hbase-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From git-site-r...@apache.org
Subject [10/51] [partial] hbase-site git commit: Published site at .
Date Sat, 30 Sep 2017 15:13:30 GMT
http://git-wip-us.apache.org/repos/asf/hbase-site/blob/d41f56fe/devapidocs/src-html/org/apache/hadoop/hbase/util/RegionSplitter.html
----------------------------------------------------------------------
diff --git a/devapidocs/src-html/org/apache/hadoop/hbase/util/RegionSplitter.html b/devapidocs/src-html/org/apache/hadoop/hbase/util/RegionSplitter.html
index c78b462..84a35ea 100644
--- a/devapidocs/src-html/org/apache/hadoop/hbase/util/RegionSplitter.html
+++ b/devapidocs/src-html/org/apache/hadoop/hbase/util/RegionSplitter.html
@@ -57,1071 +57,1142 @@
 <span class="sourceLineNo">049</span>import org.apache.hadoop.hbase.ClusterStatus;<a name="line.49"></a>
 <span class="sourceLineNo">050</span>import org.apache.hadoop.hbase.ClusterStatus.Option;<a name="line.50"></a>
 <span class="sourceLineNo">051</span>import org.apache.hadoop.hbase.HBaseConfiguration;<a name="line.51"></a>
-<span class="sourceLineNo">052</span>import org.apache.hadoop.hbase.HRegionInfo;<a name="line.52"></a>
-<span class="sourceLineNo">053</span>import org.apache.hadoop.hbase.HRegionLocation;<a name="line.53"></a>
-<span class="sourceLineNo">054</span>import org.apache.hadoop.hbase.MetaTableAccessor;<a name="line.54"></a>
-<span class="sourceLineNo">055</span>import org.apache.hadoop.hbase.ServerName;<a name="line.55"></a>
-<span class="sourceLineNo">056</span>import org.apache.hadoop.hbase.TableName;<a name="line.56"></a>
-<span class="sourceLineNo">057</span>import org.apache.yetus.audience.InterfaceAudience;<a name="line.57"></a>
-<span class="sourceLineNo">058</span>import org.apache.hadoop.hbase.client.Admin;<a name="line.58"></a>
-<span class="sourceLineNo">059</span>import org.apache.hadoop.hbase.client.ClusterConnection;<a name="line.59"></a>
-<span class="sourceLineNo">060</span>import org.apache.hadoop.hbase.client.Connection;<a name="line.60"></a>
-<span class="sourceLineNo">061</span>import org.apache.hadoop.hbase.client.ConnectionFactory;<a name="line.61"></a>
-<span class="sourceLineNo">062</span>import org.apache.hadoop.hbase.client.NoServerForRegionException;<a name="line.62"></a>
-<span class="sourceLineNo">063</span>import org.apache.hadoop.hbase.client.RegionLocator;<a name="line.63"></a>
-<span class="sourceLineNo">064</span>import org.apache.hadoop.hbase.client.Table;<a name="line.64"></a>
-<span class="sourceLineNo">065</span>import org.apache.hadoop.hbase.regionserver.HRegionFileSystem;<a name="line.65"></a>
-<span class="sourceLineNo">066</span><a name="line.66"></a>
-<span class="sourceLineNo">067</span>import org.apache.hadoop.hbase.client.ColumnFamilyDescriptorBuilder;<a name="line.67"></a>
-<span class="sourceLineNo">068</span>import org.apache.hadoop.hbase.client.TableDescriptorBuilder;<a name="line.68"></a>
-<span class="sourceLineNo">069</span>import org.apache.hadoop.hbase.shaded.com.google.common.base.Preconditions;<a name="line.69"></a>
-<span class="sourceLineNo">070</span>import org.apache.hadoop.hbase.shaded.com.google.common.collect.Lists;<a name="line.70"></a>
-<span class="sourceLineNo">071</span>import org.apache.hadoop.hbase.shaded.com.google.common.collect.Maps;<a name="line.71"></a>
-<span class="sourceLineNo">072</span>import org.apache.hadoop.hbase.shaded.com.google.common.collect.Sets;<a name="line.72"></a>
-<span class="sourceLineNo">073</span><a name="line.73"></a>
-<span class="sourceLineNo">074</span>/**<a name="line.74"></a>
-<span class="sourceLineNo">075</span> * The {@link RegionSplitter} class provides several utilities to help in the<a name="line.75"></a>
-<span class="sourceLineNo">076</span> * administration lifecycle for developers who choose to manually split regions<a name="line.76"></a>
-<span class="sourceLineNo">077</span> * instead of having HBase handle that automatically. The most useful utilities<a name="line.77"></a>
-<span class="sourceLineNo">078</span> * are:<a name="line.78"></a>
-<span class="sourceLineNo">079</span> * &lt;p&gt;<a name="line.79"></a>
-<span class="sourceLineNo">080</span> * &lt;ul&gt;<a name="line.80"></a>
-<span class="sourceLineNo">081</span> * &lt;li&gt;Create a table with a specified number of pre-split regions<a name="line.81"></a>
-<span class="sourceLineNo">082</span> * &lt;li&gt;Execute a rolling split of all regions on an existing table<a name="line.82"></a>
-<span class="sourceLineNo">083</span> * &lt;/ul&gt;<a name="line.83"></a>
-<span class="sourceLineNo">084</span> * &lt;p&gt;<a name="line.84"></a>
-<span class="sourceLineNo">085</span> * Both operations can be safely done on a live server.<a name="line.85"></a>
-<span class="sourceLineNo">086</span> * &lt;p&gt;<a name="line.86"></a>
-<span class="sourceLineNo">087</span> * &lt;b&gt;Question:&lt;/b&gt; How do I turn off automatic splitting? &lt;br&gt;<a name="line.87"></a>
-<span class="sourceLineNo">088</span> * &lt;b&gt;Answer:&lt;/b&gt; Automatic splitting is determined by the configuration value<a name="line.88"></a>
-<span class="sourceLineNo">089</span> * &lt;i&gt;HConstants.HREGION_MAX_FILESIZE&lt;/i&gt;. It is not recommended that you set this<a name="line.89"></a>
-<span class="sourceLineNo">090</span> * to Long.MAX_VALUE in case you forget about manual splits. A suggested setting<a name="line.90"></a>
-<span class="sourceLineNo">091</span> * is 100GB, which would result in &amp;gt; 1hr major compactions if reached.<a name="line.91"></a>
-<span class="sourceLineNo">092</span> * &lt;p&gt;<a name="line.92"></a>
-<span class="sourceLineNo">093</span> * &lt;b&gt;Question:&lt;/b&gt; Why did the original authors decide to manually split? &lt;br&gt;<a name="line.93"></a>
-<span class="sourceLineNo">094</span> * &lt;b&gt;Answer:&lt;/b&gt; Specific workload characteristics of our use case allowed us<a name="line.94"></a>
-<span class="sourceLineNo">095</span> * to benefit from a manual split system.<a name="line.95"></a>
-<span class="sourceLineNo">096</span> * &lt;p&gt;<a name="line.96"></a>
-<span class="sourceLineNo">097</span> * &lt;ul&gt;<a name="line.97"></a>
-<span class="sourceLineNo">098</span> * &lt;li&gt;Data (~1k) that would grow instead of being replaced<a name="line.98"></a>
-<span class="sourceLineNo">099</span> * &lt;li&gt;Data growth was roughly uniform across all regions<a name="line.99"></a>
-<span class="sourceLineNo">100</span> * &lt;li&gt;OLTP workload. Data loss is a big deal.<a name="line.100"></a>
-<span class="sourceLineNo">101</span> * &lt;/ul&gt;<a name="line.101"></a>
-<span class="sourceLineNo">102</span> * &lt;p&gt;<a name="line.102"></a>
-<span class="sourceLineNo">103</span> * &lt;b&gt;Question:&lt;/b&gt; Why is manual splitting good for this workload? &lt;br&gt;<a name="line.103"></a>
-<span class="sourceLineNo">104</span> * &lt;b&gt;Answer:&lt;/b&gt; Although automated splitting is not a bad option, there are<a name="line.104"></a>
-<span class="sourceLineNo">105</span> * benefits to manual splitting.<a name="line.105"></a>
-<span class="sourceLineNo">106</span> * &lt;p&gt;<a name="line.106"></a>
-<span class="sourceLineNo">107</span> * &lt;ul&gt;<a name="line.107"></a>
-<span class="sourceLineNo">108</span> * &lt;li&gt;With growing amounts of data, splits will continually be needed. Since<a name="line.108"></a>
-<span class="sourceLineNo">109</span> * you always know exactly what regions you have, long-term debugging and<a name="line.109"></a>
-<span class="sourceLineNo">110</span> * profiling is much easier with manual splits. It is hard to trace the logs to<a name="line.110"></a>
-<span class="sourceLineNo">111</span> * understand region level problems if it keeps splitting and getting renamed.<a name="line.111"></a>
-<span class="sourceLineNo">112</span> * &lt;li&gt;Data offlining bugs + unknown number of split regions == oh crap! If an<a name="line.112"></a>
-<span class="sourceLineNo">113</span> * WAL or StoreFile was mistakenly unprocessed by HBase due to a weird bug and<a name="line.113"></a>
-<span class="sourceLineNo">114</span> * you notice it a day or so later, you can be assured that the regions<a name="line.114"></a>
-<span class="sourceLineNo">115</span> * specified in these files are the same as the current regions and you have<a name="line.115"></a>
-<span class="sourceLineNo">116</span> * less headaches trying to restore/replay your data.<a name="line.116"></a>
-<span class="sourceLineNo">117</span> * &lt;li&gt;You can finely tune your compaction algorithm. With roughly uniform data<a name="line.117"></a>
-<span class="sourceLineNo">118</span> * growth, it's easy to cause split / compaction storms as the regions all<a name="line.118"></a>
-<span class="sourceLineNo">119</span> * roughly hit the same data size at the same time. With manual splits, you can<a name="line.119"></a>
-<span class="sourceLineNo">120</span> * let staggered, time-based major compactions spread out your network IO load.<a name="line.120"></a>
-<span class="sourceLineNo">121</span> * &lt;/ul&gt;<a name="line.121"></a>
-<span class="sourceLineNo">122</span> * &lt;p&gt;<a name="line.122"></a>
-<span class="sourceLineNo">123</span> * &lt;b&gt;Question:&lt;/b&gt; What's the optimal number of pre-split regions to create? &lt;br&gt;<a name="line.123"></a>
-<span class="sourceLineNo">124</span> * &lt;b&gt;Answer:&lt;/b&gt; Mileage will vary depending upon your application.<a name="line.124"></a>
-<span class="sourceLineNo">125</span> * &lt;p&gt;<a name="line.125"></a>
-<span class="sourceLineNo">126</span> * The short answer for our application is that we started with 10 pre-split<a name="line.126"></a>
-<span class="sourceLineNo">127</span> * regions / server and watched our data growth over time. It's better to err on<a name="line.127"></a>
-<span class="sourceLineNo">128</span> * the side of too little regions and rolling split later.<a name="line.128"></a>
-<span class="sourceLineNo">129</span> * &lt;p&gt;<a name="line.129"></a>
-<span class="sourceLineNo">130</span> * The more complicated answer is that this depends upon the largest storefile<a name="line.130"></a>
-<span class="sourceLineNo">131</span> * in your region. With a growing data size, this will get larger over time. You<a name="line.131"></a>
-<span class="sourceLineNo">132</span> * want the largest region to be just big enough that the<a name="line.132"></a>
-<span class="sourceLineNo">133</span> * {@link org.apache.hadoop.hbase.regionserver.HStore} compact<a name="line.133"></a>
-<span class="sourceLineNo">134</span> * selection algorithm only compacts it due to a timed major. If you don't, your<a name="line.134"></a>
-<span class="sourceLineNo">135</span> * cluster can be prone to compaction storms as the algorithm decides to run<a name="line.135"></a>
-<span class="sourceLineNo">136</span> * major compactions on a large series of regions all at once. Note that<a name="line.136"></a>
-<span class="sourceLineNo">137</span> * compaction storms are due to the uniform data growth, not the manual split<a name="line.137"></a>
-<span class="sourceLineNo">138</span> * decision.<a name="line.138"></a>
-<span class="sourceLineNo">139</span> * &lt;p&gt;<a name="line.139"></a>
-<span class="sourceLineNo">140</span> * If you pre-split your regions too thin, you can increase the major compaction<a name="line.140"></a>
-<span class="sourceLineNo">141</span> * interval by configuring HConstants.MAJOR_COMPACTION_PERIOD. If your data size<a name="line.141"></a>
-<span class="sourceLineNo">142</span> * grows too large, use this script to perform a network IO safe rolling split<a name="line.142"></a>
-<span class="sourceLineNo">143</span> * of all regions.<a name="line.143"></a>
-<span class="sourceLineNo">144</span> */<a name="line.144"></a>
-<span class="sourceLineNo">145</span>@InterfaceAudience.Private<a name="line.145"></a>
-<span class="sourceLineNo">146</span>public class RegionSplitter {<a name="line.146"></a>
-<span class="sourceLineNo">147</span>  private static final Log LOG = LogFactory.getLog(RegionSplitter.class);<a name="line.147"></a>
-<span class="sourceLineNo">148</span><a name="line.148"></a>
-<span class="sourceLineNo">149</span>  /**<a name="line.149"></a>
-<span class="sourceLineNo">150</span>   * A generic interface for the RegionSplitter code to use for all it's<a name="line.150"></a>
-<span class="sourceLineNo">151</span>   * functionality. Note that the original authors of this code use<a name="line.151"></a>
-<span class="sourceLineNo">152</span>   * {@link HexStringSplit} to partition their table and set it as default, but<a name="line.152"></a>
-<span class="sourceLineNo">153</span>   * provided this for your custom algorithm. To use, create a new derived class<a name="line.153"></a>
-<span class="sourceLineNo">154</span>   * from this interface and call {@link RegionSplitter#createPresplitTable} or<a name="line.154"></a>
-<span class="sourceLineNo">155</span>   * RegionSplitter#rollingSplit(TableName, SplitAlgorithm, Configuration) with the<a name="line.155"></a>
-<span class="sourceLineNo">156</span>   * argument splitClassName giving the name of your class.<a name="line.156"></a>
-<span class="sourceLineNo">157</span>   */<a name="line.157"></a>
-<span class="sourceLineNo">158</span>  public interface SplitAlgorithm {<a name="line.158"></a>
-<span class="sourceLineNo">159</span>    /**<a name="line.159"></a>
-<span class="sourceLineNo">160</span>     * Split a pre-existing region into 2 regions.<a name="line.160"></a>
-<span class="sourceLineNo">161</span>     *<a name="line.161"></a>
-<span class="sourceLineNo">162</span>     * @param start<a name="line.162"></a>
-<span class="sourceLineNo">163</span>     *          first row (inclusive)<a name="line.163"></a>
-<span class="sourceLineNo">164</span>     * @param end<a name="line.164"></a>
-<span class="sourceLineNo">165</span>     *          last row (exclusive)<a name="line.165"></a>
-<span class="sourceLineNo">166</span>     * @return the split row to use<a name="line.166"></a>
-<span class="sourceLineNo">167</span>     */<a name="line.167"></a>
-<span class="sourceLineNo">168</span>    byte[] split(byte[] start, byte[] end);<a name="line.168"></a>
-<span class="sourceLineNo">169</span><a name="line.169"></a>
-<span class="sourceLineNo">170</span>    /**<a name="line.170"></a>
-<span class="sourceLineNo">171</span>     * Split an entire table.<a name="line.171"></a>
-<span class="sourceLineNo">172</span>     *<a name="line.172"></a>
-<span class="sourceLineNo">173</span>     * @param numRegions<a name="line.173"></a>
-<span class="sourceLineNo">174</span>     *          number of regions to split the table into<a name="line.174"></a>
-<span class="sourceLineNo">175</span>     *<a name="line.175"></a>
-<span class="sourceLineNo">176</span>     * @throws RuntimeException<a name="line.176"></a>
-<span class="sourceLineNo">177</span>     *           user input is validated at this time. may throw a runtime<a name="line.177"></a>
-<span class="sourceLineNo">178</span>     *           exception in response to a parse failure<a name="line.178"></a>
-<span class="sourceLineNo">179</span>     * @return array of split keys for the initial regions of the table. The<a name="line.179"></a>
-<span class="sourceLineNo">180</span>     *         length of the returned array should be numRegions-1.<a name="line.180"></a>
-<span class="sourceLineNo">181</span>     */<a name="line.181"></a>
-<span class="sourceLineNo">182</span>    byte[][] split(int numRegions);<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>     * In HBase, the first row is represented by an empty byte array. This might<a name="line.185"></a>
-<span class="sourceLineNo">186</span>     * cause problems with your split algorithm or row printing. All your APIs<a name="line.186"></a>
-<span class="sourceLineNo">187</span>     * will be passed firstRow() instead of empty array.<a name="line.187"></a>
+<span class="sourceLineNo">052</span>import org.apache.hadoop.hbase.HConstants;<a name="line.52"></a>
+<span class="sourceLineNo">053</span>import org.apache.hadoop.hbase.HRegionInfo;<a name="line.53"></a>
+<span class="sourceLineNo">054</span>import org.apache.hadoop.hbase.HRegionLocation;<a name="line.54"></a>
+<span class="sourceLineNo">055</span>import org.apache.hadoop.hbase.MetaTableAccessor;<a name="line.55"></a>
+<span class="sourceLineNo">056</span>import org.apache.hadoop.hbase.ServerName;<a name="line.56"></a>
+<span class="sourceLineNo">057</span>import org.apache.hadoop.hbase.TableName;<a name="line.57"></a>
+<span class="sourceLineNo">058</span>import org.apache.yetus.audience.InterfaceAudience;<a name="line.58"></a>
+<span class="sourceLineNo">059</span>import org.apache.hadoop.hbase.client.Admin;<a name="line.59"></a>
+<span class="sourceLineNo">060</span>import org.apache.hadoop.hbase.client.ClusterConnection;<a name="line.60"></a>
+<span class="sourceLineNo">061</span>import org.apache.hadoop.hbase.client.Connection;<a name="line.61"></a>
+<span class="sourceLineNo">062</span>import org.apache.hadoop.hbase.client.ConnectionFactory;<a name="line.62"></a>
+<span class="sourceLineNo">063</span>import org.apache.hadoop.hbase.client.NoServerForRegionException;<a name="line.63"></a>
+<span class="sourceLineNo">064</span>import org.apache.hadoop.hbase.client.RegionLocator;<a name="line.64"></a>
+<span class="sourceLineNo">065</span>import org.apache.hadoop.hbase.client.Table;<a name="line.65"></a>
+<span class="sourceLineNo">066</span>import org.apache.hadoop.hbase.regionserver.HRegionFileSystem;<a name="line.66"></a>
+<span class="sourceLineNo">067</span><a name="line.67"></a>
+<span class="sourceLineNo">068</span>import org.apache.hadoop.hbase.client.ColumnFamilyDescriptorBuilder;<a name="line.68"></a>
+<span class="sourceLineNo">069</span>import org.apache.hadoop.hbase.client.TableDescriptorBuilder;<a name="line.69"></a>
+<span class="sourceLineNo">070</span>import org.apache.hadoop.hbase.shaded.com.google.common.base.Preconditions;<a name="line.70"></a>
+<span class="sourceLineNo">071</span>import org.apache.hadoop.hbase.shaded.com.google.common.collect.Lists;<a name="line.71"></a>
+<span class="sourceLineNo">072</span>import org.apache.hadoop.hbase.shaded.com.google.common.collect.Maps;<a name="line.72"></a>
+<span class="sourceLineNo">073</span>import org.apache.hadoop.hbase.shaded.com.google.common.collect.Sets;<a name="line.73"></a>
+<span class="sourceLineNo">074</span><a name="line.74"></a>
+<span class="sourceLineNo">075</span>/**<a name="line.75"></a>
+<span class="sourceLineNo">076</span> * The {@link RegionSplitter} class provides several utilities to help in the<a name="line.76"></a>
+<span class="sourceLineNo">077</span> * administration lifecycle for developers who choose to manually split regions<a name="line.77"></a>
+<span class="sourceLineNo">078</span> * instead of having HBase handle that automatically. The most useful utilities<a name="line.78"></a>
+<span class="sourceLineNo">079</span> * are:<a name="line.79"></a>
+<span class="sourceLineNo">080</span> * &lt;p&gt;<a name="line.80"></a>
+<span class="sourceLineNo">081</span> * &lt;ul&gt;<a name="line.81"></a>
+<span class="sourceLineNo">082</span> * &lt;li&gt;Create a table with a specified number of pre-split regions<a name="line.82"></a>
+<span class="sourceLineNo">083</span> * &lt;li&gt;Execute a rolling split of all regions on an existing table<a name="line.83"></a>
+<span class="sourceLineNo">084</span> * &lt;/ul&gt;<a name="line.84"></a>
+<span class="sourceLineNo">085</span> * &lt;p&gt;<a name="line.85"></a>
+<span class="sourceLineNo">086</span> * Both operations can be safely done on a live server.<a name="line.86"></a>
+<span class="sourceLineNo">087</span> * &lt;p&gt;<a name="line.87"></a>
+<span class="sourceLineNo">088</span> * &lt;b&gt;Question:&lt;/b&gt; How do I turn off automatic splitting? &lt;br&gt;<a name="line.88"></a>
+<span class="sourceLineNo">089</span> * &lt;b&gt;Answer:&lt;/b&gt; Automatic splitting is determined by the configuration value<a name="line.89"></a>
+<span class="sourceLineNo">090</span> * &lt;i&gt;HConstants.HREGION_MAX_FILESIZE&lt;/i&gt;. It is not recommended that you set this<a name="line.90"></a>
+<span class="sourceLineNo">091</span> * to Long.MAX_VALUE in case you forget about manual splits. A suggested setting<a name="line.91"></a>
+<span class="sourceLineNo">092</span> * is 100GB, which would result in &amp;gt; 1hr major compactions if reached.<a name="line.92"></a>
+<span class="sourceLineNo">093</span> * &lt;p&gt;<a name="line.93"></a>
+<span class="sourceLineNo">094</span> * &lt;b&gt;Question:&lt;/b&gt; Why did the original authors decide to manually split? &lt;br&gt;<a name="line.94"></a>
+<span class="sourceLineNo">095</span> * &lt;b&gt;Answer:&lt;/b&gt; Specific workload characteristics of our use case allowed us<a name="line.95"></a>
+<span class="sourceLineNo">096</span> * to benefit from a manual split system.<a name="line.96"></a>
+<span class="sourceLineNo">097</span> * &lt;p&gt;<a name="line.97"></a>
+<span class="sourceLineNo">098</span> * &lt;ul&gt;<a name="line.98"></a>
+<span class="sourceLineNo">099</span> * &lt;li&gt;Data (~1k) that would grow instead of being replaced<a name="line.99"></a>
+<span class="sourceLineNo">100</span> * &lt;li&gt;Data growth was roughly uniform across all regions<a name="line.100"></a>
+<span class="sourceLineNo">101</span> * &lt;li&gt;OLTP workload. Data loss is a big deal.<a name="line.101"></a>
+<span class="sourceLineNo">102</span> * &lt;/ul&gt;<a name="line.102"></a>
+<span class="sourceLineNo">103</span> * &lt;p&gt;<a name="line.103"></a>
+<span class="sourceLineNo">104</span> * &lt;b&gt;Question:&lt;/b&gt; Why is manual splitting good for this workload? &lt;br&gt;<a name="line.104"></a>
+<span class="sourceLineNo">105</span> * &lt;b&gt;Answer:&lt;/b&gt; Although automated splitting is not a bad option, there are<a name="line.105"></a>
+<span class="sourceLineNo">106</span> * benefits to manual splitting.<a name="line.106"></a>
+<span class="sourceLineNo">107</span> * &lt;p&gt;<a name="line.107"></a>
+<span class="sourceLineNo">108</span> * &lt;ul&gt;<a name="line.108"></a>
+<span class="sourceLineNo">109</span> * &lt;li&gt;With growing amounts of data, splits will continually be needed. Since<a name="line.109"></a>
+<span class="sourceLineNo">110</span> * you always know exactly what regions you have, long-term debugging and<a name="line.110"></a>
+<span class="sourceLineNo">111</span> * profiling is much easier with manual splits. It is hard to trace the logs to<a name="line.111"></a>
+<span class="sourceLineNo">112</span> * understand region level problems if it keeps splitting and getting renamed.<a name="line.112"></a>
+<span class="sourceLineNo">113</span> * &lt;li&gt;Data offlining bugs + unknown number of split regions == oh crap! If an<a name="line.113"></a>
+<span class="sourceLineNo">114</span> * WAL or StoreFile was mistakenly unprocessed by HBase due to a weird bug and<a name="line.114"></a>
+<span class="sourceLineNo">115</span> * you notice it a day or so later, you can be assured that the regions<a name="line.115"></a>
+<span class="sourceLineNo">116</span> * specified in these files are the same as the current regions and you have<a name="line.116"></a>
+<span class="sourceLineNo">117</span> * less headaches trying to restore/replay your data.<a name="line.117"></a>
+<span class="sourceLineNo">118</span> * &lt;li&gt;You can finely tune your compaction algorithm. With roughly uniform data<a name="line.118"></a>
+<span class="sourceLineNo">119</span> * growth, it's easy to cause split / compaction storms as the regions all<a name="line.119"></a>
+<span class="sourceLineNo">120</span> * roughly hit the same data size at the same time. With manual splits, you can<a name="line.120"></a>
+<span class="sourceLineNo">121</span> * let staggered, time-based major compactions spread out your network IO load.<a name="line.121"></a>
+<span class="sourceLineNo">122</span> * &lt;/ul&gt;<a name="line.122"></a>
+<span class="sourceLineNo">123</span> * &lt;p&gt;<a name="line.123"></a>
+<span class="sourceLineNo">124</span> * &lt;b&gt;Question:&lt;/b&gt; What's the optimal number of pre-split regions to create? &lt;br&gt;<a name="line.124"></a>
+<span class="sourceLineNo">125</span> * &lt;b&gt;Answer:&lt;/b&gt; Mileage will vary depending upon your application.<a name="line.125"></a>
+<span class="sourceLineNo">126</span> * &lt;p&gt;<a name="line.126"></a>
+<span class="sourceLineNo">127</span> * The short answer for our application is that we started with 10 pre-split<a name="line.127"></a>
+<span class="sourceLineNo">128</span> * regions / server and watched our data growth over time. It's better to err on<a name="line.128"></a>
+<span class="sourceLineNo">129</span> * the side of too little regions and rolling split later.<a name="line.129"></a>
+<span class="sourceLineNo">130</span> * &lt;p&gt;<a name="line.130"></a>
+<span class="sourceLineNo">131</span> * The more complicated answer is that this depends upon the largest storefile<a name="line.131"></a>
+<span class="sourceLineNo">132</span> * in your region. With a growing data size, this will get larger over time. You<a name="line.132"></a>
+<span class="sourceLineNo">133</span> * want the largest region to be just big enough that the<a name="line.133"></a>
+<span class="sourceLineNo">134</span> * {@link org.apache.hadoop.hbase.regionserver.HStore} compact<a name="line.134"></a>
+<span class="sourceLineNo">135</span> * selection algorithm only compacts it due to a timed major. If you don't, your<a name="line.135"></a>
+<span class="sourceLineNo">136</span> * cluster can be prone to compaction storms as the algorithm decides to run<a name="line.136"></a>
+<span class="sourceLineNo">137</span> * major compactions on a large series of regions all at once. Note that<a name="line.137"></a>
+<span class="sourceLineNo">138</span> * compaction storms are due to the uniform data growth, not the manual split<a name="line.138"></a>
+<span class="sourceLineNo">139</span> * decision.<a name="line.139"></a>
+<span class="sourceLineNo">140</span> * &lt;p&gt;<a name="line.140"></a>
+<span class="sourceLineNo">141</span> * If you pre-split your regions too thin, you can increase the major compaction<a name="line.141"></a>
+<span class="sourceLineNo">142</span> * interval by configuring HConstants.MAJOR_COMPACTION_PERIOD. If your data size<a name="line.142"></a>
+<span class="sourceLineNo">143</span> * grows too large, use this script to perform a network IO safe rolling split<a name="line.143"></a>
+<span class="sourceLineNo">144</span> * of all regions.<a name="line.144"></a>
+<span class="sourceLineNo">145</span> */<a name="line.145"></a>
+<span class="sourceLineNo">146</span>@InterfaceAudience.Private<a name="line.146"></a>
+<span class="sourceLineNo">147</span>public class RegionSplitter {<a name="line.147"></a>
+<span class="sourceLineNo">148</span>  private static final Log LOG = LogFactory.getLog(RegionSplitter.class);<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>   * A generic interface for the RegionSplitter code to use for all it's<a name="line.151"></a>
+<span class="sourceLineNo">152</span>   * functionality. Note that the original authors of this code use<a name="line.152"></a>
+<span class="sourceLineNo">153</span>   * {@link HexStringSplit} to partition their table and set it as default, but<a name="line.153"></a>
+<span class="sourceLineNo">154</span>   * provided this for your custom algorithm. To use, create a new derived class<a name="line.154"></a>
+<span class="sourceLineNo">155</span>   * from this interface and call {@link RegionSplitter#createPresplitTable} or<a name="line.155"></a>
+<span class="sourceLineNo">156</span>   * RegionSplitter#rollingSplit(TableName, SplitAlgorithm, Configuration) with the<a name="line.156"></a>
+<span class="sourceLineNo">157</span>   * argument splitClassName giving the name of your class.<a name="line.157"></a>
+<span class="sourceLineNo">158</span>   */<a name="line.158"></a>
+<span class="sourceLineNo">159</span>  public interface SplitAlgorithm {<a name="line.159"></a>
+<span class="sourceLineNo">160</span>    /**<a name="line.160"></a>
+<span class="sourceLineNo">161</span>     * Split a pre-existing region into 2 regions.<a name="line.161"></a>
+<span class="sourceLineNo">162</span>     *<a name="line.162"></a>
+<span class="sourceLineNo">163</span>     * @param start<a name="line.163"></a>
+<span class="sourceLineNo">164</span>     *          first row (inclusive)<a name="line.164"></a>
+<span class="sourceLineNo">165</span>     * @param end<a name="line.165"></a>
+<span class="sourceLineNo">166</span>     *          last row (exclusive)<a name="line.166"></a>
+<span class="sourceLineNo">167</span>     * @return the split row to use<a name="line.167"></a>
+<span class="sourceLineNo">168</span>     */<a name="line.168"></a>
+<span class="sourceLineNo">169</span>    byte[] split(byte[] start, byte[] end);<a name="line.169"></a>
+<span class="sourceLineNo">170</span><a name="line.170"></a>
+<span class="sourceLineNo">171</span>    /**<a name="line.171"></a>
+<span class="sourceLineNo">172</span>     * Split an entire table.<a name="line.172"></a>
+<span class="sourceLineNo">173</span>     *<a name="line.173"></a>
+<span class="sourceLineNo">174</span>     * @param numRegions<a name="line.174"></a>
+<span class="sourceLineNo">175</span>     *          number of regions to split the table into<a name="line.175"></a>
+<span class="sourceLineNo">176</span>     *<a name="line.176"></a>
+<span class="sourceLineNo">177</span>     * @throws RuntimeException<a name="line.177"></a>
+<span class="sourceLineNo">178</span>     *           user input is validated at this time. may throw a runtime<a name="line.178"></a>
+<span class="sourceLineNo">179</span>     *           exception in response to a parse failure<a name="line.179"></a>
+<span class="sourceLineNo">180</span>     * @return array of split keys for the initial regions of the table. The<a name="line.180"></a>
+<span class="sourceLineNo">181</span>     *         length of the returned array should be numRegions-1.<a name="line.181"></a>
+<span class="sourceLineNo">182</span>     */<a name="line.182"></a>
+<span class="sourceLineNo">183</span>    byte[][] split(int numRegions);<a name="line.183"></a>
+<span class="sourceLineNo">184</span><a name="line.184"></a>
+<span class="sourceLineNo">185</span>    /**<a name="line.185"></a>
+<span class="sourceLineNo">186</span>     * Some MapReduce jobs may want to run multiple mappers per region,<a name="line.186"></a>
+<span class="sourceLineNo">187</span>     * this is intended for such usecase.<a name="line.187"></a>
 <span class="sourceLineNo">188</span>     *<a name="line.188"></a>
-<span class="sourceLineNo">189</span>     * @return your representation of your first row<a name="line.189"></a>
-<span class="sourceLineNo">190</span>     */<a name="line.190"></a>
-<span class="sourceLineNo">191</span>    byte[] firstRow();<a name="line.191"></a>
-<span class="sourceLineNo">192</span><a name="line.192"></a>
-<span class="sourceLineNo">193</span>    /**<a name="line.193"></a>
-<span class="sourceLineNo">194</span>     * In HBase, the last row is represented by an empty byte array. This might<a name="line.194"></a>
-<span class="sourceLineNo">195</span>     * cause problems with your split algorithm or row printing. All your APIs<a name="line.195"></a>
-<span class="sourceLineNo">196</span>     * will be passed firstRow() instead of empty array.<a name="line.196"></a>
-<span class="sourceLineNo">197</span>     *<a name="line.197"></a>
-<span class="sourceLineNo">198</span>     * @return your representation of your last row<a name="line.198"></a>
-<span class="sourceLineNo">199</span>     */<a name="line.199"></a>
-<span class="sourceLineNo">200</span>    byte[] lastRow();<a name="line.200"></a>
-<span class="sourceLineNo">201</span><a name="line.201"></a>
-<span class="sourceLineNo">202</span>    /**<a name="line.202"></a>
-<span class="sourceLineNo">203</span>     * In HBase, the last row is represented by an empty byte array. Set this<a name="line.203"></a>
-<span class="sourceLineNo">204</span>     * value to help the split code understand how to evenly divide the first<a name="line.204"></a>
-<span class="sourceLineNo">205</span>     * region.<a name="line.205"></a>
-<span class="sourceLineNo">206</span>     *<a name="line.206"></a>
-<span class="sourceLineNo">207</span>     * @param userInput<a name="line.207"></a>
-<span class="sourceLineNo">208</span>     *          raw user input (may throw RuntimeException on parse failure)<a name="line.208"></a>
-<span class="sourceLineNo">209</span>     */<a name="line.209"></a>
-<span class="sourceLineNo">210</span>    void setFirstRow(String userInput);<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>     * In HBase, the last row is represented by an empty byte array. Set this<a name="line.213"></a>
-<span class="sourceLineNo">214</span>     * value to help the split code understand how to evenly divide the last<a name="line.214"></a>
-<span class="sourceLineNo">215</span>     * region. Note that this last row is inclusive for all rows sharing the<a name="line.215"></a>
-<span class="sourceLineNo">216</span>     * same prefix.<a name="line.216"></a>
-<span class="sourceLineNo">217</span>     *<a name="line.217"></a>
-<span class="sourceLineNo">218</span>     * @param userInput<a name="line.218"></a>
-<span class="sourceLineNo">219</span>     *          raw user input (may throw RuntimeException on parse failure)<a name="line.219"></a>
-<span class="sourceLineNo">220</span>     */<a name="line.220"></a>
-<span class="sourceLineNo">221</span>    void setLastRow(String userInput);<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>     * @param input<a name="line.224"></a>
-<span class="sourceLineNo">225</span>     *          user or file input for row<a name="line.225"></a>
-<span class="sourceLineNo">226</span>     * @return byte array representation of this row for HBase<a name="line.226"></a>
-<span class="sourceLineNo">227</span>     */<a name="line.227"></a>
-<span class="sourceLineNo">228</span>    byte[] strToRow(String input);<a name="line.228"></a>
-<span class="sourceLineNo">229</span><a name="line.229"></a>
-<span class="sourceLineNo">230</span>    /**<a name="line.230"></a>
-<span class="sourceLineNo">231</span>     * @param row<a name="line.231"></a>
-<span class="sourceLineNo">232</span>     *          byte array representing a row in HBase<a name="line.232"></a>
-<span class="sourceLineNo">233</span>     * @return String to use for debug &amp;amp; file printing<a name="line.233"></a>
-<span class="sourceLineNo">234</span>     */<a name="line.234"></a>
-<span class="sourceLineNo">235</span>    String rowToStr(byte[] row);<a name="line.235"></a>
-<span class="sourceLineNo">236</span><a name="line.236"></a>
-<span class="sourceLineNo">237</span>    /**<a name="line.237"></a>
-<span class="sourceLineNo">238</span>     * @return the separator character to use when storing / printing the row<a name="line.238"></a>
+<span class="sourceLineNo">189</span>     * @param start first row (inclusive)<a name="line.189"></a>
+<span class="sourceLineNo">190</span>     * @param end last row (exclusive)<a name="line.190"></a>
+<span class="sourceLineNo">191</span>     * @param numSplits number of splits to generate<a name="line.191"></a>
+<span class="sourceLineNo">192</span>     * @param inclusive whether start and end are returned as split points<a name="line.192"></a>
+<span class="sourceLineNo">193</span>     */<a name="line.193"></a>
+<span class="sourceLineNo">194</span>    byte[][] split(byte[] start, byte[] end, int numSplits, boolean inclusive);<a name="line.194"></a>
+<span class="sourceLineNo">195</span><a name="line.195"></a>
+<span class="sourceLineNo">196</span>    /**<a name="line.196"></a>
+<span class="sourceLineNo">197</span>     * In HBase, the first row is represented by an empty byte array. This might<a name="line.197"></a>
+<span class="sourceLineNo">198</span>     * cause problems with your split algorithm or row printing. All your APIs<a name="line.198"></a>
+<span class="sourceLineNo">199</span>     * will be passed firstRow() instead of empty array.<a name="line.199"></a>
+<span class="sourceLineNo">200</span>     *<a name="line.200"></a>
+<span class="sourceLineNo">201</span>     * @return your representation of your first row<a name="line.201"></a>
+<span class="sourceLineNo">202</span>     */<a name="line.202"></a>
+<span class="sourceLineNo">203</span>    byte[] firstRow();<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>     * In HBase, the last row is represented by an empty byte array. This might<a name="line.206"></a>
+<span class="sourceLineNo">207</span>     * cause problems with your split algorithm or row printing. All your APIs<a name="line.207"></a>
+<span class="sourceLineNo">208</span>     * will be passed firstRow() instead of empty array.<a name="line.208"></a>
+<span class="sourceLineNo">209</span>     *<a name="line.209"></a>
+<span class="sourceLineNo">210</span>     * @return your representation of your last row<a name="line.210"></a>
+<span class="sourceLineNo">211</span>     */<a name="line.211"></a>
+<span class="sourceLineNo">212</span>    byte[] lastRow();<a name="line.212"></a>
+<span class="sourceLineNo">213</span><a name="line.213"></a>
+<span class="sourceLineNo">214</span>    /**<a name="line.214"></a>
+<span class="sourceLineNo">215</span>     * In HBase, the last row is represented by an empty byte array. Set this<a name="line.215"></a>
+<span class="sourceLineNo">216</span>     * value to help the split code understand how to evenly divide the first<a name="line.216"></a>
+<span class="sourceLineNo">217</span>     * region.<a name="line.217"></a>
+<span class="sourceLineNo">218</span>     *<a name="line.218"></a>
+<span class="sourceLineNo">219</span>     * @param userInput<a name="line.219"></a>
+<span class="sourceLineNo">220</span>     *          raw user input (may throw RuntimeException on parse failure)<a name="line.220"></a>
+<span class="sourceLineNo">221</span>     */<a name="line.221"></a>
+<span class="sourceLineNo">222</span>    void setFirstRow(String userInput);<a name="line.222"></a>
+<span class="sourceLineNo">223</span><a name="line.223"></a>
+<span class="sourceLineNo">224</span>    /**<a name="line.224"></a>
+<span class="sourceLineNo">225</span>     * In HBase, the last row is represented by an empty byte array. Set this<a name="line.225"></a>
+<span class="sourceLineNo">226</span>     * value to help the split code understand how to evenly divide the last<a name="line.226"></a>
+<span class="sourceLineNo">227</span>     * region. Note that this last row is inclusive for all rows sharing the<a name="line.227"></a>
+<span class="sourceLineNo">228</span>     * same prefix.<a name="line.228"></a>
+<span class="sourceLineNo">229</span>     *<a name="line.229"></a>
+<span class="sourceLineNo">230</span>     * @param userInput<a name="line.230"></a>
+<span class="sourceLineNo">231</span>     *          raw user input (may throw RuntimeException on parse failure)<a name="line.231"></a>
+<span class="sourceLineNo">232</span>     */<a name="line.232"></a>
+<span class="sourceLineNo">233</span>    void setLastRow(String userInput);<a name="line.233"></a>
+<span class="sourceLineNo">234</span><a name="line.234"></a>
+<span class="sourceLineNo">235</span>    /**<a name="line.235"></a>
+<span class="sourceLineNo">236</span>     * @param input<a name="line.236"></a>
+<span class="sourceLineNo">237</span>     *          user or file input for row<a name="line.237"></a>
+<span class="sourceLineNo">238</span>     * @return byte array representation of this row for HBase<a name="line.238"></a>
 <span class="sourceLineNo">239</span>     */<a name="line.239"></a>
-<span class="sourceLineNo">240</span>    String separator();<a name="line.240"></a>
+<span class="sourceLineNo">240</span>    byte[] strToRow(String input);<a name="line.240"></a>
 <span class="sourceLineNo">241</span><a name="line.241"></a>
 <span class="sourceLineNo">242</span>    /**<a name="line.242"></a>
-<span class="sourceLineNo">243</span>     * Set the first row<a name="line.243"></a>
-<span class="sourceLineNo">244</span>     * @param userInput byte array of the row key.<a name="line.244"></a>
-<span class="sourceLineNo">245</span>     */<a name="line.245"></a>
-<span class="sourceLineNo">246</span>    void setFirstRow(byte[] userInput);<a name="line.246"></a>
-<span class="sourceLineNo">247</span><a name="line.247"></a>
-<span class="sourceLineNo">248</span>    /**<a name="line.248"></a>
-<span class="sourceLineNo">249</span>     * Set the last row<a name="line.249"></a>
-<span class="sourceLineNo">250</span>     * @param userInput byte array of the row key.<a name="line.250"></a>
+<span class="sourceLineNo">243</span>     * @param row<a name="line.243"></a>
+<span class="sourceLineNo">244</span>     *          byte array representing a row in HBase<a name="line.244"></a>
+<span class="sourceLineNo">245</span>     * @return String to use for debug &amp;amp; file printing<a name="line.245"></a>
+<span class="sourceLineNo">246</span>     */<a name="line.246"></a>
+<span class="sourceLineNo">247</span>    String rowToStr(byte[] row);<a name="line.247"></a>
+<span class="sourceLineNo">248</span><a name="line.248"></a>
+<span class="sourceLineNo">249</span>    /**<a name="line.249"></a>
+<span class="sourceLineNo">250</span>     * @return the separator character to use when storing / printing the row<a name="line.250"></a>
 <span class="sourceLineNo">251</span>     */<a name="line.251"></a>
-<span class="sourceLineNo">252</span>    void setLastRow(byte[] userInput);<a name="line.252"></a>
-<span class="sourceLineNo">253</span>  }<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>   * The main function for the RegionSplitter application. Common uses:<a name="line.256"></a>
-<span class="sourceLineNo">257</span>   * &lt;p&gt;<a name="line.257"></a>
-<span class="sourceLineNo">258</span>   * &lt;ul&gt;<a name="line.258"></a>
-<span class="sourceLineNo">259</span>   * &lt;li&gt;create a table named 'myTable' with 60 pre-split regions containing 2<a name="line.259"></a>
-<span class="sourceLineNo">260</span>   * column families 'test' &amp;amp; 'rs', assuming the keys are hex-encoded ASCII:<a name="line.260"></a>
-<span class="sourceLineNo">261</span>   * &lt;ul&gt;<a name="line.261"></a>
-<span class="sourceLineNo">262</span>   * &lt;li&gt;bin/hbase org.apache.hadoop.hbase.util.RegionSplitter -c 60 -f test:rs<a name="line.262"></a>
-<span class="sourceLineNo">263</span>   * myTable HexStringSplit<a name="line.263"></a>
-<span class="sourceLineNo">264</span>   * &lt;/ul&gt;<a name="line.264"></a>
-<span class="sourceLineNo">265</span>   * &lt;li&gt;perform a rolling split of 'myTable' (i.e. 60 =&amp;gt; 120 regions), # 2<a name="line.265"></a>
-<span class="sourceLineNo">266</span>   * outstanding splits at a time, assuming keys are uniformly distributed<a name="line.266"></a>
-<span class="sourceLineNo">267</span>   * bytes:<a name="line.267"></a>
-<span class="sourceLineNo">268</span>   * &lt;ul&gt;<a name="line.268"></a>
-<span class="sourceLineNo">269</span>   * &lt;li&gt;bin/hbase org.apache.hadoop.hbase.util.RegionSplitter -r -o 2 myTable<a name="line.269"></a>
-<span class="sourceLineNo">270</span>   * UniformSplit<a name="line.270"></a>
-<span class="sourceLineNo">271</span>   * &lt;/ul&gt;<a name="line.271"></a>
-<span class="sourceLineNo">272</span>   * &lt;/ul&gt;<a name="line.272"></a>
-<span class="sourceLineNo">273</span>   *<a name="line.273"></a>
-<span class="sourceLineNo">274</span>   * There are two SplitAlgorithms built into RegionSplitter, HexStringSplit<a name="line.274"></a>
-<span class="sourceLineNo">275</span>   * and UniformSplit. These are different strategies for choosing region<a name="line.275"></a>
-<span class="sourceLineNo">276</span>   * boundaries. See their source code for details.<a name="line.276"></a>
-<span class="sourceLineNo">277</span>   *<a name="line.277"></a>
-<span class="sourceLineNo">278</span>   * @param args<a name="line.278"></a>
-<span class="sourceLineNo">279</span>   *          Usage: RegionSplitter &amp;lt;TABLE&amp;gt; &amp;lt;SPLITALGORITHM&amp;gt;<a name="line.279"></a>
-<span class="sourceLineNo">280</span>   *          &amp;lt;-c &amp;lt;# regions&amp;gt; -f &amp;lt;family:family:...&amp;gt; | -r<a name="line.280"></a>
-<span class="sourceLineNo">281</span>   *          [-o &amp;lt;# outstanding splits&amp;gt;]&amp;gt;<a name="line.281"></a>
-<span class="sourceLineNo">282</span>   *          [-D &amp;lt;conf.param=value&amp;gt;]<a name="line.282"></a>
-<span class="sourceLineNo">283</span>   * @throws IOException<a name="line.283"></a>
-<span class="sourceLineNo">284</span>   *           HBase IO problem<a name="line.284"></a>
-<span class="sourceLineNo">285</span>   * @throws InterruptedException<a name="line.285"></a>
-<span class="sourceLineNo">286</span>   *           user requested exit<a name="line.286"></a>
-<span class="sourceLineNo">287</span>   * @throws ParseException<a name="line.287"></a>
-<span class="sourceLineNo">288</span>   *           problem parsing user input<a name="line.288"></a>
-<span class="sourceLineNo">289</span>   */<a name="line.289"></a>
-<span class="sourceLineNo">290</span>  @SuppressWarnings("static-access")<a name="line.290"></a>
-<span class="sourceLineNo">291</span>  public static void main(String[] args) throws IOException,<a name="line.291"></a>
-<span class="sourceLineNo">292</span>      InterruptedException, ParseException {<a name="line.292"></a>
-<span class="sourceLineNo">293</span>    Configuration conf = HBaseConfiguration.create();<a name="line.293"></a>
-<span class="sourceLineNo">294</span><a name="line.294"></a>
-<span class="sourceLineNo">295</span>    // parse user input<a name="line.295"></a>
-<span class="sourceLineNo">296</span>    Options opt = new Options();<a name="line.296"></a>
-<span class="sourceLineNo">297</span>    opt.addOption(OptionBuilder.withArgName("property=value").hasArg()<a name="line.297"></a>
-<span class="sourceLineNo">298</span>        .withDescription("Override HBase Configuration Settings").create("D"));<a name="line.298"></a>
-<span class="sourceLineNo">299</span>    opt.addOption(OptionBuilder.withArgName("region count").hasArg()<a name="line.299"></a>
-<span class="sourceLineNo">300</span>        .withDescription(<a name="line.300"></a>
-<span class="sourceLineNo">301</span>            "Create a new table with a pre-split number of regions")<a name="line.301"></a>
-<span class="sourceLineNo">302</span>        .create("c"));<a name="line.302"></a>
-<span class="sourceLineNo">303</span>    opt.addOption(OptionBuilder.withArgName("family:family:...").hasArg()<a name="line.303"></a>
-<span class="sourceLineNo">304</span>        .withDescription(<a name="line.304"></a>
-<span class="sourceLineNo">305</span>            "Column Families to create with new table.  Required with -c")<a name="line.305"></a>
-<span class="sourceLineNo">306</span>        .create("f"));<a name="line.306"></a>
-<span class="sourceLineNo">307</span>    opt.addOption("h", false, "Print this usage help");<a name="line.307"></a>
-<span class="sourceLineNo">308</span>    opt.addOption("r", false, "Perform a rolling split of an existing region");<a name="line.308"></a>
-<span class="sourceLineNo">309</span>    opt.addOption(OptionBuilder.withArgName("count").hasArg().withDescription(<a name="line.309"></a>
-<span class="sourceLineNo">310</span>        "Max outstanding splits that have unfinished major compactions")<a name="line.310"></a>
-<span class="sourceLineNo">311</span>        .create("o"));<a name="line.311"></a>
-<span class="sourceLineNo">312</span>    opt.addOption(null, "firstrow", true,<a name="line.312"></a>
-<span class="sourceLineNo">313</span>        "First Row in Table for Split Algorithm");<a name="line.313"></a>
-<span class="sourceLineNo">314</span>    opt.addOption(null, "lastrow", true,<a name="line.314"></a>
-<span class="sourceLineNo">315</span>        "Last Row in Table for Split Algorithm");<a name="line.315"></a>
-<span class="sourceLineNo">316</span>    opt.addOption(null, "risky", false,<a name="line.316"></a>
-<span class="sourceLineNo">317</span>        "Skip verification steps to complete quickly."<a name="line.317"></a>
-<span class="sourceLineNo">318</span>            + "STRONGLY DISCOURAGED for production systems.  ");<a name="line.318"></a>
-<span class="sourceLineNo">319</span>    CommandLine cmd = new GnuParser().parse(opt, args);<a name="line.319"></a>
-<span class="sourceLineNo">320</span><a name="line.320"></a>
-<span class="sourceLineNo">321</span>    if (cmd.hasOption("D")) {<a name="line.321"></a>
-<span class="sourceLineNo">322</span>      for (String confOpt : cmd.getOptionValues("D")) {<a name="line.322"></a>
-<span class="sourceLineNo">323</span>        String[] kv = confOpt.split("=", 2);<a name="line.323"></a>
-<span class="sourceLineNo">324</span>        if (kv.length == 2) {<a name="line.324"></a>
-<span class="sourceLineNo">325</span>          conf.set(kv[0], kv[1]);<a name="line.325"></a>
-<span class="sourceLineNo">326</span>          LOG.debug("-D configuration override: " + kv[0] + "=" + kv[1]);<a name="line.326"></a>
-<span class="sourceLineNo">327</span>        } else {<a name="line.327"></a>
-<span class="sourceLineNo">328</span>          throw new ParseException("-D option format invalid: " + confOpt);<a name="line.328"></a>
-<span class="sourceLineNo">329</span>        }<a name="line.329"></a>
-<span class="sourceLineNo">330</span>      }<a name="line.330"></a>
-<span class="sourceLineNo">331</span>    }<a name="line.331"></a>
+<span class="sourceLineNo">252</span>    String separator();<a name="line.252"></a>
+<span class="sourceLineNo">253</span><a name="line.253"></a>
+<span class="sourceLineNo">254</span>    /**<a name="line.254"></a>
+<span class="sourceLineNo">255</span>     * Set the first row<a name="line.255"></a>
+<span class="sourceLineNo">256</span>     * @param userInput byte array of the row key.<a name="line.256"></a>
+<span class="sourceLineNo">257</span>     */<a name="line.257"></a>
+<span class="sourceLineNo">258</span>    void setFirstRow(byte[] userInput);<a name="line.258"></a>
+<span class="sourceLineNo">259</span><a name="line.259"></a>
+<span class="sourceLineNo">260</span>    /**<a name="line.260"></a>
+<span class="sourceLineNo">261</span>     * Set the last row<a name="line.261"></a>
+<span class="sourceLineNo">262</span>     * @param userInput byte array of the row key.<a name="line.262"></a>
+<span class="sourceLineNo">263</span>     */<a name="line.263"></a>
+<span class="sourceLineNo">264</span>    void setLastRow(byte[] userInput);<a name="line.264"></a>
+<span class="sourceLineNo">265</span>  }<a name="line.265"></a>
+<span class="sourceLineNo">266</span><a name="line.266"></a>
+<span class="sourceLineNo">267</span>  /**<a name="line.267"></a>
+<span class="sourceLineNo">268</span>   * The main function for the RegionSplitter application. Common uses:<a name="line.268"></a>
+<span class="sourceLineNo">269</span>   * &lt;p&gt;<a name="line.269"></a>
+<span class="sourceLineNo">270</span>   * &lt;ul&gt;<a name="line.270"></a>
+<span class="sourceLineNo">271</span>   * &lt;li&gt;create a table named 'myTable' with 60 pre-split regions containing 2<a name="line.271"></a>
+<span class="sourceLineNo">272</span>   * column families 'test' &amp;amp; 'rs', assuming the keys are hex-encoded ASCII:<a name="line.272"></a>
+<span class="sourceLineNo">273</span>   * &lt;ul&gt;<a name="line.273"></a>
+<span class="sourceLineNo">274</span>   * &lt;li&gt;bin/hbase org.apache.hadoop.hbase.util.RegionSplitter -c 60 -f test:rs<a name="line.274"></a>
+<span class="sourceLineNo">275</span>   * myTable HexStringSplit<a name="line.275"></a>
+<span class="sourceLineNo">276</span>   * &lt;/ul&gt;<a name="line.276"></a>
+<span class="sourceLineNo">277</span>   * &lt;li&gt;perform a rolling split of 'myTable' (i.e. 60 =&amp;gt; 120 regions), # 2<a name="line.277"></a>
+<span class="sourceLineNo">278</span>   * outstanding splits at a time, assuming keys are uniformly distributed<a name="line.278"></a>
+<span class="sourceLineNo">279</span>   * bytes:<a name="line.279"></a>
+<span class="sourceLineNo">280</span>   * &lt;ul&gt;<a name="line.280"></a>
+<span class="sourceLineNo">281</span>   * &lt;li&gt;bin/hbase org.apache.hadoop.hbase.util.RegionSplitter -r -o 2 myTable<a name="line.281"></a>
+<span class="sourceLineNo">282</span>   * UniformSplit<a name="line.282"></a>
+<span class="sourceLineNo">283</span>   * &lt;/ul&gt;<a name="line.283"></a>
+<span class="sourceLineNo">284</span>   * &lt;/ul&gt;<a name="line.284"></a>
+<span class="sourceLineNo">285</span>   *<a name="line.285"></a>
+<span class="sourceLineNo">286</span>   * There are two SplitAlgorithms built into RegionSplitter, HexStringSplit<a name="line.286"></a>
+<span class="sourceLineNo">287</span>   * and UniformSplit. These are different strategies for choosing region<a name="line.287"></a>
+<span class="sourceLineNo">288</span>   * boundaries. See their source code for details.<a name="line.288"></a>
+<span class="sourceLineNo">289</span>   *<a name="line.289"></a>
+<span class="sourceLineNo">290</span>   * @param args<a name="line.290"></a>
+<span class="sourceLineNo">291</span>   *          Usage: RegionSplitter &amp;lt;TABLE&amp;gt; &amp;lt;SPLITALGORITHM&amp;gt;<a name="line.291"></a>
+<span class="sourceLineNo">292</span>   *          &amp;lt;-c &amp;lt;# regions&amp;gt; -f &amp;lt;family:family:...&amp;gt; | -r<a name="line.292"></a>
+<span class="sourceLineNo">293</span>   *          [-o &amp;lt;# outstanding splits&amp;gt;]&amp;gt;<a name="line.293"></a>
+<span class="sourceLineNo">294</span>   *          [-D &amp;lt;conf.param=value&amp;gt;]<a name="line.294"></a>
+<span class="sourceLineNo">295</span>   * @throws IOException<a name="line.295"></a>
+<span class="sourceLineNo">296</span>   *           HBase IO problem<a name="line.296"></a>
+<span class="sourceLineNo">297</span>   * @throws InterruptedException<a name="line.297"></a>
+<span class="sourceLineNo">298</span>   *           user requested exit<a name="line.298"></a>
+<span class="sourceLineNo">299</span>   * @throws ParseException<a name="line.299"></a>
+<span class="sourceLineNo">300</span>   *           problem parsing user input<a name="line.300"></a>
+<span class="sourceLineNo">301</span>   */<a name="line.301"></a>
+<span class="sourceLineNo">302</span>  @SuppressWarnings("static-access")<a name="line.302"></a>
+<span class="sourceLineNo">303</span>  public static void main(String[] args) throws IOException,<a name="line.303"></a>
+<span class="sourceLineNo">304</span>      InterruptedException, ParseException {<a name="line.304"></a>
+<span class="sourceLineNo">305</span>    Configuration conf = HBaseConfiguration.create();<a name="line.305"></a>
+<span class="sourceLineNo">306</span><a name="line.306"></a>
+<span class="sourceLineNo">307</span>    // parse user input<a name="line.307"></a>
+<span class="sourceLineNo">308</span>    Options opt = new Options();<a name="line.308"></a>
+<span class="sourceLineNo">309</span>    opt.addOption(OptionBuilder.withArgName("property=value").hasArg()<a name="line.309"></a>
+<span class="sourceLineNo">310</span>        .withDescription("Override HBase Configuration Settings").create("D"));<a name="line.310"></a>
+<span class="sourceLineNo">311</span>    opt.addOption(OptionBuilder.withArgName("region count").hasArg()<a name="line.311"></a>
+<span class="sourceLineNo">312</span>        .withDescription(<a name="line.312"></a>
+<span class="sourceLineNo">313</span>            "Create a new table with a pre-split number of regions")<a name="line.313"></a>
+<span class="sourceLineNo">314</span>        .create("c"));<a name="line.314"></a>
+<span class="sourceLineNo">315</span>    opt.addOption(OptionBuilder.withArgName("family:family:...").hasArg()<a name="line.315"></a>
+<span class="sourceLineNo">316</span>        .withDescription(<a name="line.316"></a>
+<span class="sourceLineNo">317</span>            "Column Families to create with new table.  Required with -c")<a name="line.317"></a>
+<span class="sourceLineNo">318</span>        .create("f"));<a name="line.318"></a>
+<span class="sourceLineNo">319</span>    opt.addOption("h", false, "Print this usage help");<a name="line.319"></a>
+<span class="sourceLineNo">320</span>    opt.addOption("r", false, "Perform a rolling split of an existing region");<a name="line.320"></a>
+<span class="sourceLineNo">321</span>    opt.addOption(OptionBuilder.withArgName("count").hasArg().withDescription(<a name="line.321"></a>
+<span class="sourceLineNo">322</span>        "Max outstanding splits that have unfinished major compactions")<a name="line.322"></a>
+<span class="sourceLineNo">323</span>        .create("o"));<a name="line.323"></a>
+<span class="sourceLineNo">324</span>    opt.addOption(null, "firstrow", true,<a name="line.324"></a>
+<span class="sourceLineNo">325</span>        "First Row in Table for Split Algorithm");<a name="line.325"></a>
+<span class="sourceLineNo">326</span>    opt.addOption(null, "lastrow", true,<a name="line.326"></a>
+<span class="sourceLineNo">327</span>        "Last Row in Table for Split Algorithm");<a name="line.327"></a>
+<span class="sourceLineNo">328</span>    opt.addOption(null, "risky", false,<a name="line.328"></a>
+<span class="sourceLineNo">329</span>        "Skip verification steps to complete quickly."<a name="line.329"></a>
+<span class="sourceLineNo">330</span>            + "STRONGLY DISCOURAGED for production systems.  ");<a name="line.330"></a>
+<span class="sourceLineNo">331</span>    CommandLine cmd = new GnuParser().parse(opt, args);<a name="line.331"></a>
 <span class="sourceLineNo">332</span><a name="line.332"></a>
-<span class="sourceLineNo">333</span>    if (cmd.hasOption("risky")) {<a name="line.333"></a>
-<span class="sourceLineNo">334</span>      conf.setBoolean("split.verify", false);<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>    boolean createTable = cmd.hasOption("c") &amp;&amp; cmd.hasOption("f");<a name="line.337"></a>
-<span class="sourceLineNo">338</span>    boolean rollingSplit = cmd.hasOption("r");<a name="line.338"></a>
-<span class="sourceLineNo">339</span>    boolean oneOperOnly = createTable ^ rollingSplit;<a name="line.339"></a>
-<span class="sourceLineNo">340</span><a name="line.340"></a>
-<span class="sourceLineNo">341</span>    if (2 != cmd.getArgList().size() || !oneOperOnly || cmd.hasOption("h")) {<a name="line.341"></a>
-<span class="sourceLineNo">342</span>      new HelpFormatter().printHelp("RegionSplitter &lt;TABLE&gt; &lt;SPLITALGORITHM&gt;\n"+<a name="line.342"></a>
-<span class="sourceLineNo">343</span>          "SPLITALGORITHM is a java class name of a class implementing " +<a name="line.343"></a>
-<span class="sourceLineNo">344</span>          "SplitAlgorithm, or one of the special strings HexStringSplit " +<a name="line.344"></a>
-<span class="sourceLineNo">345</span>          "or UniformSplit, which are built-in split algorithms. " +<a name="line.345"></a>
-<span class="sourceLineNo">346</span>          "HexStringSplit treats keys as hexadecimal ASCII, and " +<a name="line.346"></a>
-<span class="sourceLineNo">347</span>          "UniformSplit treats keys as arbitrary bytes.", opt);<a name="line.347"></a>
-<span class="sourceLineNo">348</span>      return;<a name="line.348"></a>
-<span class="sourceLineNo">349</span>    }<a name="line.349"></a>
-<span class="sourceLineNo">350</span>    TableName tableName = TableName.valueOf(cmd.getArgs()[0]);<a name="line.350"></a>
-<span class="sourceLineNo">351</span>    String splitClass = cmd.getArgs()[1];<a name="line.351"></a>
-<span class="sourceLineNo">352</span>    SplitAlgorithm splitAlgo = newSplitAlgoInstance(conf, splitClass);<a name="line.352"></a>
-<span class="sourceLineNo">353</span><a name="line.353"></a>
-<span class="sourceLineNo">354</span>    if (cmd.hasOption("firstrow")) {<a name="line.354"></a>
-<span class="sourceLineNo">355</span>      splitAlgo.setFirstRow(cmd.getOptionValue("firstrow"));<a name="line.355"></a>
-<span class="sourceLineNo">356</span>    }<a name="line.356"></a>
-<span class="sourceLineNo">357</span>    if (cmd.hasOption("lastrow")) {<a name="line.357"></a>
-<span class="sourceLineNo">358</span>      splitAlgo.setLastRow(cmd.getOptionValue("lastrow"));<a name="line.358"></a>
-<span class="sourceLineNo">359</span>    }<a name="line.359"></a>
-<span class="sourceLineNo">360</span><a name="line.360"></a>
-<span class="sourceLineNo">361</span>    if (createTable) {<a name="line.361"></a>
-<span class="sourceLineNo">362</span>      conf.set("split.count", cmd.getOptionValue("c"));<a name="line.362"></a>
-<span class="sourceLineNo">363</span>      createPresplitTable(tableName, splitAlgo, cmd.getOptionValue("f").split(":"), conf);<a name="line.363"></a>
-<span class="sourceLineNo">364</span>    }<a name="line.364"></a>
+<span class="sourceLineNo">333</span>    if (cmd.hasOption("D")) {<a name="line.333"></a>
+<span class="sourceLineNo">334</span>      for (String confOpt : cmd.getOptionValues("D")) {<a name="line.334"></a>
+<span class="sourceLineNo">335</span>        String[] kv = confOpt.split("=", 2);<a name="line.335"></a>
+<span class="sourceLineNo">336</span>        if (kv.length == 2) {<a name="line.336"></a>
+<span class="sourceLineNo">337</span>          conf.set(kv[0], kv[1]);<a name="line.337"></a>
+<span class="sourceLineNo">338</span>          LOG.debug("-D configuration override: " + kv[0] + "=" + kv[1]);<a name="line.338"></a>
+<span class="sourceLineNo">339</span>        } else {<a name="line.339"></a>
+<span class="sourceLineNo">340</span>          throw new ParseException("-D option format invalid: " + confOpt);<a name="line.340"></a>
+<span class="sourceLineNo">341</span>        }<a name="line.341"></a>
+<span class="sourceLineNo">342</span>      }<a name="line.342"></a>
+<span class="sourceLineNo">343</span>    }<a name="line.343"></a>
+<span class="sourceLineNo">344</span><a name="line.344"></a>
+<span class="sourceLineNo">345</span>    if (cmd.hasOption("risky")) {<a name="line.345"></a>
+<span class="sourceLineNo">346</span>      conf.setBoolean("split.verify", false);<a name="line.346"></a>
+<span class="sourceLineNo">347</span>    }<a name="line.347"></a>
+<span class="sourceLineNo">348</span><a name="line.348"></a>
+<span class="sourceLineNo">349</span>    boolean createTable = cmd.hasOption("c") &amp;&amp; cmd.hasOption("f");<a name="line.349"></a>
+<span class="sourceLineNo">350</span>    boolean rollingSplit = cmd.hasOption("r");<a name="line.350"></a>
+<span class="sourceLineNo">351</span>    boolean oneOperOnly = createTable ^ rollingSplit;<a name="line.351"></a>
+<span class="sourceLineNo">352</span><a name="line.352"></a>
+<span class="sourceLineNo">353</span>    if (2 != cmd.getArgList().size() || !oneOperOnly || cmd.hasOption("h")) {<a name="line.353"></a>
+<span class="sourceLineNo">354</span>      new HelpFormatter().printHelp("RegionSplitter &lt;TABLE&gt; &lt;SPLITALGORITHM&gt;\n"+<a name="line.354"></a>
+<span class="sourceLineNo">355</span>          "SPLITALGORITHM is a java class name of a class implementing " +<a name="line.355"></a>
+<span class="sourceLineNo">356</span>          "SplitAlgorithm, or one of the special strings HexStringSplit " +<a name="line.356"></a>
+<span class="sourceLineNo">357</span>          "or UniformSplit, which are built-in split algorithms. " +<a name="line.357"></a>
+<span class="sourceLineNo">358</span>          "HexStringSplit treats keys as hexadecimal ASCII, and " +<a name="line.358"></a>
+<span class="sourceLineNo">359</span>          "UniformSplit treats keys as arbitrary bytes.", opt);<a name="line.359"></a>
+<span class="sourceLineNo">360</span>      return;<a name="line.360"></a>
+<span class="sourceLineNo">361</span>    }<a name="line.361"></a>
+<span class="sourceLineNo">362</span>    TableName tableName = TableName.valueOf(cmd.getArgs()[0]);<a name="line.362"></a>
+<span class="sourceLineNo">363</span>    String splitClass = cmd.getArgs()[1];<a name="line.363"></a>
+<span class="sourceLineNo">364</span>    SplitAlgorithm splitAlgo = newSplitAlgoInstance(conf, splitClass);<a name="line.364"></a>
 <span class="sourceLineNo">365</span><a name="line.365"></a>
-<span class="sourceLineNo">366</span>    if (rollingSplit) {<a name="line.366"></a>
-<span class="sourceLineNo">367</span>      if (cmd.hasOption("o")) {<a name="line.367"></a>
-<span class="sourceLineNo">368</span>        conf.set("split.outstanding", cmd.getOptionValue("o"));<a name="line.368"></a>
-<span class="sourceLineNo">369</span>      }<a name="line.369"></a>
-<span class="sourceLineNo">370</span>      rollingSplit(tableName, splitAlgo, conf);<a name="line.370"></a>
+<span class="sourceLineNo">366</span>    if (cmd.hasOption("firstrow")) {<a name="line.366"></a>
+<span class="sourceLineNo">367</span>      splitAlgo.setFirstRow(cmd.getOptionValue("firstrow"));<a name="line.367"></a>
+<span class="sourceLineNo">368</span>    }<a name="line.368"></a>
+<span class="sourceLineNo">369</span>    if (cmd.hasOption("lastrow")) {<a name="line.369"></a>
+<span class="sourceLineNo">370</span>      splitAlgo.setLastRow(cmd.getOptionValue("lastrow"));<a name="line.370"></a>
 <span class="sourceLineNo">371</span>    }<a name="line.371"></a>
-<span class="sourceLineNo">372</span>  }<a name="line.372"></a>
-<span class="sourceLineNo">373</span><a name="line.373"></a>
-<span class="sourceLineNo">374</span>  static void createPresplitTable(TableName tableName, SplitAlgorithm splitAlgo,<a name="line.374"></a>
-<span class="sourceLineNo">375</span>          String[] columnFamilies, Configuration conf)<a name="line.375"></a>
-<span class="sourceLineNo">376</span>  throws IOException, InterruptedException {<a name="line.376"></a>
-<span class="sourceLineNo">377</span>    final int splitCount = conf.getInt("split.count", 0);<a name="line.377"></a>
-<span class="sourceLineNo">378</span>    Preconditions.checkArgument(splitCount &gt; 1, "Split count must be &gt; 1");<a name="line.378"></a>
-<span class="sourceLineNo">379</span><a name="line.379"></a>
-<span class="sourceLineNo">380</span>    Preconditions.checkArgument(columnFamilies.length &gt; 0,<a name="line.380"></a>
-<span class="sourceLineNo">381</span>        "Must specify at least one column family. ");<a name="line.381"></a>
-<span class="sourceLineNo">382</span>    LOG.debug("Creating table " + tableName + " with " + columnFamilies.length<a name="line.382"></a>
-<span class="sourceLineNo">383</span>        + " column families.  Presplitting to " + splitCount + " regions");<a name="line.383"></a>
-<span class="sourceLineNo">384</span><a name="line.384"></a>
-<span class="sourceLineNo">385</span>    TableDescriptorBuilder builder = TableDescriptorBuilder.newBuilder(tableName);<a name="line.385"></a>
-<span class="sourceLineNo">386</span>    for (String cf : columnFamilies) {<a name="line.386"></a>
-<span class="sourceLineNo">387</span>      builder.addColumnFamily(ColumnFamilyDescriptorBuilder.of(cf));<a name="line.387"></a>
-<span class="sourceLineNo">388</span>    }<a name="line.388"></a>
-<span class="sourceLineNo">389</span>    try (Connection connection = ConnectionFactory.createConnection(conf)) {<a name="line.389"></a>
-<span class="sourceLineNo">390</span>      Admin admin = connection.getAdmin();<a name="line.390"></a>
-<span class="sourceLineNo">391</span>      try {<a name="line.391"></a>
-<span class="sourceLineNo">392</span>        Preconditions.checkArgument(!admin.tableExists(tableName),<a name="line.392"></a>
-<span class="sourceLineNo">393</span>          "Table already exists: " + tableName);<a name="line.393"></a>
-<span class="sourceLineNo">394</span>        admin.createTable(builder.build(), splitAlgo.split(splitCount));<a name="line.394"></a>
-<span class="sourceLineNo">395</span>      } finally {<a name="line.395"></a>
-<span class="sourceLineNo">396</span>        admin.close();<a name="line.396"></a>
-<span class="sourceLineNo">397</span>      }<a name="line.397"></a>
-<span class="sourceLineNo">398</span>      LOG.debug("Table created!  Waiting for regions to show online in META...");<a name="line.398"></a>
-<span class="sourceLineNo">399</span>      if (!conf.getBoolean("split.verify", true)) {<a name="line.399"></a>
-<span class="sourceLineNo">400</span>        // NOTE: createTable is synchronous on the table, but not on the regions<a name="line.400"></a>
-<span class="sourceLineNo">401</span>        int onlineRegions = 0;<a name="line.401"></a>
-<span class="sourceLineNo">402</span>        while (onlineRegions &lt; splitCount) {<a name="line.402"></a>
-<span class="sourceLineNo">403</span>          onlineRegions = MetaTableAccessor.getRegionCount(connection, tableName);<a name="line.403"></a>
-<span class="sourceLineNo">404</span>          LOG.debug(onlineRegions + " of " + splitCount + " regions online...");<a name="line.404"></a>
-<span class="sourceLineNo">405</span>          if (onlineRegions &lt; splitCount) {<a name="line.405"></a>
-<span class="sourceLineNo">406</span>            Thread.sleep(10 * 1000); // sleep<a name="line.406"></a>
-<span class="sourceLineNo">407</span>          }<a name="line.407"></a>
-<span class="sourceLineNo">408</span>        }<a name="line.408"></a>
+<span class="sourceLineNo">372</span><a name="line.372"></a>
+<span class="sourceLineNo">373</span>    if (createTable) {<a name="line.373"></a>
+<span class="sourceLineNo">374</span>      conf.set("split.count", cmd.getOptionValue("c"));<a name="line.374"></a>
+<span class="sourceLineNo">375</span>      createPresplitTable(tableName, splitAlgo, cmd.getOptionValue("f").split(":"), conf);<a name="line.375"></a>
+<span class="sourceLineNo">376</span>    }<a name="line.376"></a>
+<span class="sourceLineNo">377</span><a name="line.377"></a>
+<span class="sourceLineNo">378</span>    if (rollingSplit) {<a name="line.378"></a>
+<span class="sourceLineNo">379</span>      if (cmd.hasOption("o")) {<a name="line.379"></a>
+<span class="sourceLineNo">380</span>        conf.set("split.outstanding", cmd.getOptionValue("o"));<a name="line.380"></a>
+<span class="sourceLineNo">381</span>      }<a name="line.381"></a>
+<span class="sourceLineNo">382</span>      rollingSplit(tableName, splitAlgo, conf);<a name="line.382"></a>
+<span class="sourceLineNo">383</span>    }<a name="line.383"></a>
+<span class="sourceLineNo">384</span>  }<a name="line.384"></a>
+<span class="sourceLineNo">385</span><a name="line.385"></a>
+<span class="sourceLineNo">386</span>  static void createPresplitTable(TableName tableName, SplitAlgorithm splitAlgo,<a name="line.386"></a>
+<span class="sourceLineNo">387</span>          String[] columnFamilies, Configuration conf)<a name="line.387"></a>
+<span class="sourceLineNo">388</span>  throws IOException, InterruptedException {<a name="line.388"></a>
+<span class="sourceLineNo">389</span>    final int splitCount = conf.getInt("split.count", 0);<a name="line.389"></a>
+<span class="sourceLineNo">390</span>    Preconditions.checkArgument(splitCount &gt; 1, "Split count must be &gt; 1");<a name="line.390"></a>
+<span class="sourceLineNo">391</span><a name="line.391"></a>
+<span class="sourceLineNo">392</span>    Preconditions.checkArgument(columnFamilies.length &gt; 0,<a name="line.392"></a>
+<span class="sourceLineNo">393</span>        "Must specify at least one column family. ");<a name="line.393"></a>
+<span class="sourceLineNo">394</span>    LOG.debug("Creating table " + tableName + " with " + columnFamilies.length<a name="line.394"></a>
+<span class="sourceLineNo">395</span>        + " column families.  Presplitting to " + splitCount + " regions");<a name="line.395"></a>
+<span class="sourceLineNo">396</span><a name="line.396"></a>
+<span class="sourceLineNo">397</span>    TableDescriptorBuilder builder = TableDescriptorBuilder.newBuilder(tableName);<a name="line.397"></a>
+<span class="sourceLineNo">398</span>    for (String cf : columnFamilies) {<a name="line.398"></a>
+<span class="sourceLineNo">399</span>      builder.addColumnFamily(ColumnFamilyDescriptorBuilder.of(cf));<a name="line.399"></a>
+<span class="sourceLineNo">400</span>    }<a name="line.400"></a>
+<span class="sourceLineNo">401</span>    try (Connection connection = ConnectionFactory.createConnection(conf)) {<a name="line.401"></a>
+<span class="sourceLineNo">402</span>      Admin admin = connection.getAdmin();<a name="line.402"></a>
+<span class="sourceLineNo">403</span>      try {<a name="line.403"></a>
+<span class="sourceLineNo">404</span>        Preconditions.checkArgument(!admin.tableExists(tableName),<a name="line.404"></a>
+<span class="sourceLineNo">405</span>          "Table already exists: " + tableName);<a name="line.405"></a>
+<span class="sourceLineNo">406</span>        admin.createTable(builder.build(), splitAlgo.split(splitCount));<a name="line.406"></a>
+<span class="sourceLineNo">407</span>      } finally {<a name="line.407"></a>
+<span class="sourceLineNo">408</span>        admin.close();<a name="line.408"></a>
 <span class="sourceLineNo">409</span>      }<a name="line.409"></a>
-<span class="sourceLineNo">410</span>      LOG.debug("Finished creating table with " + splitCount + " regions");<a name="line.410"></a>
-<span class="sourceLineNo">411</span>    }<a name="line.411"></a>
-<span class="sourceLineNo">412</span>  }<a name="line.412"></a>
-<span class="sourceLineNo">413</span><a name="line.413"></a>
-<span class="sourceLineNo">414</span>  /**<a name="line.414"></a>
-<span class="sourceLineNo">415</span>   * Alternative getCurrentNrHRS which is no longer available.<a name="line.415"></a>
-<span class="sourceLineNo">416</span>   * @param connection<a name="line.416"></a>
-<span class="sourceLineNo">417</span>   * @return Rough count of regionservers out on cluster.<a name="line.417"></a>
-<span class="sourceLineNo">418</span>   * @throws IOException <a name="line.418"></a>
-<span class="sourceLineNo">419</span>   */<a name="line.419"></a>
-<span class="sourceLineNo">420</span>  private static int getRegionServerCount(final Connection connection) throws IOException {<a name="line.420"></a>
-<span class="sourceLineNo">421</span>    try (Admin admin = connection.getAdmin()) {<a name="line.421"></a>
-<span class="sourceLineNo">422</span>      ClusterStatus status = admin.getClusterStatus(EnumSet.of(Option.LIVE_SERVERS));<a name="line.422"></a>
-<span class="sourceLineNo">423</span>      Collection&lt;ServerName&gt; servers = status.getServers();<a name="line.423"></a>
-<span class="sourceLineNo">424</span>      return servers == null || servers.isEmpty()? 0: servers.size();<a name="line.424"></a>
-<span class="sourceLineNo">425</span>    }<a name="line.425"></a>
-<span class="sourceLineNo">426</span>  }<a name="line.426"></a>
-<span class="sourceLineNo">427</span><a name="line.427"></a>
-<span class="sourceLineNo">428</span>  private static byte [] readFile(final FileSystem fs, final Path path) throws IOException {<a name="line.428"></a>
-<span class="sourceLineNo">429</span>    FSDataInputStream tmpIn = fs.open(path);<a name="line.429"></a>
-<span class="sourceLineNo">430</span>    try {<a name="line.430"></a>
-<span class="sourceLineNo">431</span>      byte [] rawData = new byte[tmpIn.available()];<a name="line.431"></a>
-<span class="sourceLineNo">432</span>      tmpIn.readFully(rawData);<a name="line.432"></a>
-<span class="sourceLineNo">433</span>      return rawData;<a name="line.433"></a>
-<span class="sourceLineNo">434</span>    } finally {<a name="line.434"></a>
-<span class="sourceLineNo">435</span>      tmpIn.close();<a name="line.435"></a>
-<span class="sourceLineNo">436</span>    }<a name="line.436"></a>
-<span class="sourceLineNo">437</span>  }<a name="line.437"></a>
-<span class="sourceLineNo">438</span><a name="line.438"></a>
-<span class="sourceLineNo">439</span>  static void rollingSplit(TableName tableName, SplitAlgorithm splitAlgo, Configuration conf)<a name="line.439"></a>
-<span class="sourceLineNo">440</span>  throws IOException, InterruptedException {<a name="line.440"></a>
-<span class="sourceLineNo">441</span>    final int minOS = conf.getInt("split.outstanding", 2);<a name="line.441"></a>
-<span class="sourceLineNo">442</span>    try (Connection connection = ConnectionFactory.createConnection(conf)) {<a name="line.442"></a>
-<span class="sourceLineNo">443</span>      // Max outstanding splits. default == 50% of servers<a name="line.443"></a>
-<span class="sourceLineNo">444</span>      final int MAX_OUTSTANDING = Math.max(getRegionServerCount(connection) / 2, minOS);<a name="line.444"></a>
-<span class="sourceLineNo">445</span><a name="line.445"></a>
-<span class="sourceLineNo">446</span>      Path hbDir = FSUtils.getRootDir(conf);<a name="line.446"></a>
-<span class="sourceLineNo">447</span>      Path tableDir = FSUtils.getTableDir(hbDir, tableName);<a name="line.447"></a>
-<span class="sourceLineNo">448</span>      Path splitFile = new Path(tableDir, "_balancedSplit");<a name="line.448"></a>
-<span class="sourceLineNo">449</span>      FileSystem fs = FileSystem.get(conf);<a name="line.449"></a>
+<span class="sourceLineNo">410</span>      LOG.debug("Table created!  Waiting for regions to show online in META...");<a name="line.410"></a>
+<span class="sourceLineNo">411</span>      if (!conf.getBoolean("split.verify", true)) {<a name="line.411"></a>
+<span class="sourceLineNo">412</span>        // NOTE: createTable is synchronous on the table, but not on the regions<a name="line.412"></a>
+<span class="sourceLineNo">413</span>        int onlineRegions = 0;<a name="line.413"></a>
+<span class="sourceLineNo">414</span>        while (onlineRegions &lt; splitCount) {<a name="line.414"></a>
+<span class="sourceLineNo">415</span>          onlineRegions = MetaTableAccessor.getRegionCount(connection, tableName);<a name="line.415"></a>
+<span class="sourceLineNo">416</span>          LOG.debug(onlineRegions + " of " + splitCount + " regions online...");<a name="line.416"></a>
+<span class="sourceLineNo">417</span>          if (onlineRegions &lt; splitCount) {<a name="line.417"></a>
+<span class="sourceLineNo">418</span>            Thread.sleep(10 * 1000); // sleep<a name="line.418"></a>
+<span class="sourceLineNo">419</span>          }<a name="line.419"></a>
+<span class="sourceLineNo">420</span>        }<a name="line.420"></a>
+<span class="sourceLineNo">421</span>      }<a name="line.421"></a>
+<span class="sourceLineNo">422</span>      LOG.debug("Finished creating table with " + splitCount + " regions");<a name="line.422"></a>
+<span class="sourceLineNo">423</span>    }<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>  /**<a name="line.426"></a>
+<span class="sourceLineNo">427</span>   * Alternative getCurrentNrHRS which is no longer available.<a name="line.427"></a>
+<span class="sourceLineNo">428</span>   * @param connection<a name="line.428"></a>
+<span class="sourceLineNo">429</span>   * @return Rough count of regionservers out on cluster.<a name="line.429"></a>
+<span class="sourceLineNo">430</span>   * @throws IOException <a name="line.430"></a>
+<span class="sourceLineNo">431</span>   */<a name="line.431"></a>
+<span class="sourceLineNo">432</span>  private static int getRegionServerCount(final Connection connection) throws IOException {<a name="line.432"></a>
+<span class="sourceLineNo">433</span>    try (Admin admin = connection.getAdmin()) {<a name="line.433"></a>
+<span class="sourceLineNo">434</span>      ClusterStatus status = admin.getClusterStatus(EnumSet.of(Option.LIVE_SERVERS));<a name="line.434"></a>
+<span class="sourceLineNo">435</span>      Collection&lt;ServerName&gt; servers = status.getServers();<a name="line.435"></a>
+<span class="sourceLineNo">436</span>      return servers == null || servers.isEmpty()? 0: servers.size();<a name="line.436"></a>
+<span class="sourceLineNo">437</span>    }<a name="line.437"></a>
+<span class="sourceLineNo">438</span>  }<a name="line.438"></a>
+<span class="sourceLineNo">439</span><a name="line.439"></a>
+<span class="sourceLineNo">440</span>  private static byte [] readFile(final FileSystem fs, final Path path) throws IOException {<a name="line.440"></a>
+<span class="sourceLineNo">441</span>    FSDataInputStream tmpIn = fs.open(path);<a name="line.441"></a>
+<span class="sourceLineNo">442</span>    try {<a name="line.442"></a>
+<span class="sourceLineNo">443</span>      byte [] rawData = new byte[tmpIn.available()];<a name="line.443"></a>
+<span class="sourceLineNo">444</span>      tmpIn.readFully(rawData);<a name="line.444"></a>
+<span class="sourceLineNo">445</span>      return rawData;<a name="line.445"></a>
+<span class="sourceLineNo">446</span>    } finally {<a name="line.446"></a>
+<span class="sourceLineNo">447</span>      tmpIn.close();<a name="line.447"></a>
+<span class="sourceLineNo">448</span>    }<a name="line.448"></a>
+<span class="sourceLineNo">449</span>  }<a name="line.449"></a>
 <span class="sourceLineNo">450</span><a name="line.450"></a>
-<span class="sourceLineNo">451</span>      // Get a list of daughter regions to create<a name="line.451"></a>
-<span class="sourceLineNo">452</span>      LinkedList&lt;Pair&lt;byte[], byte[]&gt;&gt; tmpRegionSet = null;<a name="line.452"></a>
-<span class="sourceLineNo">453</span>      try (Table table = connection.getTable(tableName)) {<a name="line.453"></a>
-<span class="sourceLineNo">454</span>        tmpRegionSet = getSplits(connection, tableName, splitAlgo);<a name="line.454"></a>
-<span class="sourceLineNo">455</span>      }<a name="line.455"></a>
-<span class="sourceLineNo">456</span>      LinkedList&lt;Pair&lt;byte[], byte[]&gt;&gt; outstanding = Lists.newLinkedList();<a name="line.456"></a>
-<span class="sourceLineNo">457</span>      int splitCount = 0;<a name="line.457"></a>
-<span class="sourceLineNo">458</span>      final int origCount = tmpRegionSet.size();<a name="line.458"></a>
-<span class="sourceLineNo">459</span><a name="line.459"></a>
-<span class="sourceLineNo">460</span>      // all splits must compact &amp; we have 1 compact thread, so 2 split<a name="line.460"></a>
-<span class="sourceLineNo">461</span>      // requests to the same RS can stall the outstanding split queue.<a name="line.461"></a>
-<span class="sourceLineNo">462</span>      // To fix, group the regions into an RS pool and round-robin through it<a name="line.462"></a>
-<span class="sourceLineNo">463</span>      LOG.debug("Bucketing regions by regionserver...");<a name="line.463"></a>
-<span class="sourceLineNo">464</span>      TreeMap&lt;ServerName, LinkedList&lt;Pair&lt;byte[], byte[]&gt;&gt;&gt; daughterRegions =<a name="line.464"></a>
-<span class="sourceLineNo">465</span>          Maps.newTreeMap();<a name="line.465"></a>
-<span class="sourceLineNo">466</span>      // Get a regionLocator.  Need it in below.<a name="line.466"></a>
-<span class="sourceLineNo">467</span>      try (RegionLocator regionLocator = connection.getRegionLocator(tableName)) {<a name="line.467"></a>
-<span class="sourceLineNo">468</span>        for (Pair&lt;byte[], byte[]&gt; dr : tmpRegionSet) {<a name="line.468"></a>
-<span class="sourceLineNo">469</span>          ServerName rsLocation = regionLocator.getRegionLocation(dr.getSecond()).getServerName();<a name="line.469"></a>
-<span class="sourceLineNo">470</span>          if (!daughterRegions.containsKey(rsLocation)) {<a name="line.470"></a>
-<span class="sourceLineNo">471</span>            LinkedList&lt;Pair&lt;byte[], byte[]&gt;&gt; entry = Lists.newLinkedList();<a name="line.471"></a>
-<span class="sourceLineNo">472</span>            daughterRegions.put(rsLocation, entry);<a name="line.472"></a>
-<span class="sourceLineNo">473</span>          }<a name="line.473"></a>
-<span class="sourceLineNo">474</span>          daughterRegions.get(rsLocation).add(dr);<a name="line.474"></a>
-<span class="sourceLineNo">475</span>        }<a name="line.475"></a>
-<span class="sourceLineNo">476</span>        LOG.debug("Done with bucketing.  Split time!");<a name="line.476"></a>
-<span class="sourceLineNo">477</span>        long startTime = System.currentTimeMillis();<a name="line.477"></a>
-<span class="sourceLineNo">478</span><a name="line.478"></a>
-<span class="sourceLineNo">479</span>        // Open the split file and modify it as splits finish<a name="line.479"></a>
-<span class="sourceLineNo">480</span>        byte[] rawData = readFile(fs, splitFile);<a name="line.480"></a>
-<span class="sourceLineNo">481</span><a name="line.481"></a>
-<span class="sourceLineNo">482</span>        FSDataOutputStream splitOut = fs.create(splitFile);<a name="line.482"></a>
-<span class="sourceLineNo">483</span>        try {<a name="line.483"></a>
-<span class="sourceLineNo">484</span>          splitOut.write(rawData);<a name="line.484"></a>
-<span class="sourceLineNo">485</span><a name="line.485"></a>
-<span class="sourceLineNo">486</span>          try {<a name="line.486"></a>
-<span class="sourceLineNo">487</span>            // *** split code ***<a name="line.487"></a>
-<span class="sourceLineNo">488</span>            while (!daughterRegions.isEmpty()) {<a name="line.488"></a>
-<span class="sourceLineNo">489</span>              LOG.debug(daughterRegions.size() + " RS have regions to splt.");<a name="line.489"></a>
+<span class="sourceLineNo">451</span>  static void rollingSplit(TableName tableName, SplitAlgorithm splitAlgo, Configuration conf)<a name="line.451"></a>
+<span class="sourceLineNo">452</span>  throws IOException, InterruptedException {<a name="line.452"></a>
+<span class="sourceLineNo">453</span>    final int minOS = conf.getInt("split.outstanding", 2);<a name="line.453"></a>
+<span class="sourceLineNo">454</span>    try (Connection connection = ConnectionFactory.createConnection(conf)) {<a name="line.454"></a>
+<span class="sourceLineNo">455</span>      // Max outstanding splits. default == 50% of servers<a name="line.455"></a>
+<span class="sourceLineNo">456</span>      final int MAX_OUTSTANDING = Math.max(getRegionServerCount(connection) / 2, minOS);<a name="line.456"></a>
+<span class="sourceLineNo">457</span><a name="line.457"></a>
+<span class="sourceLineNo">458</span>      Path hbDir = FSUtils.getRootDir(conf);<a name="line.458"></a>
+<span class="sourceLineNo">459</span>      Path tableDir = FSUtils.getTableDir(hbDir, tableName);<a name="line.459"></a>
+<span class="sourceLineNo">460</span>      Path splitFile = new Path(tableDir, "_balancedSplit");<a name="line.460"></a>
+<span class="sourceLineNo">461</span>      FileSystem fs = FileSystem.get(conf);<a name="line.461"></a>
+<span class="sourceLineNo">462</span><a name="line.462"></a>
+<span class="sourceLineNo">463</span>      // Get a list of daughter regions to create<a name="line.463"></a>
+<span class="sourceLineNo">464</span>      LinkedList&lt;Pair&lt;byte[], byte[]&gt;&gt; tmpRegionSet = null;<a name="line.464"></a>
+<span class="sourceLineNo">465</span>      try (Table table = connection.getTable(tableName)) {<a name="line.465"></a>
+<span class="sourceLineNo">466</span>        tmpRegionSet = getSplits(connection, tableName, splitAlgo);<a name="line.466"></a>
+<span class="sourceLineNo">467</span>      }<a name="line.467"></a>
+<span class="sourceLineNo">468</span>      LinkedList&lt;Pair&lt;byte[], byte[]&gt;&gt; outstanding = Lists.newLinkedList();<a name="line.468"></a>
+<span class="sourceLineNo">469</span>      int splitCount = 0;<a name="line.469"></a>
+<span class="sourceLineNo">470</span>      final int origCount = tmpRegionSet.size();<a name="line.470"></a>
+<span class="sourceLineNo">471</span><a name="line.471"></a>
+<span class="sourceLineNo">472</span>      // all splits must compact &amp; we have 1 compact thread, so 2 split<a name="line.472"></a>
+<span class="sourceLineNo">473</span>      // requests to the same RS can stall the outstandin

<TRUNCATED>

Mime
View raw message