span class="sourceLineNo">171</span>    try {<a name="line.171"></a>
+<span class="sourceLineNo">172</span>      if (getTable() == null) {<a name="line.172"></a>
+<span class="sourceLineNo">173</span>        // initialize() must not have been implemented in the subclass.<a name="line.173"></a>
+<span class="sourceLineNo">174</span>        throw new IOException(INITIALIZATION_ERROR);<a name="line.174"></a>
+<span class="sourceLineNo">175</span>      }<a name="line.175"></a>
+<span class="sourceLineNo">176</span>    } catch (IllegalStateException exception) {<a name="line.176"></a>
+<span class="sourceLineNo">177</span>      throw new IOException(INITIALIZATION_ERROR, exception);<a name="line.177"></a>
+<span class="sourceLineNo">178</span>    }<a name="line.178"></a>
+<span class="sourceLineNo">179</span>    TableSplit tSplit = (TableSplit) split;<a name="line.179"></a>
+<span class="sourceLineNo">180</span>    LOG.info("Input split length: " + StringUtils.humanReadableInt(tSplit.getLength()) + " bytes.");<a name="line.180"></a>
+<span class="sourceLineNo">181</span>    final TableRecordReader trr =<a name="line.181"></a>
+<span class="sourceLineNo">182</span>        this.tableRecordReader != null ? this.tableRecordReader : new TableRecordReader();<a name="line.182"></a>
+<span class="sourceLineNo">183</span>    Scan sc = new Scan(this.scan);<a name="line.183"></a>
+<span class="sourceLineNo">184</span>    sc.setStartRow(tSplit.getStartRow());<a name="line.184"></a>
+<span class="sourceLineNo">185</span>    sc.setStopRow(tSplit.getEndRow());<a name="line.185"></a>
+<span class="sourceLineNo">186</span>    trr.setScan(sc);<a name="line.186"></a>
+<span class="sourceLineNo">187</span>    trr.setTable(getTable());<a name="line.187"></a>
+<span class="sourceLineNo">188</span>    return new RecordReader&lt;ImmutableBytesWritable, Result&gt;() {<a name="line.188"></a>
 <span class="sourceLineNo">189</span><a name="line.189"></a>
 <span class="sourceLineNo">190</span>      @Override<a name="line.190"></a>
-<span class="sourceLineNo">191</span>      public Result getCurrentValue() throws IOException, InterruptedException {<a name="line.191"></a>
-<span class="sourceLineNo">192</span>        return trr.getCurrentValue();<a name="line.192"></a>
-<span class="sourceLineNo">193</span>      }<a name="line.193"></a>
-<span class="sourceLineNo">194</span><a name="line.194"></a>
-<span class="sourceLineNo">195</span>      @Override<a name="line.195"></a>
-<span class="sourceLineNo">196</span>      public float getProgress() throws IOException, InterruptedException {<a name="line.196"></a>
-<span class="sourceLineNo">197</span>        return trr.getProgress();<a name="line.197"></a>
-<span class="sourceLineNo">198</span>      }<a name="line.198"></a>
-<span class="sourceLineNo">199</span><a name="line.199"></a>
-<span class="sourceLineNo">200</span>      @Override<a name="line.200"></a>
-<span class="sourceLineNo">201</span>      public void initialize(InputSplit inputsplit, TaskAttemptContext context) throws IOException,<a name="line.201"></a>
-<span class="sourceLineNo">202</span>          InterruptedException {<a name="line.202"></a>
-<span class="sourceLineNo">203</span>        trr.initialize(inputsplit, context);<a name="line.203"></a>
+<span class="sourceLineNo">191</span>      public void close() throws IOException {<a name="line.191"></a>
+<span class="sourceLineNo">192</span>        trr.close();<a name="line.192"></a>
+<span class="sourceLineNo">193</span>        closeTable();<a name="line.193"></a>
+<span class="sourceLineNo">194</span>      }<a name="line.194"></a>
+<span class="sourceLineNo">195</span><a name="line.195"></a>
+<span class="sourceLineNo">196</span>      @Override<a name="line.196"></a>
+<span class="sourceLineNo">197</span>      public ImmutableBytesWritable getCurrentKey() throws IOException, InterruptedException {<a name="line.197"></a>
+<span class="sourceLineNo">198</span>        return trr.getCurrentKey();<a name="line.198"></a>
+<span class="sourceLineNo">199</span>      }<a name="line.199"></a>
+<span class="sourceLineNo">200</span><a name="line.200"></a>
+<span class="sourceLineNo">201</span>      @Override<a name="line.201"></a>
+<span class="sourceLineNo">202</span>      public Result getCurrentValue() throws IOException, InterruptedException {<a name="line.202"></a>
+<span class="sourceLineNo">203</span>        return trr.getCurrentValue();<a name="line.203"></a>
 <span class="sourceLineNo">204</span>      }<a name="line.204"></a>
 <span class="sourceLineNo">205</span><a name="line.205"></a>
 <span class="sourceLineNo">206</span>      @Override<a name="line.206"></a>
-<span class="sourceLineNo">207</span>      public boolean nextKeyValue() throws IOException, InterruptedException {<a name="line.207"></a>
-<span class="sourceLineNo">208</span>        return trr.nextKeyValue();<a name="line.208"></a>
+<span class="sourceLineNo">207</span>      public float getProgress() throws IOException, InterruptedException {<a name="line.207"></a>
+<span class="sourceLineNo">208</span>        return trr.getProgress();<a name="line.208"></a>
 <span class="sourceLineNo">209</span>      }<a name="line.209"></a>
-<span class="sourceLineNo">210</span>    };<a name="line.210"></a>
-<span class="sourceLineNo">211</span>  }<a name="line.211"></a>
-<span class="sourceLineNo">212</span><a name="line.212"></a>
-<span class="sourceLineNo">213</span>  protected Pair&lt;byte[][],byte[][]&gt; getStartEndKeys() throws IOException {<a name="line.213"></a>
-<span class="sourceLineNo">214</span>    return getRegionLocator().getStartEndKeys();<a name="line.214"></a>
-<span class="sourceLineNo">215</span>  }<a name="line.215"></a>
+<span class="sourceLineNo">210</span><a name="line.210"></a>
+<span class="sourceLineNo">211</span>      @Override<a name="line.211"></a>
+<span class="sourceLineNo">212</span>      public void initialize(InputSplit inputsplit, TaskAttemptContext context) throws IOException,<a name="line.212"></a>
+<span class="sourceLineNo">213</span>          InterruptedException {<a name="line.213"></a>
+<span class="sourceLineNo">214</span>        trr.initialize(inputsplit, context);<a name="line.214"></a>
+<span class="sourceLineNo">215</span>      }<a name="line.215"></a>
 <span class="sourceLineNo">216</span><a name="line.216"></a>
-<span class="sourceLineNo">217</span>  /**<a name="line.217"></a>
-<span class="sourceLineNo">218</span>   * Calculates the splits that will serve as input for the map tasks. The<a name="line.218"></a>
-<span class="sourceLineNo">219</span>   * number of splits matches the number of regions in a table.<a name="line.219"></a>
-<span class="sourceLineNo">220</span>   *<a name="line.220"></a>
-<span class="sourceLineNo">221</span>   * @param context  The current job context.<a name="line.221"></a>
-<span class="sourceLineNo">222</span>   * @return The list of input splits.<a name="line.222"></a>
-<span class="sourceLineNo">223</span>   * @throws IOException When creating the list of splits fails.<a name="line.223"></a>
-<span class="sourceLineNo">224</span>   * @see org.apache.hadoop.mapreduce.InputFormat#getSplits(<a name="line.224"></a>
-<span class="sourceLineNo">225</span>   *   org.apache.hadoop.mapreduce.JobContext)<a name="line.225"></a>
-<span class="sourceLineNo">226</span>   */<a name="line.226"></a>
-<span class="sourceLineNo">227</span>  @Override<a name="line.227"></a>
-<span class="sourceLineNo">228</span>  public List&lt;InputSplit&gt; getSplits(JobContext context) throws IOException {<a name="line.228"></a>
-<span class="sourceLineNo">229</span>    boolean closeOnFinish = false;<a name="line.229"></a>
-<span class="sourceLineNo">230</span><a name="line.230"></a>
-<span class="sourceLineNo">231</span>    // Just in case a subclass is relying on JobConfigurable magic.<a name="line.231"></a>
-<span class="sourceLineNo">232</span>    if (table == null) {<a name="line.232"></a>
-<span class="sourceLineNo">233</span>      initialize(context);<a name="line.233"></a>
-<span class="sourceLineNo">234</span>      closeOnFinish = true;<a name="line.234"></a>
-<span class="sourceLineNo">235</span>    }<a name="line.235"></a>
-<span class="sourceLineNo">236</span><a name="line.236"></a>
-<span class="sourceLineNo">237</span>    // null check in case our child overrides getTable to not throw.<a name="line.237"></a>
-<span class="sourceLineNo">238</span>    try {<a name="line.238"></a>
-<span class="sourceLineNo">239</span>      if (getTable() == null) {<a name="line.239"></a>
-<span class="sourceLineNo">240</span>        // initialize() must not have been implemented in the subclass.<a name="line.240"></a>
-<span class="sourceLineNo">241</span>        throw new IOException(INITIALIZATION_ERROR);<a name="line.241"></a>
-<span class="sourceLineNo">242</span>      }<a name="line.242"></a>
-<span class="sourceLineNo">243</span>    } catch (IllegalStateException exception) {<a name="line.243"></a>
-<span class="sourceLineNo">244</span>      throw new IOException(INITIALIZATION_ERROR, exception);<a name="line.244"></a>
-<span class="sourceLineNo">245</span>    }<a name="line.245"></a>
-<span class="sourceLineNo">246</span><a name="line.246"></a>
+<span class="sourceLineNo">217</span>      @Override<a name="line.217"></a>
+<span class="sourceLineNo">218</span>      public boolean nextKeyValue() throws IOException, InterruptedException {<a name="line.218"></a>
+<span class="sourceLineNo">219</span>        return trr.nextKeyValue();<a name="line.219"></a>
+<span class="sourceLineNo">220</span>      }<a name="line.220"></a>
+<span class="sourceLineNo">221</span>    };<a name="line.221"></a>
+<span class="sourceLineNo">222</span>  }<a name="line.222"></a>
+<span class="sourceLineNo">223</span><a name="line.223"></a>
+<span class="sourceLineNo">224</span>  protected Pair&lt;byte[][],byte[][]&gt; getStartEndKeys() throws IOException {<a name="line.224"></a>
+<span class="sourceLineNo">225</span>    return getRegionLocator().getStartEndKeys();<a name="line.225"></a>
+<span class="sourceLineNo">226</span>  }<a name="line.226"></a>
+<span class="sourceLineNo">227</span><a name="line.227"></a>
+<span class="sourceLineNo">228</span>  /**<a name="line.228"></a>
+<span class="sourceLineNo">229</span>   * Calculates the splits that will serve as input for the map tasks.<a name="line.229"></a>
+<span class="sourceLineNo">230</span>   * @param context  The current job context.<a name="line.230"></a>
+<span class="sourceLineNo">231</span>   * @return The list of input splits.<a name="line.231"></a>
+<span class="sourceLineNo">232</span>   * @throws IOException When creating the list of splits fails.<a name="line.232"></a>
+<span class="sourceLineNo">233</span>   * @see org.apache.hadoop.mapreduce.InputFormat#getSplits(<a name="line.233"></a>
+<span class="sourceLineNo">234</span>   *   org.apache.hadoop.mapreduce.JobContext)<a name="line.234"></a>
+<span class="sourceLineNo">235</span>   */<a name="line.235"></a>
+<span class="sourceLineNo">236</span>  @Override<a name="line.236"></a>
+<span class="sourceLineNo">237</span>  public List&lt;InputSplit&gt; getSplits(JobContext context) throws IOException {<a name="line.237"></a>
+<span class="sourceLineNo">238</span>    boolean closeOnFinish = false;<a name="line.238"></a>
+<span class="sourceLineNo">239</span><a name="line.239"></a>
+<span class="sourceLineNo">240</span>    // Just in case a subclass is relying on JobConfigurable magic.<a name="line.240"></a>
+<span class="sourceLineNo">241</span>    if (table == null) {<a name="line.241"></a>
+<span class="sourceLineNo">242</span>      initialize(context);<a name="line.242"></a>
+<span class="sourceLineNo">243</span>      closeOnFinish = true;<a name="line.243"></a>
+<span class="sourceLineNo">244</span>    }<a name="line.244"></a>
+<span class="sourceLineNo">245</span><a name="line.245"></a>
+<span class="sourceLineNo">246</span>    // null check in case our child overrides getTable to not throw.<a name="line.246"></a>
 <span class="sourceLineNo">247</span>    try {<a name="line.247"></a>
-<span class="sourceLineNo">248</span>      RegionSizeCalculator sizeCalculator =<a name="line.248"></a>
-<span class="sourceLineNo">249</span>          new RegionSizeCalculator(getRegionLocator(), getAdmin());<a name="line.249"></a>
-<span class="sourceLineNo">250</span><a name="line.250"></a>
-<span class="sourceLineNo">251</span>      TableName tableName = getTable().getName();<a name="line.251"></a>
-<span class="sourceLineNo">252</span><a name="line.252"></a>
-<span class="sourceLineNo">253</span>      Pair&lt;byte[][], byte[][]&gt; keys = getStartEndKeys();<a name="line.253"></a>
-<span class="sourceLineNo">254</span>      if (keys == null || keys.getFirst() == null ||<a name="line.254"></a>
-<span class="sourceLineNo">255</span>          keys.getFirst().length == 0) {<a name="line.255"></a>
-<span class="sourceLineNo">256</span>        HRegionLocation regLoc =<a name="line.256"></a>
-<span class="sourceLineNo">257</span>            getRegionLocator().getRegionLocation(HConstants.EMPTY_BYTE_ARRAY, false);<a name="line.257"></a>
-<span class="sourceLineNo">258</span>        if (null == regLoc) {<a name="line.258"></a>
-<span class="sourceLineNo">259</span>          throw new IOException("Expecting at least one region.");<a name="line.259"></a>
-<span class="sourceLineNo">260</span>        }<a name="line.260"></a>
-<span class="sourceLineNo">261</span>        List&lt;InputSplit&gt; splits = new ArrayList&lt;&gt;(1);<a name="line.261"></a>
-<span class="sourceLineNo">262</span>        long regionSize = sizeCalculator.getRegionSize(regLoc.getRegionInfo().getRegionName());<a name="line.262"></a>
-<span class="sourceLineNo">263</span>        TableSplit split = new TableSplit(tableName, scan,<a name="line.263"></a>
-<span class="sourceLineNo">264</span>            HConstants.EMPTY_BYTE_ARRAY, HConstants.EMPTY_BYTE_ARRAY, regLoc<a name="line.264"></a>
-<span class="sourceLineNo">265</span>                .getHostnamePort().split(Addressing.HOSTNAME_PORT_SEPARATOR)[0], regionSize);<a name="line.265"></a>
-<span class="sourceLineNo">266</span>        splits.add(split);<a name="line.266"></a>
-<span class="sourceLineNo">267</span>        return splits;<a name="line.267"></a>
+<span class="sourceLineNo">248</span>      if (getTable() == null) {<a name="line.248"></a>
+<span class="sourceLineNo">249</span>        // initialize() must not have been implemented in the subclass.<a name="line.249"></a>
+<span class="sourceLineNo">250</span>        throw new IOException(INITIALIZATION_ERROR);<a name="line.250"></a>
+<span class="sourceLineNo">251</span>      }<a name="line.251"></a>
+<span class="sourceLineNo">252</span>    } catch (IllegalStateException exception) {<a name="line.252"></a>
+<span class="sourceLineNo">253</span>      throw new IOException(INITIALIZATION_ERROR, exception);<a name="line.253"></a>
+<span class="sourceLineNo">254</span>    }<a name="line.254"></a>
+<span class="sourceLineNo">255</span><a name="line.255"></a>
+<span class="sourceLineNo">256</span>    try {<a name="line.256"></a>
+<span class="sourceLineNo">257</span>      List&lt;InputSplit&gt; splits = oneInputSplitPerRegion();<a name="line.257"></a>
+<span class="sourceLineNo">258</span><a name="line.258"></a>
+<span class="sourceLineNo">259</span>      // set same number of mappers for each region<a name="line.259"></a>
+<span class="sourceLineNo">260</span>      if (context.getConfiguration().get(NUM_MAPPERS_PER_REGION) != null) {<a name="line.260"></a>
+<span class="sourceLineNo">261</span>        int nSplitsPerRegion = context.getConfiguration().getInt(NUM_MAPPERS_PER_REGION, 1);<a name="line.261"></a>
+<span class="sourceLineNo">262</span>        List&lt;InputSplit&gt; res = new ArrayList&lt;&gt;();<a name="line.262"></a>
+<span class="sourceLineNo">263</span>        for (int i = 0; i &lt; splits.size(); i++) {<a name="line.263"></a>
+<span class="sourceLineNo">264</span>          List&lt;InputSplit&gt; tmp = createNInputSplitsUniform(splits.get(i), nSplitsPerRegion);<a name="line.264"></a>
+<span class="sourceLineNo">265</span>          res.addAll(tmp);<a name="line.265"></a>
+<span class="sourceLineNo">266</span>        }<a name="line.266"></a>
+<span class="sourceLineNo">267</span>        return res;<a name="line.267"></a>
 <span class="sourceLineNo">268</span>      }<a name="line.268"></a>
-<span class="sourceLineNo">269</span>      List&lt;InputSplit&gt; splits = new ArrayList&lt;&gt;(keys.getFirst().length);<a name="line.269"></a>
-<span class="sourceLineNo">270</span>      for (int i = 0; i &lt; keys.getFirst().length; i++) {<a name="line.270"></a>
-<span class="sourceLineNo">271</span>        if (!includeRegionInSplit(keys.getFirst()[i], keys.getSecond()[i])) {<a name="line.271"></a>
-<span class="sourceLineNo">272</span>          continue;<a name="line.272"></a>
-<span class="sourceLineNo">273</span>        }<a name="line.273"></a>
-<span class="sourceLineNo">274</span><a name="line.274"></a>
-<span class="sourceLineNo">275</span>        byte[] startRow = scan.getStartRow();<a name="line.275"></a>
-<span class="sourceLineNo">276</span>        byte[] stopRow = scan.getStopRow();<a name="line.276"></a>
-<span class="sourceLineNo">277</span>        // determine if the given start an stop key fall into the region<a name="line.277"></a>
-<span class="sourceLineNo">278</span>        if ((startRow.length == 0 || keys.getSecond()[i].length == 0 ||<a name="line.278"></a>
-<span class="sourceLineNo">279</span>            Bytes.compareTo(startRow, keys.getSecond()[i]) &lt; 0) &amp;&amp;<a name="line.279"></a>
-<span class="sourceLineNo">280</span>            (stopRow.length == 0 ||<a name="line.280"></a>
-<span class="sourceLineNo">281</span>             Bytes.compareTo(stopRow, keys.getFirst()[i]) &gt; 0)) {<a name="line.281"></a>
-<span class="sourceLineNo">282</span>          byte[] splitStart = startRow.length == 0 ||<a name="line.282"></a>
-<span class="sourceLineNo">283</span>            Bytes.compareTo(keys.getFirst()[i], startRow) &gt;= 0 ?<a name="line.283"></a>
-<span class="sourceLineNo">284</span>              keys.getFirst()[i] : startRow;<a name="line.284"></a>
-<span class="sourceLineNo">285</span>          byte[] splitStop = (stopRow.length == 0 ||<a name="line.285"></a>
-<span class="sourceLineNo">286</span>            Bytes.compareTo(keys.getSecond()[i], stopRow) &lt;= 0) &amp;&amp;<a name="line.286"></a>
-<span class="sourceLineNo">287</span>            keys.getSecond()[i].length &gt; 0 ?<a name="line.287"></a>
-<span class="sourceLineNo">288</span>              keys.getSecond()[i] : stopRow;<a name="line.288"></a>
-<span class="sourceLineNo">289</span><a name="line.289"></a>
-<span class="sourceLineNo">290</span>          HRegionLocation location = getRegionLocator().getRegionLocation(keys.getFirst()[i], false);<a name="line.290"></a>
-<span class="sourceLineNo">291</span>          // The below InetSocketAddress creation does a name resolution.<a name="line.291"></a>
-<span class="sourceLineNo">292</span>          InetSocketAddress isa = new InetSocketAddress(location.getHostname(), location.getPort());<a name="line.292"></a>
-<span class="sourceLineNo">293</span>          if (isa.isUnresolved()) {<a name="line.293"></a>
-<span class="sourceLineNo">294</span>            LOG.warn("Failed resolve " + isa);<a name="line.294"></a>
-<span class="sourceLineNo">295</span>          }<a name="line.295"></a>
-<span class="sourceLineNo">296</span>          InetAddress regionAddress = isa.getAddress();<a name="line.296"></a>
-<span class="sourceLineNo">297</span>          String regionLocation;<a name="line.297"></a>
-<span class="sourceLineNo">298</span>          regionLocation = reverseDNS(regionAddress);<a name="line.298"></a>
-<span class="sourceLineNo">299</span><a name="line.299"></a>
-<span class="sourceLineNo">300</span>          byte[] regionName = location.getRegionInfo().getRegionName();<a name="line.300"></a>
-<span class="sourceLineNo">301</span>          String encodedRegionName = location.getRegionInfo().getEncodedName();<a name="line.301"></a>
-<span class="sourceLineNo">302</span>          long regionSize = sizeCalculator.getRegionSize(regionName);<a name="line.302"></a>
-<span class="sourceLineNo">303</span>          TableSplit split = new TableSplit(tableName, scan,<a name="line.303"></a>
-<span class="sourceLineNo">304</span>            splitStart, splitStop, regionLocation, encodedRegionName, regionSize);<a name="line.304"></a>
-<span class="sourceLineNo">305</span>          splits.add(split);<a name="line.305"></a>
-<span class="sourceLineNo">306</span>          if (LOG.isDebugEnabled()) {<a name="line.306"></a>
-<span class="sourceLineNo">307</span>            LOG.debug("getSplits: split -&gt; " + i + " -&gt; " + split);<a name="line.307"></a>
-<span class="sourceLineNo">308</span>          }<a name="line.308"></a>
-<span class="sourceLineNo">309</span>        }<a name="line.309"></a>
-<span class="sourceLineNo">310</span>      }<a name="line.310"></a>
-<span class="sourceLineNo">311</span>      //The default value of "hbase.mapreduce.input.autobalance" is false, which means not enabled.<a name="line.311"></a>
-<span class="sourceLineNo">312</span>      boolean enableAutoBalance = context.getConfiguration()<a name="line.312"></a>
-<span class="sourceLineNo">313</span>        .getBoolean(MAPREDUCE_INPUT_AUTOBALANCE, false);<a name="line.313"></a>
-<span class="sourceLineNo">314</span>      if (enableAutoBalance) {<a name="line.314"></a>
-<span class="sourceLineNo">315</span>        long totalRegionSize=0;<a name="line.315"></a>
-<span class="sourceLineNo">316</span>        for (int i = 0; i &lt; splits.size(); i++){<a name="line.316"></a>
-<span class="sourceLineNo">317</span>          TableSplit ts = (TableSplit)splits.get(i);<a name="line.317"></a>
-<span class="sourceLineNo">318</span>          totalRegionSize += ts.getLength();<a name="line.318"></a>
-<span class="sourceLineNo">319</span>        }<a name="line.319"></a>
-<span class="sourceLineNo">320</span>        long averageRegionSize = totalRegionSize / splits.size();<a name="line.320"></a>
-<span class="sourceLineNo">321</span>        // the averageRegionSize must be positive.<a name="line.321"></a>
-<span class="sourceLineNo">322</span>        if (averageRegionSize &lt;= 0) {<a name="line.322"></a>
-<span class="sourceLineNo">323</span>            LOG.warn("The averageRegionSize is not positive: "+ averageRegionSize + ", " +<a name="line.323"></a>
-<span class="sourceLineNo">324</span>                    "set it to 1.");<a name="line.324"></a>
-<span class="sourceLineNo">325</span>            averageRegionSize = 1;<a name="line.325"></a>
-<span class="sourceLineNo">326</span>        }<a name="line.326"></a>
-<span class="sourceLineNo">327</span>        return calculateRebalancedSplits(splits, context, averageRegionSize);<a name="line.327"></a>
-<span class="sourceLineNo">328</span>      } else {<a name="line.328"></a>
-<span class="sourceLineNo">329</span>        return splits;<a name="line.329"></a>
-<span class="sourceLineNo">330</span>      }<a name="line.330"></a>
-<span class="sourceLineNo">331</span>    } finally {<a name="line.331"></a>
-<span class="sourceLineNo">332</span>      if (closeOnFinish) {<a name="line.332"></a>
-<span class="sourceLineNo">333</span>        closeTable();<a name="line.333"></a>
-<span class="sourceLineNo">334</span>      }<a name="line.334"></a>
-<span class="sourceLineNo">335</span>    }<a name="line.335"></a>
-<span class="sourceLineNo">336</span>  }<a name="line.336"></a>
-<span class="sourceLineNo">337</span><a name="line.337"></a>
-<span class="sourceLineNo">338</span>  String reverseDNS(InetAddress ipAddress) throws UnknownHostException {<a name="line.338"></a>
-<span class="sourceLineNo">339</span>    String hostName = this.reverseDNSCacheMap.get(ipAddress);<a name="line.339"></a>
-<span class="sourceLineNo">340</span>    if (hostName == null) {<a name="line.340"></a>
-<span class="sourceLineNo">341</span>      String ipAddressString = null;<a name="line.341"></a>
-<span class="sourceLineNo">342</span>      try {<a name="line.342"></a>
-<span class="sourceLineNo">343</span>        ipAddressString = DNS.reverseDns(ipAddress, null);<a name="line.343"></a>
-<span class="sourceLineNo">344</span>      } catch (Exception e) {<a name="line.344"></a>
-<span class="sourceLineNo">345</span>        // We can use InetAddress in case the jndi failed to pull up the reverse DNS entry from the<a name="line.345"></a>
-<span class="sourceLineNo">346</span>        // name service. Also, in case of ipv6, we need to use the InetAddress since resolving<a name="line.346"></a>
-<span class="sourceLineNo">347</span>        // reverse DNS using jndi doesn't work well with ipv6 addresses.<a name="line.347"></a>
-<span class="sourceLineNo">348</span>        ipAddressString = InetAddress.getByName(ipAddress.getHostAddress()).getHostName();<a name="line.348"></a>
-<span class="sourceLineNo">349</span>      }<a name="line.349"></a>
-<span class="sourceLineNo">350</span>      if (ipAddressString == null) throw new UnknownHostException("No host found for " + ipAddress);<a name="line.350"></a>
-<span class="sourceLineNo">351</span>      hostName = Strings.domainNamePointerToHostName(ipAddressString);<a name="line.351"></a>
-<span class="sourceLineNo">352</span>      this.reverseDNSCacheMap.put(ipAddress, hostName);<a name="line.352"></a>
-<span class="sourceLineNo">353</span>    }<a name="line.353"></a>
-<span class="sourceLineNo">354</span>    return hostName;<a name="line.354"></a>
-<span class="sourceLineNo">355</span>  }<a name="line.355"></a>
-<span class="sourceLineNo">356</span><a name="line.356"></a>
-<span class="sourceLineNo">357</span>  /**<a name="line.357"></a>
-<span class="sourceLineNo">358</span>   * Calculates the number of MapReduce input splits for the map tasks. The number of<a name="line.358"></a>
-<span class="sourceLineNo">359</span>   * MapReduce input splits depends on the average region size and the "data skew ratio" user set in<a name="line.359"></a>
-<span class="sourceLineNo">360</span>   * configuration.<a name="line.360"></a>
-<span class="sourceLineNo">361</span>   *<a name="line.361"></a>
-<span class="sourceLineNo">362</span>   * @param list  The list of input splits before balance.<a name="line.362"></a>
-<span class="sourceLineNo">363</span>   * @param context  The current job context.<a name="line.363"></a>
-<span class="sourceLineNo">364</span>   * @param average  The average size of all regions .<a name="line.364"></a>
-<span class="sourceLineNo">365</span>   * @return The list of input splits.<a name="line.365"></a>
-<span class="sourceLineNo">366</span>   * @throws IOException When creating the list of splits fails.<a name="line.366"></a>
-<span class="sourceLineNo">367</span>   * @see org.apache.hadoop.mapreduce.InputFormat#getSplits(<a name="line.367"></a>
-<span class="sourceLineNo">368</span>   *   org.apache.hadoop.mapreduce.JobContext)<a name="line.368"></a>
-<span class="sourceLineNo">369</span>   */<a name="line.369"></a>
-<span class="sourceLineNo">370</span>  private List&lt;InputSplit&gt; calculateRebalancedSplits(List&lt;InputSplit&gt; list, JobContext context,<a name="line.370"></a>
-<span class="sourceLineNo">371</span>                                               long average) throws IOException {<a name="line.371"></a>
-<span class="sourceLineNo">372</span>    List&lt;InputSplit&gt; resultList = new ArrayList&lt;&gt;();<a name="line.372"></a>
-<span class="sourceLineNo">373</span>    Configuration conf = context.getConfiguration();<a name="line.373"></a>
-<span class="sourceLineNo">374</span>    //The default data skew ratio is 3<a name="line.374"></a>
-<span class="sourceLineNo">375</span>    long dataSkewRatio = conf.getLong(INPUT_AUTOBALANCE_MAXSKEWRATIO, 3);<a name="line.375"></a>
-<span class="sourceLineNo">376</span>    //It determines which mode to use: text key mode or binary key mode. The default is text mode.<a name="line.376"></a>
-<span class="sourceLineNo">377</span>    boolean isTextKey = context.getConfiguration().getBoolean(TABLE_ROW_TEXTKEY, true);<a name="line.377"></a>
-<span class="sourceLineNo">378</span>    long dataSkewThreshold = dataSkewRatio * average;<a name="line.378"></a>
-<span class="sourceLineNo">379</span>    int count = 0;<a name="line.379"></a>
-<span class="sourceLineNo">380</span>    while (count &lt; list.size()) {<a name="line.380"></a>
-<span class="sourceLineNo">381</span>      TableSplit ts = (TableSplit)list.get(count);<a name="line.381"></a>
-<span class="sourceLineNo">382</span>      TableName tableName = ts.getTable();<a name="line.382"></a>
-<span class="sourceLineNo">383</span>      String regionLocation = ts.getRegionLocation();<a name="line.383"></a>
-<span class="sourceLineNo">384</span>      String encodedRegionName = ts.getEncodedRegionName();<a name="line.384"></a>
-<span class="sourceLineNo">385</span>      long regionSize = ts.getLength();<a name="line.385"></a>
-<span class="sourceLineNo">386</span>      if (regionSize &gt;= dataSkewThreshold) {<a name="line.386"></a>
-<span class="sourceLineNo">387</span>        // if the current region size is large than the data skew threshold,<a name="line.387"></a>
-<span class="sourceLineNo">388</span>        // split the region into two MapReduce input splits.<a name="line.388"></a>
-<span class="sourceLineNo">389</span>        byte[] splitKey = getSplitKey(ts.getStartRow(), ts.getEndRow(), isTextKey);<a name="line.389"></a>
-<span class="sourceLineNo">390</span>        if (Arrays.equals(ts.getEndRow(), splitKey)) {<a name="line.390"></a>
-<span class="sourceLineNo">391</span>          // Not splitting since the end key is the same as the split key<a name="line.391"></a>
-<span class="sourceLineNo">392</span>          resultList.add(ts);<a name="line.392"></a>
-<span class="sourceLineNo">393</span>        } else {<a name="line.393"></a>
-<span class="sourceLineNo">394</span>          //Set the size of child TableSplit as 1/2 of the region size. The exact size of the<a name="line.394"></a>
-<span class="sourceLineNo">395</span>          // MapReduce input splits is not far off.<a name="line.395"></a>
-<span class="sourceLineNo">396</span>          TableSplit t1 = new TableSplit(tableName, scan, ts.getStartRow(), splitKey,<a name="line.396"></a>
-<span class="sourceLineNo">397</span>              regionLocation, regionSize / 2);<a name="line.397"></a>
-<span class="sourceLineNo">398</span>          TableSplit t2 = new TableSplit(tableName, scan, splitKey, ts.getEndRow(), regionLocation,<a name="line.398"></a>
-<span class="sourceLineNo">399</span>              regionSize - regionSize / 2);<a name="line.399"></a>
-<span class="sourceLineNo">400</span>          resultList.add(t1);<a name="line.400"></a>
-<span class="sourceLineNo">401</span>          resultList.add(t2);<a name="line.401"></a>
-<span class="sourceLineNo">402</span>        }<a name="line.402"></a>
-<span class="sourceLineNo">403</span>        count++;<a name="line.403"></a>
-<span class="sourceLineNo">404</span>      } else if (regionSize &gt;= average) {<a name="line.404"></a>
-<span class="sourceLineNo">405</span>        // if the region size between average size and data skew threshold size,<a name="line.405"></a>
-<span class="sourceLineNo">406</span>        // make this region as one MapReduce input split.<a name="line.406"></a>
-<span class="sourceLineNo">407</span>        resultList.add(ts);<a name="line.407"></a>
-<span class="sourceLineNo">408</span>        count++;<a name="line.408"></a>
-<span class="sourceLineNo">409</span>      } else {<a name="line.409"></a>
-<span class="sourceLineNo">410</span>        // if the total size of several small continuous regions less than the average region size,<a name="line.410"></a>
-<span class="sourceLineNo">411</span>        // combine them into one MapReduce input split.<a name="line.411"></a>
-<span class="sourceLineNo">412</span>        long totalSize = regionSize;<a name="line.412"></a>
-<span class="sourceLineNo">413</span>        byte[] splitStartKey = ts.getStartRow();<a name="line.413"></a>
-<span class="sourceLineNo">414</span>        byte[] splitEndKey = ts.getEndRow();<a name="line.414"></a>
-<span class="sourceLineNo">415</span>        count++;<a name="line.415"></a>
-<span class="sourceLineNo">416</span>        for (; count &lt; list.size(); count++) {<a name="line.416"></a>
-<span class="sourceLineNo">417</span>          TableSplit nextRegion = (TableSplit)list.get(count);<a name="line.417"></a>
-<span class="sourceLineNo">418</span>          long nextRegionSize = nextRegion.getLength();<a name="line.418"></a>
-<span class="sourceLineNo">419</span>          if (totalSize + nextRegionSize &lt;= dataSkewThreshold) {<a name="line.419"></a>
-<span class="sourceLineNo">420</span>            totalSize = totalSize + nextRegionSize;<a name="line.420"></a>
-<span class="sourceLineNo">421</span>            splitEndKey = nextRegion.getEndRow();<a name="line.421"></a>
-<span class="sourceLineNo">422</span>          } else {<a name="line.422"></a>
-<span class="sourceLineNo">423</span>            break;<a name="line.423"></a>
-<span class="sourceLineNo">424</span>          }<a name="line.424"></a>
-<span class="sourceLineNo">425</span>        }<a name="line.425"></a>
-<span class="sourceLineNo">426</span>        TableSplit t = new TableSplit(tableName, scan, splitStartKey, splitEndKey,<a name="line.426"></a>
-<span class="sourceLineNo">427</span>                regionLocation, encodedRegionName, totalSize);<a name="line.427"></a>
-<span class="sourceLineNo">428</span>        resultList.add(t);<a name="line.428"></a>
-<span class="sourceLineNo">429</span>      }<a name="line.429"></a>
-<span class="sourceLineNo">430</span>    }<a name="line.430"></a>
-<span class="sourceLineNo">431</span>    return resultList;<a name="line.431"></a>
-<span class="sourceLineNo">432</span>  }<a name="line.432"></a>
-<span class="sourceLineNo">433</span><a name="line.433"></a>
-<span class="sourceLineNo">434</span>  /**<a name="line.434"></a>
-<span class="sourceLineNo">435</span>   * select a split point in the region. The selection of the split point is based on an uniform<a name="line.435"></a>
-<span class="sourceLineNo">436</span>   * distribution assumption for the keys in a region.<a name="line.436"></a>
-<span class="sourceLineNo">437</span>   * Here are some examples:<a name="line.437"></a>
-<span class="sourceLineNo">438</span>   *<a name="line.438"></a>
-<span class="sourceLineNo">439</span>   * &lt;table&gt;<a name="line.439"></a>
-<span class="sourceLineNo">440</span>   *   &lt;tr&gt;<a name="line.440"></a>
-<span class="sourceLineNo">441</span>   *     &lt;th&gt;start key&lt;/th&gt;<a name="line.441"></a>
-<span class="sourceLineNo">442</span>   *     &lt;th&gt;end key&lt;/th&gt;<a name="line.442"></a>
-<span class="sourceLineNo">443</span>   *     &lt;th&gt;is text&lt;/th&gt;<a name="line.443"></a>
-<span class="sourceLineNo">444</span>   *     &lt;th&gt;split point&lt;/th&gt;<a name="line.444"></a>
-<span class="sourceLineNo">445</span>   *   &lt;/tr&gt;<a name="line.445"></a>
-<span class="sourceLineNo">446</span>   *   &lt;tr&gt;<a name="line.446"></a>
-<span class="sourceLineNo">447</span>   *     &lt;td&gt;'a', 'a', 'a', 'b', 'c', 'd', 'e', 'f', 'g'&lt;/td&gt;<a name="line.447"></a>
-<span class="sourceLineNo">448</span>   *     &lt;td&gt;'a', 'a', 'a', 'f', 'f', 'f'&lt;/td&gt;<a name="line.448"></a>
-<span class="sourceLineNo">449</span>   *     &lt;td&gt;true&lt;/td&gt;<a name="line.449"></a>
-<span class="sourceLineNo">450</span>   *     &lt;td&gt;'a', 'a', 'a', 'd', 'd', -78, 50, -77, 51&lt;/td&gt;<a name="line.450"></a>
-<span class="sourceLineNo">451</span>   *   &lt;/tr&gt;<a name="line.451"></a>
-<span class="sourceLineNo">452</span>   *   &lt;tr&gt;<a name="line.452"></a>
-<span class="sourceLineNo">453</span>   *     &lt;td&gt;'1', '1', '1', '0', '0', '0'&lt;/td&gt;<a name="line.453"></a>
-<span class="sourceLineNo">454</span>   *     &lt;td&gt;'1', '1', '2', '5', '7', '9', '0'&lt;/td&gt;<a name="line.454"></a>
-<span class="sourceLineNo">455</span>   *     &lt;td&gt;true&lt;/td&gt;<a name="line.455"></a>
-<span class="sourceLineNo">456</span>   *     &lt;td&gt;'1', '1', '1', -78, -77, -76, -104&lt;/td&gt;<a name="line.456"></a>
-<span class="sourceLineNo">457</span>   *   &lt;/tr&gt;<a name="line.457"></a>
-<span class="sourceLineNo">458</span>   *   &lt;tr&gt;<a name="line.458"></a>
-<span class="sourceLineNo">459</span>   *     &lt;td&gt;'1', '1', '1', '0'&lt;/td&gt;<a name="line.459"></a>
-<span class="sourceLineNo">460</span>   *     &lt;td&gt;'1', '1', '2', '0'&lt;/td&gt;<a name="line.460"></a>
-<span class="sourceLineNo">461</span>   *     &lt;td&gt;true&lt;/td&gt;<a name="line.461"></a>
-<span class="sourceLineNo">462</span>   *     &lt;td&gt;'1', '1', '1', -80&lt;/td&gt;<a name="line.462"></a>
-<span class="sourceLineNo">463</span>   *   &lt;/tr&gt;<a name="line.463"></a>
-<span class="sourceLineNo">464</span>   *   &lt;tr&gt;<a name="line.464"></a>
-<span class="sourceLineNo">465</span>   *     &lt;td&gt;13, -19, 126, 127&lt;/td&gt;<a name="line.465"></a>
-<span class="sourceLineNo">466</span>   *     &lt;td&gt;13, -19, 127, 0&lt;/td&gt;<a name="line.466"></a>
-<span class="sourceLineNo">467</span>   *     &lt;td&gt;false&lt;/td&gt;<a name="line.467"></a>
-<span class="sourceLineNo">468</span>   *     &lt;td&gt;13, -19, 126, -65&lt;/td&gt;<a name="line.468"></a>
-<span class="sourceLineNo">469</span>   *   &lt;/tr&gt;<a name="line.469"></a>
-<span class="sourceLineNo">470</span>   * &lt;/table&gt;<a name="line.470"></a>
-<span class="sourceLineNo">471</span>   *<a name="line.471"></a>
-<span class="sourceLineNo">472</span>   * Set this function as "public static", make it easier for test.<a name="line.472"></a>
-<span class="sourceLineNo">473</span>   *<a name="line.473"></a>
-<span class="sourceLineNo">474</span>   * @param start Start key of the region<a name="line.474"></a>
-<span class="sourceLineNo">475</span>   * @param end End key of the region<a name="line.475"></a>
-<span class="sourceLineNo">476</span>   * @param isText It determines to use text key mode or binary key mode<a name="line.476"></a>
-<span class="sourceLineNo">477</span>   * @return The split point in the region.<a name="line.477"></a>
-<span class="sourceLineNo">478</span>   */<a name="line.478"></a>
-<span class="sourceLineNo">479</span>  @InterfaceAudience.Private<a name="line.479"></a>
-<span class="sourceLineNo">480</span>  public static byte[] getSplitKey(byte[] start, byte[] end, boolean isText) {<a name="line.480"></a>
-<span class="sourceLineNo">481</span>    byte upperLimitByte;<a name="line.481"></a>
-<span class="sourceLineNo">482</span>    byte lowerLimitByte;<a name="line.482"></a>
-<span class="sourceLineNo">483</span>    //Use text mode or binary mode.<a name="line.483"></a>
-<span class="sourceLineNo">484</span>    if (isText) {<a name="line.484"></a>
-<span class="sourceLineNo">485</span>      //The range of text char set in ASCII is [32,126], the lower limit is space and the upper<a name="line.485"></a>
-<span class="sourceLineNo">486</span>      // limit is '~'.<a name="line.486"></a>
-<span class="sourceLineNo">487</span>      upperLimitByte = '~';<a name="line.487"></a>
-<span class="sourceLineNo">488</span>      lowerLimitByte = ' ';<a name="line.488"></a>
-<span class="sourceLineNo">489</span>    } else {<a name="line.489"></a>
-<span class="sourceLineNo">490</span>      upperLimitByte = -1;<a name="line.490"></a>
-<span class="sourceLineNo">491</span>      lowerLimitByte = 0;<a name="line.491"></a>
-<span class="sourceLineNo">492</span>    }<a name="line.492"></a>
-<span class="sourceLineNo">493</span>    // For special case<a name="line.493"></a>
-<span class="sourceLineNo">494</span>    // Example 1 : startkey=null, endkey="hhhqqqwww", splitKey="h"<a name="line.494"></a>
-<span class="sourceLineNo">495</span>    // Example 2 (text key mode): startKey="ffffaaa", endKey=null, splitkey="f~~~~~~"<a name="line.495"></a>
-<span class="sourceLineNo">496</span>    if (start.length == 0 &amp;&amp; end.length == 0){<a name="line.496"></a>
-<span class="sourceLineNo">497</span>      return new byte[]{(byte) ((lowerLimitByte + upperLimitByte) / 2)};<a name="line.497"></a>
-<span class="sourceLineNo">498</span>    }<a name="line.498"></a>
-<span class="sourceLineNo">499</span>    if (start.length == 0 &amp;&amp; end.length != 0){<a name="line.499"></a>
-<span class="sourceLineNo">500</span>      return new byte[]{ end[0] };<a name="line.500"></a>
-<span class="sourceLineNo">501</span>    }<a name="line.501"></a>
-<span class="sourceLineNo">502</span>    if (start.length != 0 &amp;&amp; end.length == 0){<a name="line.502"></a>
-<span class="sourceLineNo">503</span>      byte[] result =new byte[start.length];<a name="line.503"></a>
-<span class="sourceLineNo">504</span>      result[0]=start[0];<a name="line.504"></a>
-<span class="sourceLineNo">505</span>      for (int k = 1; k &lt; start.length; k++){<a name="line.505"></a>
-<span class="sourceLineNo">506</span>          result[k] = upperLimitByte;<a name="line.506"></a>
-<span class="sourceLineNo">507</span>      }<a name="line.507"></a>
-<span class="sourceLineNo">508</span>      return result;<a name="line.508"></a>
-<span class="sourceLineNo">509</span>    }<a name="line.509"></a>
-<span class="sourceLineNo">510</span>    return Bytes.split(start, end, false, 1)[1];<a name="line.510"></a>
-<span class="sourceLineNo">511</span>  }<a name="line.511"></a>
-<span class="sourceLineNo">512</span><a name="line.512"></a>
-<span class="sourceLineNo">513</span>  /**<a name="line.513"></a>
-<span class="sourceLineNo">514</span>   * Test if the given region is to be included in the InputSplit while splitting<a name="line.514"></a>
-<span class="sourceLineNo">515</span>   * the regions of a table.<a name="line.515"></a>
-<span class="sourceLineNo">516</span>   * &lt;p&gt;<a name="line.516"></a>
-<span class="sourceLineNo">517</span>   * This optimization is effective when there is a specific reasoning to exclude an entire region from the M-R job,<a name="line.517"></a>
-<span class="sourceLineNo">518</span>   * (and hence, not contributing to the InputSplit), given the start and end keys of the same. &lt;br&gt;<a name="line.518"></a>
-<span class="sourceLineNo">519</span>   * Useful when we need to remember the last-processed top record and revisit the [last, current) interval for M-R processing,<a name="line.519"></a>
-<span class="sourceLineNo">520</span>   * continuously. In addition to reducing InputSplits, reduces the load on the region server as well, due to the ordering of the keys.<a name="line.520"></a>
-<span class="sourceLineNo">521</span>   * &lt;br&gt;<a name="line.521"></a>
-<span class="sourceLineNo">522</span>   * &lt;br&gt;<a name="line.522"></a>
-<span class="sourceLineNo">523</span>   * Note: It is possible that &lt;code&gt;endKey.length() == 0 &lt;/code&gt; , for the last (recent) region.<a name="line.523"></a>
+<span class="sourceLineNo">269</span><a name="line.269"></a>
+<span class="sourceLineNo">270</span>      //The default value of "hbase.mapreduce.input.autobalance" is false.<a name="line.270"></a>
+<span class="sourceLineNo">271</span>      if (context.getConfiguration().getBoolean(MAPREDUCE_INPUT_AUTOBALANCE, false) != false) {<a name="line.271"></a>
+<span class="sourceLineNo">272</span>        long maxAveRegionSize = context.getConfiguration().getInt(MAX_AVERAGE_REGION_SIZE, 8*1073741824);<a name="line.272"></a>
+<span class="sourceLineNo">273</span>        return calculateAutoBalancedSplits(splits, maxAveRegionSize);<a name="line.273"></a>
+<span class="sourceLineNo">274</span>      }<a name="line.274"></a>
+<span class="sourceLineNo">275</span><a name="line.275"></a>
+<span class="sourceLineNo">276</span>      // return one mapper per region<a name="line.276"></a>
+<span class="sourceLineNo">277</span>      return splits;<a name="line.277"></a>
+<span class="sourceLineNo">278</span>    } finally {<a name="line.278"></a>
+<span class="sourceLineNo">279</span>      if (closeOnFinish) {<a name="line.279"></a>
+<span class="sourceLineNo">280</span>        closeTable();<a name="line.280"></a>
+<span class="sourceLineNo">281</span>      }<a name="line.281"></a>
+<span class="sourceLineNo">282</span>    }<a name="line.282"></a>
+<span class="sourceLineNo">283</span>  }<a name="line.283"></a>
+<span class="sourceLineNo">284</span><a name="line.284"></a>
+<span class="sourceLineNo">285</span>  /**<a name="line.285"></a>
+<span class="sourceLineNo">286</span>   * Create one InputSplit per region<a name="line.286"></a>
+<span class="sourceLineNo">287</span>   *<a name="line.287"></a>
+<span class="sourceLineNo">288</span>   * @return The list of InputSplit for all the regions<a name="line.288"></a>
+<span class="sourceLineNo">289</span>   * @throws IOException<a name="line.289"></a>
+<span class="sourceLineNo">290</span>   */<a name="line.290"></a>
+<span class="sourceLineNo">291</span>  private List&lt;InputSplit&gt; oneInputSplitPerRegion() throws IOException {<a name="line.291"></a>
+<span class="sourceLineNo">292</span>    RegionSizeCalculator sizeCalculator =<a name="line.292"></a>
+<span class="sourceLineNo">293</span>        new RegionSizeCalculator(getRegionLocator(), getAdmin());<a name="line.293"></a>
+<span class="sourceLineNo">294</span><a name="line.294"></a>
+<span class="sourceLineNo">295</span>    TableName tableName = getTable().getName();<a name="line.295"></a>
+<span class="sourceLineNo">296</span><a name="line.296"></a>
+<span class="sourceLineNo">297</span>    Pair&lt;byte[][], byte[][]&gt; keys = getStartEndKeys();<a name="line.297"></a>
+<span class="sourceLineNo">298</span>    if (keys == null || keys.getFirst() == null ||<a name="line.298"></a>
+<span class="sourceLineNo">299</span>        keys.getFirst().length == 0) {<a name="line.299"></a>
+<span class="sourceLineNo">300</span>      HRegionLocation regLoc =<a name="line.300"></a>
+<span class="sourceLineNo">301</span>          getRegionLocator().getRegionLocation(HConstants.EMPTY_BYTE_ARRAY, false);<a name="line.301"></a>
+<span class="sourceLineNo">302</span>      if (null == regLoc) {<a name="line.302"></a>
+<span class="sourceLineNo">303</span>        throw new IOException("Expecting at least one region.");<a name="line.303"></a>
+<span class="sourceLineNo">304</span>      }<a name="line.304"></a>
+<span class="sourceLineNo">305</span>      List&lt;InputSplit&gt; splits = new ArrayList&lt;&gt;(1);<a name="line.305"></a>
+<span class="sourceLineNo">306</span>      long regionSize = sizeCalculator.getRegionSize(regLoc.getRegionInfo().getRegionName());<a name="line.306"></a>
+<span class="sourceLineNo">307</span>      TableSplit split = new TableSplit(tableName, scan,<a name="line.307"></a>
+<span class="sourceLineNo">308</span>          HConstants.EMPTY_BYTE_ARRAY, HConstants.EMPTY_BYTE_ARRAY, regLoc<a name="line.308"></a>
+<span class="sourceLineNo">309</span>          .getHostnamePort().split(Addressing.HOSTNAME_PORT_SEPARATOR)[0], regionSize);<a name="line.309"></a>
+<span class="sourceLineNo">310</span>      splits.add(split);<a name="line.310"></a>
+<span class="sourceLineNo">311</span>      return splits;<a name="line.311"></a>
+<span class="sourceLineNo">312</span>    }<a name="line.312"></a>
+<span class="sourceLineNo">313</span>    List&lt;InputSplit&gt; splits = new ArrayList&lt;&gt;(keys.getFirst().length);<a name="line.313"></a>
+<span class="sourceLineNo">314</span>    for (int i = 0; i &lt; keys.getFirst().length; i++) {<a name="line.314"></a>
+<span class="sourceLineNo">315</span>      if (!includeRegionInSplit(keys.getFirst()[i], keys.getSecond()[i])) {<a name="line.315"></a>
+<span class="sourceLineNo">316</span>        continue;<a name="line.316"></a>
+<span class="sourceLineNo">317</span>      }<a name="line.317"></a>
+<span class="sourceLineNo">318</span><a name="line.318"></a>
+<span class="sourceLineNo">319</span>      byte[] startRow = scan.getStartRow();<a name="line.319"></a>
+<span class="sourceLineNo">320</span>      byte[] stopRow = scan.getStopRow();<a name="line.320"></a>
+<span class="sourceLineNo">321</span>      // determine if the given start an stop key fall into the region<a name="line.321"></a>
+<span class="sourceLineNo">322</span>      if ((startRow.length == 0 || keys.getSecond()[i].length == 0 ||<a name="line.322"></a>
+<span class="sourceLineNo">323</span>          Bytes.compareTo(startRow, keys.getSecond()[i]) &lt; 0) &amp;&amp;<a name="line.323"></a>
+<span class="sourceLineNo">324</span>          (stopRow.length == 0 ||<a name="line.324"></a>
+<span class="sourceLineNo">325</span>              Bytes.compareTo(stopRow, keys.getFirst()[i]) &gt; 0)) {<a name="line.325"></a>
+<span class="sourceLineNo">326</span>        byte[] splitStart = startRow.length == 0 ||<a name="line.326"></a>
+<span class="sourceLineNo">327</span>            Bytes.compareTo(keys.getFirst()[i], startRow) &gt;= 0 ?<a name="line.327"></a>
+<span class="sourceLineNo">328</span>            keys.getFirst()[i] : startRow;<a name="line.328"></a>
+<span class="sourceLineNo">329</span>        byte[] splitStop = (stopRow.length == 0 ||<a name="line.329"></a>
+<span class="sourceLineNo">330</span>            Bytes.compareTo(keys.getSecond()[i], stopRow) &lt;= 0) &amp;&amp;<a name="line.330"></a>
+<span class="sourceLineNo">331</span>            keys.getSecond()[i].length &gt; 0 ?<a name="line.331"></a>
+<span class="sourceLineNo">332</span>            keys.getSecond()[i] : stopRow;<a name="line.332"></a>
+<span class="sourceLineNo">333</span><a name="line.333"></a>
+<span class="sourceLineNo">334</span>        HRegionLocation location = getRegionLocator().getRegionLocation(keys.getFirst()[i], false);<a name="line.334"></a>
+<span class="sourceLineNo">335</span>        // The below InetSocketAddress creation does a name resolution.<a name="line.335"></a>
+<span class="sourceLineNo">336</span>        InetSocketAddress isa = new InetSocketAddress(location.getHostname(), location.getPort());<a name="line.336"></a>
+<span class="sourceLineNo">337</span>        if (isa.isUnresolved()) {<a name="line.337"></a>
+<span class="sourceLineNo">338</span>          LOG.warn("Failed resolve " + isa);<a name="line.338"></a>
+<span class="sourceLineNo">339</span>        }<a name="line.339"></a>
+<span class="sourceLineNo">340</span>        InetAddress regionAddress = isa.getAddress();<a name="line.340"></a>
+<span class="sourceLineNo">341</span>        String regionLocation;<a name="line.341"></a>
+<span class="sourceLineNo">342</span>        regionLocation = reverseDNS(regionAddress);<a name="line.342"></a>
+<span class="sourceLineNo">343</span><a name="line.343"></a>
+<span class="sourceLineNo">344</span>        byte[] regionName = location.getRegionInfo().getRegionName();<a name="line.344"></a>
+<span class="sourceLineNo">345</span>        String encodedRegionName = location.getRegionInfo().getEncodedName();<a name="line.345"></a>
+<span class="sourceLineNo">346</span>        long regionSize = sizeCalculator.getRegionSize(regionName);<a name="line.346"></a>
+<span class="sourceLineNo">347</span>        TableSplit split = new TableSplit(tableName, scan,<a name="line.347"></a>
+<span class="sourceLineNo">348</span>            splitStart, splitStop, regionLocation, encodedRegionName, regionSize);<a name="line.348"></a>
+<span class="sourceLineNo">349</span>        splits.add(split);<a name="line.349"></a>
+<span class="sourceLineNo">350</span>        if (LOG.isDebugEnabled()) {<a name="line.350"></a>
+<span class="sourceLineNo">351</span>          LOG.debug("getSplits: split -&gt; " + i + " -&gt; " + split);<a name="line.351"></a>
+<span class="sourceLineNo">352</span>        }<a name="line.352"></a>
+<span class="sourceLineNo">353</span>      }<a name="line.353"></a>
+<span class="sourceLineNo">354</span>    }<a name="line.354"></a>
+<span class="sourceLineNo">355</span>    return splits;<a name="line.355"></a>
+<span class="sourceLineNo">356</span>  }<a name="line.356"></a>
+<span class="sourceLineNo">357</span><a name="line.357"></a>
+<span class="sourceLineNo">358</span>  /**<a name="line.358"></a>
+<span class="sourceLineNo">359</span>   * Create n splits for one InputSplit, For now only support uniform distribution<a name="line.359"></a>
+<span class="sourceLineNo">360</span>   * @param split A TableSplit corresponding to a range of rowkeys<a name="line.360"></a>
+<span class="sourceLineNo">361</span>   * @param n     Number of ranges after splitting.  Pass 1 means no split for the range<a name="line.361"></a>
+<span class="sourceLineNo">362</span>   *              Pass 2 if you want to split the range in two;<a name="line.362"></a>
+<span class="sourceLineNo">363</span>   * @return A list of TableSplit, the size of the list is n<a name="line.363"></a>
+<span class="sourceLineNo">364</span>   * @throws IllegalArgumentIOException<a name="line.364"></a>
+<span class="sourceLineNo">365</span>   */<a name="line.365"></a>
+<span class="sourceLineNo">366</span>  protected List&lt;InputSplit&gt; createNInputSplitsUniform(InputSplit split, int n)<a name="line.366"></a>
+<span class="sourceLineNo">367</span>      throws IllegalArgumentIOException {<a name="line.367"></a>
+<span class="sourceLineNo">368</span>    if (split == null || !(split instanceof TableSplit)) {<a name="line.368"></a>
+<span class="sourceLineNo">369</span>      throw new IllegalArgumentIOException(<a name="line.369"></a>
+<span class="sourceLineNo">370</span>          "InputSplit for CreateNSplitsPerRegion can not be null + "<a name="line.370"></a>
+<span class="sourceLineNo">371</span>              + "and should be instance of TableSplit");<a name="line.371"></a>
+<span class="sourceLineNo">372</span>    }<a name="line.372"></a>
+<span class="sourceLineNo">373</span>    //if n &lt; 1, then still continue using n = 1<a name="line.373"></a>
+<span class="sourceLineNo">374</span>    n = n &lt; 1 ? 1 : n;<a name="line.374"></a>
+<span class="sourceLineNo">375</span>    List&lt;InputSplit&gt; res = new ArrayList&lt;&gt;(n);<a name="line.375"></a>
+<span class="sourceLineNo">376</span>    if (n == 1) {<a name="line.376"></a>
+<span class="sourceLineNo">377</span>      res.add(split);<a name="line.377"></a>
+<span class="sourceLineNo">378</span>      return res;<a name="line.378"></a>
+<span class="sourceLineNo">379</span>    }<a name="line.379"></a>
+<span class="sourceLineNo">380</span><a name="line.380"></a>
+<span class="sourceLineNo">381</span>    // Collect Region related information<a name="line.381"></a>
+<span class="sourceLineNo">382</span>    TableSplit ts = (TableSplit) split;<a name="line.382"></a>
+<span class="sourceLineNo">383</span>    TableName tableName = ts.getTable();<a name="line.383"></a>
+<span class="sourceLineNo">384</span>    String regionLocation = ts.getRegionLocation();<a name="line.384"></a>
+<span class="sourceLineNo">385</span>    String encodedRegionName = ts.getEncodedRegionName();<a name="line.385"></a>
+<span class="sourceLineNo">386</span>    long regionSize = ts.getLength();<a name="line.386"></a>
+<span class="sourceLineNo">387</span>    byte[] startRow = ts.getStartRow();<a name="line.387"></a>
+<span class="sourceLineNo">388</span>    byte[] endRow = ts.getEndRow();<a name="line.388"></a>
+<span class="sourceLineNo">389</span><a name="line.389"></a>
+<span class="sourceLineNo">390</span>    // For special case: startRow or endRow is empty<a name="line.390"></a>
+<span class="sourceLineNo">391</span>    if (startRow.length == 0 &amp;&amp; endRow.length == 0){<a name="line.391"></a>
+<span class="sourceLineNo">392</span>      startRow = new byte[1];<a name="line.392"></a>
+<span class="sourceLineNo">393</span>      endRow = new byte[1];<a name="line.393"></a>
+<span class="sourceLineNo">394</span>      startRow[0] = 0;<a name="line.394"></a>
+<span class="sourceLineNo">395</span>      endRow[0] = -1;<a name="line.395"></a>
+<span class="sourceLineNo">396</span>    }<a name="line.396"></a>
+<span class="sourceLineNo">397</span>    if (startRow.length == 0 &amp;&amp; endRow.length != 0){<a name="line.397"></a>
+<span class="sourceLineNo">398</span>      startRow = new byte[1];<a name="line.398"></a>
+<span class="sourceLineNo">399</span>      startRow[0] = 0;<a name="line.399"></a>
+<span class="sourceLineNo">400</span>    }<a name="line.400"></a>
+<span class="sourceLineNo">401</span>    if (startRow.length != 0 &amp;&amp; endRow.length == 0){<a name="line.401"></a>
+<span class="sourceLineNo">402</span>      endRow =new byte[startRow.length];<a name="line.402"></a>
+<span class="sourceLineNo">403</span>      for (int k = 0; k &lt; startRow.length; k++){<a name="line.403"></a>
+<span class="sourceLineNo">404</span>        endRow[k] = -1;<a name="line.404"></a>
+<span class="sourceLineNo">405</span>      }<a name="line.405"></a>
+<span class="sourceLineNo">406</span>    }<a name="line.406"></a>
+<span class="sourceLineNo">407</span><a name="line.407"></a>
+<span class="sourceLineNo">408</span>    // Split Region into n chunks evenly<a name="line.408"></a>
+<span class="sourceLineNo">409</span>    byte[][] splitKeys = Bytes.split(startRow, endRow, true, n-1);<a name="line.409"></a>
+<span class="sourceLineNo">410</span>    for (int i = 0; i &lt; splitKeys.length - 1; i++) {<a name="line.410"></a>
+<span class="sourceLineNo">411</span>      //notice that the regionSize parameter may be not very accurate<a name="line.411"></a>
+<span class="sourceLineNo">412</span>      TableSplit tsplit =<a name="line.412"></a>
+<span class="sourceLineNo">413</span>          new TableSplit(tableName, scan, splitKeys[i], splitKeys[i + 1], regionLocation,<a name="line.413"></a>
+<span class="sourceLineNo">414</span>              encodedRegionName, regionSize / n);<a name="line.414"></a>
+<span class="sourceLineNo">415</span>      res.add(tsplit);<a name="line.415"></a>
+<span class="sourceLineNo">416</span>    }<a name="line.416"></a>
+<span class="sourceLineNo">417</span>    return res;<a name="line.417"></a>
+<span class="sourceLineNo">418</span>  }<a name="line.418"></a>
+<span class="sourceLineNo">419</span>  /**<a name="line.419"></a>
+<span class="sourceLineNo">420</span>   * Calculates the number of MapReduce input splits for the map tasks. The number of<a name="line.420"></a>
+<span class="sourceLineNo">421</span>   * MapReduce input splits depends on the average region size.<a name="line.421"></a>
+<span class="sourceLineNo">422</span>   * Make it 'public' for testing<a name="line.422"></a>
+<span class="sourceLineNo">423</span>   *<a name="line.423"></a>
+<span class="sourceLineNo">424</span>   * @param splits The list of input splits before balance.<a name="line.424"></a>
+<span class="sourceLineNo">425</span>   * @param maxAverageRegionSize max Average region size for one mapper<a name="line.425"></a>
+<span class="sourceLineNo">426</span>   * @return The list of input splits.<a name="line.426"></a>
+<span class="sourceLineNo">427</span>   * @throws IOException When creating the list of splits fails.<a name="line.427"></a>
+<span class="sourceLineNo">428</span>   * @see org.apache.hadoop.mapreduce.InputFormat#getSplits(<a name="line.428"></a>
+<span class="sourceLineNo">429</span>   *org.apache.hadoop.mapreduce.JobContext)<a name="line.429"></a>
+<span class="sourceLineNo">430</span>   */<a name="line.430"></a>
+<span class="sourceLineNo">431</span>  public List&lt;InputSplit&gt; calculateAutoBalancedSplits(List&lt;InputSplit&gt; splits, long maxAverageRegionSize)<a name="line.431"></a>
+<span class="sourceLineNo">432</span>      throws IOException {<a name="line.432"></a>
+<span class="sourceLineNo">433</span>    if (splits.size() == 0) {<a name="line.433"></a>
+<span class="sourceLineNo">434</span>      return splits;<a name="line.434"></a>
+<span class="sourceLineNo">435</span>    }<a name="line.435"></a>
+<span class="sourceLineNo">436</span>    List&lt;InputSplit&gt; resultList = new ArrayList&lt;&gt;();<a name="line.436"></a>
+<span class="sourceLineNo">437</span>    long totalRegionSize = 0;<a name="line.437"></a>
+<span class="sourceLineNo">438</span>    for (int i = 0; i &lt; splits.size(); i++) {<a name="line.438"></a>
+<span class="sourceLineNo">439</span>      TableSplit ts = (TableSplit) splits.get(i);<a name="line.439"></a>
+<span class="sourceLineNo">440</span>      totalRegionSize += ts.getLength();<a name="line.440"></a>
+<span class="sourceLineNo">441</span>    }<a name="line.441"></a>
+<span class="sourceLineNo">442</span>    long averageRegionSize = totalRegionSize / splits.size();<a name="line.442"></a>
+<span class="sourceLineNo">443</span>    // totalRegionSize might be overflow, and the averageRegionSize must be positive.<a name="line.443"></a>
+<span class="sourceLineNo">444</span>    if (averageRegionSize &lt;= 0) {<a name="line.444"></a>
+<span class="sourceLineNo">445</span>      LOG.warn("The averageRegionSize is not positive: " + averageRegionSize + ", " +<a name="line.445"></a>
+<span class="sourceLineNo">446</span>          "set it to Long.MAX_VALUE " + splits.size());<a name="line.446"></a>
+<span class="sourceLineNo">447</span>      averageRegionSize = Long.MAX_VALUE / splits.size();<a name="line.447"></a>
+<span class="sourceLineNo">448</span>    }<a name="line.448"></a>
+<span class="sourceLineNo">449</span>    //if averageRegionSize is too big, change it to default as 1 GB,<a name="line.449"></a>
+<span class="sourceLineNo">450</span>    if (averageRegionSize &gt; maxAverageRegionSize) {<a name="line.450"></a>
+<span class="sourceLineNo">451</span>      averageRegionSize = maxAverageRegionSize;<a name="line.451"></a>
+<span class="sourceLineNo">452</span>    }<a name="line.452"></a>
+<span class="sourceLineNo">453</span>    // if averageRegionSize is too small, we do not need to allocate more mappers for those 'large' region<a name="line.453"></a>
+<span class="sourceLineNo">454</span>    // set default as 16M = (default hdfs block size) / 4;<a name="line.454"></a>
+<span class="sourceLineNo">455</span>    if (averageRegionSize &lt; 16 * 1048576) {<a name="line.455"></a>
+<span class="sourceLineNo">456</span>      return splits;<a name="line.456"></a>
+<span class="sourceLineNo">457</span>    }<a name="line.457"></a>
+<span class="sourceLineNo">458</span>    for (int i = 0; i &lt; splits.size(); i++) {<a name="line.458"></a>
+<span class="sourceLineNo">459</span>      TableSplit ts = (TableSplit) splits.get(i);<a name="line.459"></a>
+<span class="sourceLineNo">460</span>      TableName tableName = ts.getTable();<a name="line.460"></a>
+<span class="sourceLineNo">461</span>      String regionLocation = ts.getRegionLocation();<a name="line.461"></a>
+<span class="sourceLineNo">462</span>      String encodedRegionName = ts.getEncodedRegionName();<a name="line.462"></a>
+<span class="sourceLineNo">463</span>      long regionSize = ts.getLength();<a name="line.463"></a>
+<span class="sourceLineNo">464</span><a name="line.464"></a>
+<span class="sourceLineNo">465</span>      if (regionSize &gt;= averageRegionSize) {<a name="line.465"></a>
+<span class="sourceLineNo">466</span>        // make this region as multiple MapReduce input split.<a name="line.466"></a>
+<span class="sourceLineNo">467</span>        int n = (int) Math.round(Math.log(((double) regionSize) / ((double) averageRegionSize)) + 1.0);<a name="line.467"></a>
+<span class="sourceLineNo">468</span>        List&lt;InputSplit&gt; temp = createNInputSplitsUniform(ts, n);<a name="line.468"></a>
+<span class="sourceLineNo">469</span>        resultList.addAll(temp);<a name="line.469"></a>
+<span class="sourceLineNo">470</span>      } else {<a name="line.470"></a>
+<span class="sourceLineNo">471</span>        // if the total size of several small continuous regions less than the average region size,<a name="line.471"></a>
+<span class="sourceLineNo">472</span>        // combine them into one MapReduce input split.<a name="line.472"></a>
+<span class="sourceLineNo">473</span>        long totalSize = regionSize;<a name="line.473"></a>
+<span class="sourceLineNo">474</span>        byte[] splitStartKey = ts.getStartRow();<a name="line.474"></a>
+<span class="sourceLineNo">475</span>        byte[] splitEndKey = ts.getEndRow();<a name="line.475"></a>
+<span class="sourceLineNo">476</span>        int j = i + 1;<a name="line.476"></a>
+<span class="sourceLineNo">477</span>        while (j &lt; splits.size()) {<a name="line.477"></a>
+<span class="sourceLineNo">478</span>          TableSplit nextRegion = (TableSplit) splits.get(j);<a name="line.478"></a>
+<span class="sourceLineNo">479</span>          long nextRegionSize = nextRegion.getLength();<a name="line.479"></a>
+<span class="sourceLineNo">480</span>          if (totalSize + nextRegionSize &lt;= averageRegionSize) {<a name="line.480"></a>
+<span class="sourceLineNo">481</span>            totalSize = totalSize + nextRegionSize;<a name="line.481"></a>
+<span class="sourceLineNo">482</span>            splitEndKey = nextRegion.getEndRow();<a name="line.482"></a>
+<span class="sourceLineNo">483</span>            j++;<a name="line.483"></a>
+<span class="sourceLineNo">484</span>          } else {<a name="line.484"></a>
+<span class="sourceLineNo">485</span>            break;<a name="line.485"></a>
+<span class="sourceLineNo">486</span>          }<a name="line.486"></a>
+<span class="sourceLineNo">487</span>        }<a name="line.487"></a>
+<span class="sourceLineNo">488</span>        i = j - 1;<a name="line.488"></a>
+<span class="sourceLineNo">489</span>        TableSplit t = new TableSplit(tableName, scan, splitStartKey, splitEndKey, regionLocation,<a name="line.489"></a>
+<span class="sourceLineNo">490</span>            encodedRegionName, totalSize);<a name="line.490"></a>
+<span class="sourceLineNo">491</span>        resultList.add(t);<a name="line.491"></a>
+<span class="sourceLineNo">492</span>      }<a name="line.492"></a>
+<span class="sourceLineNo">493</span>    }<a name="line.493"></a>
+<span class="sourceLineNo">494</span>    return resultList;<a name="line.494"></a>
+<span class="sourceLineNo">495</span>  }<a name="line.495"></a>
+<span class="sourceLineNo">496</span><a name="line.496"></a>
+<span class="sourceLineNo">497</span>  String reverseDNS(InetAddress ipAddress) throws UnknownHostException {<a name="line.497"></a>
+<span class="sourceLineNo">498</span>    String hostName = this.reverseDNSCacheMap.get(ipAddress);<a name="line.498"></a>
+<span class="sourceLineNo">499</span>    if (hostName == null) {<a name="line.499"></a>
+<span class="sourceLineNo">500</span>      String ipAddressString = null;<a name="line.500"></a>
+<span class="sourceLineNo">501</span>      try {<a name="line.501"></a>
+<span class="sourceLineNo">5

<TRUNCATED>

Mime
View raw message