blob: 8661cb0ef1cd619907d6309502d76347ec54084e [file] [log] [blame]
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
<html lang="en">
<head>
<title>Source code</title>
<link rel="stylesheet" type="text/css" href="../../../../../../stylesheet.css" title="Style">
</head>
<body>
<div class="sourceContainer">
<pre><span class="sourceLineNo">001</span>/*<a name="line.1"></a>
<span class="sourceLineNo">002</span> * Licensed to the Apache Software Foundation (ASF) under one<a name="line.2"></a>
<span class="sourceLineNo">003</span> * or more contributor license agreements. See the NOTICE file<a name="line.3"></a>
<span class="sourceLineNo">004</span> * distributed with this work for additional information<a name="line.4"></a>
<span class="sourceLineNo">005</span> * regarding copyright ownership. The ASF licenses this file<a name="line.5"></a>
<span class="sourceLineNo">006</span> * to you under the Apache License, Version 2.0 (the<a name="line.6"></a>
<span class="sourceLineNo">007</span> * "License"); you may not use this file except in compliance<a name="line.7"></a>
<span class="sourceLineNo">008</span> * with the License. You may obtain a copy of the License at<a name="line.8"></a>
<span class="sourceLineNo">009</span> *<a name="line.9"></a>
<span class="sourceLineNo">010</span> * http://www.apache.org/licenses/LICENSE-2.0<a name="line.10"></a>
<span class="sourceLineNo">011</span> *<a name="line.11"></a>
<span class="sourceLineNo">012</span> * Unless required by applicable law or agreed to in writing, software<a name="line.12"></a>
<span class="sourceLineNo">013</span> * distributed under the License is distributed on an "AS IS" BASIS,<a name="line.13"></a>
<span class="sourceLineNo">014</span> * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.<a name="line.14"></a>
<span class="sourceLineNo">015</span> * See the License for the specific language governing permissions and<a name="line.15"></a>
<span class="sourceLineNo">016</span> * limitations under the License.<a name="line.16"></a>
<span class="sourceLineNo">017</span> */<a name="line.17"></a>
<span class="sourceLineNo">018</span>package org.apache.hadoop.hbase.regionserver;<a name="line.18"></a>
<span class="sourceLineNo">019</span><a name="line.19"></a>
<span class="sourceLineNo">020</span>import static org.apache.hadoop.hbase.regionserver.HStoreFile.BLOOM_FILTER_PARAM_KEY;<a name="line.20"></a>
<span class="sourceLineNo">021</span>import static org.apache.hadoop.hbase.regionserver.HStoreFile.BLOOM_FILTER_TYPE_KEY;<a name="line.21"></a>
<span class="sourceLineNo">022</span>import static org.apache.hadoop.hbase.regionserver.HStoreFile.COMPACTION_EVENT_KEY;<a name="line.22"></a>
<span class="sourceLineNo">023</span>import static org.apache.hadoop.hbase.regionserver.HStoreFile.DELETE_FAMILY_COUNT;<a name="line.23"></a>
<span class="sourceLineNo">024</span>import static org.apache.hadoop.hbase.regionserver.HStoreFile.EARLIEST_PUT_TS;<a name="line.24"></a>
<span class="sourceLineNo">025</span>import static org.apache.hadoop.hbase.regionserver.HStoreFile.MAJOR_COMPACTION_KEY;<a name="line.25"></a>
<span class="sourceLineNo">026</span>import static org.apache.hadoop.hbase.regionserver.HStoreFile.MAX_SEQ_ID_KEY;<a name="line.26"></a>
<span class="sourceLineNo">027</span>import static org.apache.hadoop.hbase.regionserver.HStoreFile.MOB_CELLS_COUNT;<a name="line.27"></a>
<span class="sourceLineNo">028</span>import static org.apache.hadoop.hbase.regionserver.HStoreFile.MOB_FILE_REFS;<a name="line.28"></a>
<span class="sourceLineNo">029</span>import static org.apache.hadoop.hbase.regionserver.HStoreFile.TIMERANGE_KEY;<a name="line.29"></a>
<span class="sourceLineNo">030</span><a name="line.30"></a>
<span class="sourceLineNo">031</span>import java.io.IOException;<a name="line.31"></a>
<span class="sourceLineNo">032</span>import java.net.InetSocketAddress;<a name="line.32"></a>
<span class="sourceLineNo">033</span>import java.util.Collection;<a name="line.33"></a>
<span class="sourceLineNo">034</span>import java.util.Collections;<a name="line.34"></a>
<span class="sourceLineNo">035</span>import java.util.HashSet;<a name="line.35"></a>
<span class="sourceLineNo">036</span>import java.util.Set;<a name="line.36"></a>
<span class="sourceLineNo">037</span>import java.util.UUID;<a name="line.37"></a>
<span class="sourceLineNo">038</span>import java.util.function.Consumer;<a name="line.38"></a>
<span class="sourceLineNo">039</span>import java.util.function.Supplier;<a name="line.39"></a>
<span class="sourceLineNo">040</span>import java.util.regex.Pattern;<a name="line.40"></a>
<span class="sourceLineNo">041</span>import java.util.stream.Collectors;<a name="line.41"></a>
<span class="sourceLineNo">042</span>import org.apache.hadoop.conf.Configuration;<a name="line.42"></a>
<span class="sourceLineNo">043</span>import org.apache.hadoop.fs.FileSystem;<a name="line.43"></a>
<span class="sourceLineNo">044</span>import org.apache.hadoop.fs.Path;<a name="line.44"></a>
<span class="sourceLineNo">045</span>import org.apache.hadoop.hbase.Cell;<a name="line.45"></a>
<span class="sourceLineNo">046</span>import org.apache.hadoop.hbase.HConstants;<a name="line.46"></a>
<span class="sourceLineNo">047</span>import org.apache.hadoop.hbase.KeyValue;<a name="line.47"></a>
<span class="sourceLineNo">048</span>import org.apache.hadoop.hbase.PrivateCellUtil;<a name="line.48"></a>
<span class="sourceLineNo">049</span>import org.apache.hadoop.hbase.TableName;<a name="line.49"></a>
<span class="sourceLineNo">050</span>import org.apache.hadoop.hbase.client.ColumnFamilyDescriptorBuilder;<a name="line.50"></a>
<span class="sourceLineNo">051</span>import org.apache.hadoop.hbase.io.hfile.CacheConfig;<a name="line.51"></a>
<span class="sourceLineNo">052</span>import org.apache.hadoop.hbase.io.hfile.HFile;<a name="line.52"></a>
<span class="sourceLineNo">053</span>import org.apache.hadoop.hbase.io.hfile.HFileContext;<a name="line.53"></a>
<span class="sourceLineNo">054</span>import org.apache.hadoop.hbase.io.hfile.HFileWriterImpl;<a name="line.54"></a>
<span class="sourceLineNo">055</span>import org.apache.hadoop.hbase.mob.MobUtils;<a name="line.55"></a>
<span class="sourceLineNo">056</span>import org.apache.hadoop.hbase.util.BloomContext;<a name="line.56"></a>
<span class="sourceLineNo">057</span>import org.apache.hadoop.hbase.util.BloomFilterFactory;<a name="line.57"></a>
<span class="sourceLineNo">058</span>import org.apache.hadoop.hbase.util.BloomFilterUtil;<a name="line.58"></a>
<span class="sourceLineNo">059</span>import org.apache.hadoop.hbase.util.BloomFilterWriter;<a name="line.59"></a>
<span class="sourceLineNo">060</span>import org.apache.hadoop.hbase.util.Bytes;<a name="line.60"></a>
<span class="sourceLineNo">061</span>import org.apache.hadoop.hbase.util.CommonFSUtils;<a name="line.61"></a>
<span class="sourceLineNo">062</span>import org.apache.hadoop.hbase.util.RowBloomContext;<a name="line.62"></a>
<span class="sourceLineNo">063</span>import org.apache.hadoop.hbase.util.RowColBloomContext;<a name="line.63"></a>
<span class="sourceLineNo">064</span>import org.apache.hadoop.hbase.util.RowPrefixFixedLengthBloomContext;<a name="line.64"></a>
<span class="sourceLineNo">065</span>import org.apache.yetus.audience.InterfaceAudience;<a name="line.65"></a>
<span class="sourceLineNo">066</span>import org.slf4j.Logger;<a name="line.66"></a>
<span class="sourceLineNo">067</span>import org.slf4j.LoggerFactory;<a name="line.67"></a>
<span class="sourceLineNo">068</span><a name="line.68"></a>
<span class="sourceLineNo">069</span>import org.apache.hbase.thirdparty.com.google.common.base.Preconditions;<a name="line.69"></a>
<span class="sourceLineNo">070</span>import org.apache.hbase.thirdparty.com.google.common.base.Strings;<a name="line.70"></a>
<span class="sourceLineNo">071</span>import org.apache.hbase.thirdparty.com.google.common.collect.SetMultimap;<a name="line.71"></a>
<span class="sourceLineNo">072</span><a name="line.72"></a>
<span class="sourceLineNo">073</span>import org.apache.hadoop.hbase.shaded.protobuf.ProtobufUtil;<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> * A StoreFile writer. Use this to read/write HBase Store Files. It is package local because it is<a name="line.76"></a>
<span class="sourceLineNo">077</span> * an implementation detail of the HBase regionserver.<a name="line.77"></a>
<span class="sourceLineNo">078</span> */<a name="line.78"></a>
<span class="sourceLineNo">079</span>@InterfaceAudience.Private<a name="line.79"></a>
<span class="sourceLineNo">080</span>public class StoreFileWriter implements CellSink, ShipperListener {<a name="line.80"></a>
<span class="sourceLineNo">081</span> private static final Logger LOG = LoggerFactory.getLogger(StoreFileWriter.class.getName());<a name="line.81"></a>
<span class="sourceLineNo">082</span> private static final Pattern dash = Pattern.compile("-");<a name="line.82"></a>
<span class="sourceLineNo">083</span> private final BloomFilterWriter generalBloomFilterWriter;<a name="line.83"></a>
<span class="sourceLineNo">084</span> private final BloomFilterWriter deleteFamilyBloomFilterWriter;<a name="line.84"></a>
<span class="sourceLineNo">085</span> private final BloomType bloomType;<a name="line.85"></a>
<span class="sourceLineNo">086</span> private byte[] bloomParam = null;<a name="line.86"></a>
<span class="sourceLineNo">087</span> private long earliestPutTs = HConstants.LATEST_TIMESTAMP;<a name="line.87"></a>
<span class="sourceLineNo">088</span> private long deleteFamilyCnt = 0;<a name="line.88"></a>
<span class="sourceLineNo">089</span> private BloomContext bloomContext = null;<a name="line.89"></a>
<span class="sourceLineNo">090</span> private BloomContext deleteFamilyBloomContext = null;<a name="line.90"></a>
<span class="sourceLineNo">091</span> private final TimeRangeTracker timeRangeTracker;<a name="line.91"></a>
<span class="sourceLineNo">092</span> private final Supplier&lt;Collection&lt;HStoreFile&gt;&gt; compactedFilesSupplier;<a name="line.92"></a>
<span class="sourceLineNo">093</span><a name="line.93"></a>
<span class="sourceLineNo">094</span> protected HFile.Writer writer;<a name="line.94"></a>
<span class="sourceLineNo">095</span><a name="line.95"></a>
<span class="sourceLineNo">096</span> /**<a name="line.96"></a>
<span class="sourceLineNo">097</span> * Creates an HFile.Writer that also write helpful meta data.<a name="line.97"></a>
<span class="sourceLineNo">098</span> * @param fs file system to write to<a name="line.98"></a>
<span class="sourceLineNo">099</span> * @param path file name to create<a name="line.99"></a>
<span class="sourceLineNo">100</span> * @param conf user configuration<a name="line.100"></a>
<span class="sourceLineNo">101</span> * @param bloomType bloom filter setting<a name="line.101"></a>
<span class="sourceLineNo">102</span> * @param maxKeys the expected maximum number of keys to be added. Was used for<a name="line.102"></a>
<span class="sourceLineNo">103</span> * Bloom filter size in {@link HFile} format version 1.<a name="line.103"></a>
<span class="sourceLineNo">104</span> * @param favoredNodes an array of favored nodes or possibly null<a name="line.104"></a>
<span class="sourceLineNo">105</span> * @param fileContext The HFile context<a name="line.105"></a>
<span class="sourceLineNo">106</span> * @param shouldDropCacheBehind Drop pages written to page cache after writing the store file.<a name="line.106"></a>
<span class="sourceLineNo">107</span> * @param compactedFilesSupplier Returns the {@link HStore} compacted files which not archived<a name="line.107"></a>
<span class="sourceLineNo">108</span> * @throws IOException problem writing to FS<a name="line.108"></a>
<span class="sourceLineNo">109</span> */<a name="line.109"></a>
<span class="sourceLineNo">110</span> private StoreFileWriter(FileSystem fs, Path path, final Configuration conf, CacheConfig cacheConf,<a name="line.110"></a>
<span class="sourceLineNo">111</span> BloomType bloomType, long maxKeys, InetSocketAddress[] favoredNodes, HFileContext fileContext,<a name="line.111"></a>
<span class="sourceLineNo">112</span> boolean shouldDropCacheBehind, Supplier&lt;Collection&lt;HStoreFile&gt;&gt; compactedFilesSupplier)<a name="line.112"></a>
<span class="sourceLineNo">113</span> throws IOException {<a name="line.113"></a>
<span class="sourceLineNo">114</span> this.compactedFilesSupplier = compactedFilesSupplier;<a name="line.114"></a>
<span class="sourceLineNo">115</span> this.timeRangeTracker = TimeRangeTracker.create(TimeRangeTracker.Type.NON_SYNC);<a name="line.115"></a>
<span class="sourceLineNo">116</span> // TODO : Change all writers to be specifically created for compaction context<a name="line.116"></a>
<span class="sourceLineNo">117</span> writer =<a name="line.117"></a>
<span class="sourceLineNo">118</span> HFile.getWriterFactory(conf, cacheConf).withPath(fs, path).withFavoredNodes(favoredNodes)<a name="line.118"></a>
<span class="sourceLineNo">119</span> .withFileContext(fileContext).withShouldDropCacheBehind(shouldDropCacheBehind).create();<a name="line.119"></a>
<span class="sourceLineNo">120</span><a name="line.120"></a>
<span class="sourceLineNo">121</span> generalBloomFilterWriter = BloomFilterFactory.createGeneralBloomAtWrite(conf, cacheConf,<a name="line.121"></a>
<span class="sourceLineNo">122</span> bloomType, (int) Math.min(maxKeys, Integer.MAX_VALUE), writer);<a name="line.122"></a>
<span class="sourceLineNo">123</span><a name="line.123"></a>
<span class="sourceLineNo">124</span> if (generalBloomFilterWriter != null) {<a name="line.124"></a>
<span class="sourceLineNo">125</span> this.bloomType = bloomType;<a name="line.125"></a>
<span class="sourceLineNo">126</span> this.bloomParam = BloomFilterUtil.getBloomFilterParam(bloomType, conf);<a name="line.126"></a>
<span class="sourceLineNo">127</span> if (LOG.isTraceEnabled()) {<a name="line.127"></a>
<span class="sourceLineNo">128</span> LOG.trace("Bloom filter type for " + path + ": " + this.bloomType + ", param: "<a name="line.128"></a>
<span class="sourceLineNo">129</span> + (bloomType == BloomType.ROWPREFIX_FIXED_LENGTH<a name="line.129"></a>
<span class="sourceLineNo">130</span> ? Bytes.toInt(bloomParam)<a name="line.130"></a>
<span class="sourceLineNo">131</span> : Bytes.toStringBinary(bloomParam))<a name="line.131"></a>
<span class="sourceLineNo">132</span> + ", " + generalBloomFilterWriter.getClass().getSimpleName());<a name="line.132"></a>
<span class="sourceLineNo">133</span> }<a name="line.133"></a>
<span class="sourceLineNo">134</span> // init bloom context<a name="line.134"></a>
<span class="sourceLineNo">135</span> switch (bloomType) {<a name="line.135"></a>
<span class="sourceLineNo">136</span> case ROW:<a name="line.136"></a>
<span class="sourceLineNo">137</span> bloomContext =<a name="line.137"></a>
<span class="sourceLineNo">138</span> new RowBloomContext(generalBloomFilterWriter, fileContext.getCellComparator());<a name="line.138"></a>
<span class="sourceLineNo">139</span> break;<a name="line.139"></a>
<span class="sourceLineNo">140</span> case ROWCOL:<a name="line.140"></a>
<span class="sourceLineNo">141</span> bloomContext =<a name="line.141"></a>
<span class="sourceLineNo">142</span> new RowColBloomContext(generalBloomFilterWriter, fileContext.getCellComparator());<a name="line.142"></a>
<span class="sourceLineNo">143</span> break;<a name="line.143"></a>
<span class="sourceLineNo">144</span> case ROWPREFIX_FIXED_LENGTH:<a name="line.144"></a>
<span class="sourceLineNo">145</span> bloomContext = new RowPrefixFixedLengthBloomContext(generalBloomFilterWriter,<a name="line.145"></a>
<span class="sourceLineNo">146</span> fileContext.getCellComparator(), Bytes.toInt(bloomParam));<a name="line.146"></a>
<span class="sourceLineNo">147</span> break;<a name="line.147"></a>
<span class="sourceLineNo">148</span> default:<a name="line.148"></a>
<span class="sourceLineNo">149</span> throw new IOException(<a name="line.149"></a>
<span class="sourceLineNo">150</span> "Invalid Bloom filter type: " + bloomType + " (ROW or ROWCOL or ROWPREFIX expected)");<a name="line.150"></a>
<span class="sourceLineNo">151</span> }<a name="line.151"></a>
<span class="sourceLineNo">152</span> } else {<a name="line.152"></a>
<span class="sourceLineNo">153</span> // Not using Bloom filters.<a name="line.153"></a>
<span class="sourceLineNo">154</span> this.bloomType = BloomType.NONE;<a name="line.154"></a>
<span class="sourceLineNo">155</span> }<a name="line.155"></a>
<span class="sourceLineNo">156</span><a name="line.156"></a>
<span class="sourceLineNo">157</span> // initialize delete family Bloom filter when there is NO RowCol Bloom<a name="line.157"></a>
<span class="sourceLineNo">158</span> // filter<a name="line.158"></a>
<span class="sourceLineNo">159</span> if (this.bloomType != BloomType.ROWCOL) {<a name="line.159"></a>
<span class="sourceLineNo">160</span> this.deleteFamilyBloomFilterWriter = BloomFilterFactory.createDeleteBloomAtWrite(conf,<a name="line.160"></a>
<span class="sourceLineNo">161</span> cacheConf, (int) Math.min(maxKeys, Integer.MAX_VALUE), writer);<a name="line.161"></a>
<span class="sourceLineNo">162</span> deleteFamilyBloomContext =<a name="line.162"></a>
<span class="sourceLineNo">163</span> new RowBloomContext(deleteFamilyBloomFilterWriter, fileContext.getCellComparator());<a name="line.163"></a>
<span class="sourceLineNo">164</span> } else {<a name="line.164"></a>
<span class="sourceLineNo">165</span> deleteFamilyBloomFilterWriter = null;<a name="line.165"></a>
<span class="sourceLineNo">166</span> }<a name="line.166"></a>
<span class="sourceLineNo">167</span> if (deleteFamilyBloomFilterWriter != null &amp;&amp; LOG.isTraceEnabled()) {<a name="line.167"></a>
<span class="sourceLineNo">168</span> LOG.trace("Delete Family Bloom filter type for " + path + ": "<a name="line.168"></a>
<span class="sourceLineNo">169</span> + deleteFamilyBloomFilterWriter.getClass().getSimpleName());<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><a name="line.172"></a>
<span class="sourceLineNo">173</span> public long getPos() throws IOException {<a name="line.173"></a>
<span class="sourceLineNo">174</span> return ((HFileWriterImpl) writer).getPos();<a name="line.174"></a>
<span class="sourceLineNo">175</span> }<a name="line.175"></a>
<span class="sourceLineNo">176</span><a name="line.176"></a>
<span class="sourceLineNo">177</span> /**<a name="line.177"></a>
<span class="sourceLineNo">178</span> * Writes meta data. Call before {@link #close()} since its written as meta data to this file.<a name="line.178"></a>
<span class="sourceLineNo">179</span> * @param maxSequenceId Maximum sequence id.<a name="line.179"></a>
<span class="sourceLineNo">180</span> * @param majorCompaction True if this file is product of a major compaction<a name="line.180"></a>
<span class="sourceLineNo">181</span> * @throws IOException problem writing to FS<a name="line.181"></a>
<span class="sourceLineNo">182</span> */<a name="line.182"></a>
<span class="sourceLineNo">183</span> public void appendMetadata(final long maxSequenceId, final boolean majorCompaction)<a name="line.183"></a>
<span class="sourceLineNo">184</span> throws IOException {<a name="line.184"></a>
<span class="sourceLineNo">185</span> appendMetadata(maxSequenceId, majorCompaction, Collections.emptySet());<a name="line.185"></a>
<span class="sourceLineNo">186</span> }<a name="line.186"></a>
<span class="sourceLineNo">187</span><a name="line.187"></a>
<span class="sourceLineNo">188</span> /**<a name="line.188"></a>
<span class="sourceLineNo">189</span> * Writes meta data. Call before {@link #close()} since its written as meta data to this file.<a name="line.189"></a>
<span class="sourceLineNo">190</span> * @param maxSequenceId Maximum sequence id.<a name="line.190"></a>
<span class="sourceLineNo">191</span> * @param majorCompaction True if this file is product of a major compaction<a name="line.191"></a>
<span class="sourceLineNo">192</span> * @param storeFiles The compacted store files to generate this new file<a name="line.192"></a>
<span class="sourceLineNo">193</span> * @throws IOException problem writing to FS<a name="line.193"></a>
<span class="sourceLineNo">194</span> */<a name="line.194"></a>
<span class="sourceLineNo">195</span> public void appendMetadata(final long maxSequenceId, final boolean majorCompaction,<a name="line.195"></a>
<span class="sourceLineNo">196</span> final Collection&lt;HStoreFile&gt; storeFiles) throws IOException {<a name="line.196"></a>
<span class="sourceLineNo">197</span> writer.appendFileInfo(MAX_SEQ_ID_KEY, Bytes.toBytes(maxSequenceId));<a name="line.197"></a>
<span class="sourceLineNo">198</span> writer.appendFileInfo(MAJOR_COMPACTION_KEY, Bytes.toBytes(majorCompaction));<a name="line.198"></a>
<span class="sourceLineNo">199</span> writer.appendFileInfo(COMPACTION_EVENT_KEY, toCompactionEventTrackerBytes(storeFiles));<a name="line.199"></a>
<span class="sourceLineNo">200</span> appendTrackedTimestampsToMetadata();<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> /**<a name="line.203"></a>
<span class="sourceLineNo">204</span> * Used when write {@link HStoreFile#COMPACTION_EVENT_KEY} to new file's file info. The compacted<a name="line.204"></a>
<span class="sourceLineNo">205</span> * store files's name is needed. But if the compacted store file is a result of compaction, it's<a name="line.205"></a>
<span class="sourceLineNo">206</span> * compacted files which still not archived is needed, too. And don't need to add compacted files<a name="line.206"></a>
<span class="sourceLineNo">207</span> * recursively. If file A, B, C compacted to new file D, and file D compacted to new file E, will<a name="line.207"></a>
<span class="sourceLineNo">208</span> * write A, B, C, D to file E's compacted files. So if file E compacted to new file F, will add E<a name="line.208"></a>
<span class="sourceLineNo">209</span> * to F's compacted files first, then add E's compacted files: A, B, C, D to it. And no need to<a name="line.209"></a>
<span class="sourceLineNo">210</span> * add D's compacted file, as D's compacted files has been in E's compacted files, too. See<a name="line.210"></a>
<span class="sourceLineNo">211</span> * HBASE-20724 for more details.<a name="line.211"></a>
<span class="sourceLineNo">212</span> * @param storeFiles The compacted store files to generate this new file<a name="line.212"></a>
<span class="sourceLineNo">213</span> * @return bytes of CompactionEventTracker<a name="line.213"></a>
<span class="sourceLineNo">214</span> */<a name="line.214"></a>
<span class="sourceLineNo">215</span> private byte[] toCompactionEventTrackerBytes(Collection&lt;HStoreFile&gt; storeFiles) {<a name="line.215"></a>
<span class="sourceLineNo">216</span> Set&lt;String&gt; notArchivedCompactedStoreFiles = this.compactedFilesSupplier.get().stream()<a name="line.216"></a>
<span class="sourceLineNo">217</span> .map(sf -&gt; sf.getPath().getName()).collect(Collectors.toSet());<a name="line.217"></a>
<span class="sourceLineNo">218</span> Set&lt;String&gt; compactedStoreFiles = new HashSet&lt;&gt;();<a name="line.218"></a>
<span class="sourceLineNo">219</span> for (HStoreFile storeFile : storeFiles) {<a name="line.219"></a>
<span class="sourceLineNo">220</span> compactedStoreFiles.add(storeFile.getFileInfo().getPath().getName());<a name="line.220"></a>
<span class="sourceLineNo">221</span> for (String csf : storeFile.getCompactedStoreFiles()) {<a name="line.221"></a>
<span class="sourceLineNo">222</span> if (notArchivedCompactedStoreFiles.contains(csf)) {<a name="line.222"></a>
<span class="sourceLineNo">223</span> compactedStoreFiles.add(csf);<a name="line.223"></a>
<span class="sourceLineNo">224</span> }<a name="line.224"></a>
<span class="sourceLineNo">225</span> }<a name="line.225"></a>
<span class="sourceLineNo">226</span> }<a name="line.226"></a>
<span class="sourceLineNo">227</span> return ProtobufUtil.toCompactionEventTrackerBytes(compactedStoreFiles);<a name="line.227"></a>
<span class="sourceLineNo">228</span> }<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> * Writes meta data. Call before {@link #close()} since its written as meta data to this file.<a name="line.231"></a>
<span class="sourceLineNo">232</span> * @param maxSequenceId Maximum sequence id.<a name="line.232"></a>
<span class="sourceLineNo">233</span> * @param majorCompaction True if this file is product of a major compaction<a name="line.233"></a>
<span class="sourceLineNo">234</span> * @param mobCellsCount The number of mob cells.<a name="line.234"></a>
<span class="sourceLineNo">235</span> * @throws IOException problem writing to FS<a name="line.235"></a>
<span class="sourceLineNo">236</span> */<a name="line.236"></a>
<span class="sourceLineNo">237</span> public void appendMetadata(final long maxSequenceId, final boolean majorCompaction,<a name="line.237"></a>
<span class="sourceLineNo">238</span> final long mobCellsCount) throws IOException {<a name="line.238"></a>
<span class="sourceLineNo">239</span> writer.appendFileInfo(MAX_SEQ_ID_KEY, Bytes.toBytes(maxSequenceId));<a name="line.239"></a>
<span class="sourceLineNo">240</span> writer.appendFileInfo(MAJOR_COMPACTION_KEY, Bytes.toBytes(majorCompaction));<a name="line.240"></a>
<span class="sourceLineNo">241</span> writer.appendFileInfo(MOB_CELLS_COUNT, Bytes.toBytes(mobCellsCount));<a name="line.241"></a>
<span class="sourceLineNo">242</span> appendTrackedTimestampsToMetadata();<a name="line.242"></a>
<span class="sourceLineNo">243</span> }<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> * Appends MOB - specific metadata (even if it is empty)<a name="line.246"></a>
<span class="sourceLineNo">247</span> * @param mobRefSet - original table -&gt; set of MOB file names<a name="line.247"></a>
<span class="sourceLineNo">248</span> * @throws IOException problem writing to FS<a name="line.248"></a>
<span class="sourceLineNo">249</span> */<a name="line.249"></a>
<span class="sourceLineNo">250</span> public void appendMobMetadata(SetMultimap&lt;TableName, String&gt; mobRefSet) throws IOException {<a name="line.250"></a>
<span class="sourceLineNo">251</span> writer.appendFileInfo(MOB_FILE_REFS, MobUtils.serializeMobFileRefs(mobRefSet));<a name="line.251"></a>
<span class="sourceLineNo">252</span> }<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> * Add TimestampRange and earliest put timestamp to Metadata<a name="line.255"></a>
<span class="sourceLineNo">256</span> */<a name="line.256"></a>
<span class="sourceLineNo">257</span> public void appendTrackedTimestampsToMetadata() throws IOException {<a name="line.257"></a>
<span class="sourceLineNo">258</span> // TODO: The StoreFileReader always converts the byte[] to TimeRange<a name="line.258"></a>
<span class="sourceLineNo">259</span> // via TimeRangeTracker, so we should write the serialization data of TimeRange directly.<a name="line.259"></a>
<span class="sourceLineNo">260</span> appendFileInfo(TIMERANGE_KEY, TimeRangeTracker.toByteArray(timeRangeTracker));<a name="line.260"></a>
<span class="sourceLineNo">261</span> appendFileInfo(EARLIEST_PUT_TS, Bytes.toBytes(earliestPutTs));<a name="line.261"></a>
<span class="sourceLineNo">262</span> }<a name="line.262"></a>
<span class="sourceLineNo">263</span><a name="line.263"></a>
<span class="sourceLineNo">264</span> /**<a name="line.264"></a>
<span class="sourceLineNo">265</span> * Record the earlest Put timestamp. If the timeRangeTracker is not set, update TimeRangeTracker<a name="line.265"></a>
<span class="sourceLineNo">266</span> * to include the timestamp of this key<a name="line.266"></a>
<span class="sourceLineNo">267</span> */<a name="line.267"></a>
<span class="sourceLineNo">268</span> public void trackTimestamps(final Cell cell) {<a name="line.268"></a>
<span class="sourceLineNo">269</span> if (KeyValue.Type.Put.getCode() == cell.getTypeByte()) {<a name="line.269"></a>
<span class="sourceLineNo">270</span> earliestPutTs = Math.min(earliestPutTs, cell.getTimestamp());<a name="line.270"></a>
<span class="sourceLineNo">271</span> }<a name="line.271"></a>
<span class="sourceLineNo">272</span> timeRangeTracker.includeTimestamp(cell);<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> private void appendGeneralBloomfilter(final Cell cell) throws IOException {<a name="line.275"></a>
<span class="sourceLineNo">276</span> if (this.generalBloomFilterWriter != null) {<a name="line.276"></a>
<span class="sourceLineNo">277</span> /*<a name="line.277"></a>
<span class="sourceLineNo">278</span> * http://2.bp.blogspot.com/_Cib_A77V54U/StZMrzaKufI/AAAAAAAAADo/ZhK7bGoJdMQ/s400/KeyValue.png<a name="line.278"></a>
<span class="sourceLineNo">279</span> * Key = RowLen + Row + FamilyLen + Column [Family + Qualifier] + Timestamp 3 Types of<a name="line.279"></a>
<span class="sourceLineNo">280</span> * Filtering: 1. Row = Row 2. RowCol = Row + Qualifier 3. RowPrefixFixedLength = Fixed Length<a name="line.280"></a>
<span class="sourceLineNo">281</span> * Row Prefix<a name="line.281"></a>
<span class="sourceLineNo">282</span> */<a name="line.282"></a>
<span class="sourceLineNo">283</span> bloomContext.writeBloom(cell);<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><a name="line.286"></a>
<span class="sourceLineNo">287</span> private void appendDeleteFamilyBloomFilter(final Cell cell) throws IOException {<a name="line.287"></a>
<span class="sourceLineNo">288</span> if (!PrivateCellUtil.isDeleteFamily(cell) &amp;&amp; !PrivateCellUtil.isDeleteFamilyVersion(cell)) {<a name="line.288"></a>
<span class="sourceLineNo">289</span> return;<a name="line.289"></a>
<span class="sourceLineNo">290</span> }<a name="line.290"></a>
<span class="sourceLineNo">291</span><a name="line.291"></a>
<span class="sourceLineNo">292</span> // increase the number of delete family in the store file<a name="line.292"></a>
<span class="sourceLineNo">293</span> deleteFamilyCnt++;<a name="line.293"></a>
<span class="sourceLineNo">294</span> if (this.deleteFamilyBloomFilterWriter != null) {<a name="line.294"></a>
<span class="sourceLineNo">295</span> deleteFamilyBloomContext.writeBloom(cell);<a name="line.295"></a>
<span class="sourceLineNo">296</span> }<a name="line.296"></a>
<span class="sourceLineNo">297</span> }<a name="line.297"></a>
<span class="sourceLineNo">298</span><a name="line.298"></a>
<span class="sourceLineNo">299</span> @Override<a name="line.299"></a>
<span class="sourceLineNo">300</span> public void append(final Cell cell) throws IOException {<a name="line.300"></a>
<span class="sourceLineNo">301</span> appendGeneralBloomfilter(cell);<a name="line.301"></a>
<span class="sourceLineNo">302</span> appendDeleteFamilyBloomFilter(cell);<a name="line.302"></a>
<span class="sourceLineNo">303</span> writer.append(cell);<a name="line.303"></a>
<span class="sourceLineNo">304</span> trackTimestamps(cell);<a name="line.304"></a>
<span class="sourceLineNo">305</span> }<a name="line.305"></a>
<span class="sourceLineNo">306</span><a name="line.306"></a>
<span class="sourceLineNo">307</span> @Override<a name="line.307"></a>
<span class="sourceLineNo">308</span> public void beforeShipped() throws IOException {<a name="line.308"></a>
<span class="sourceLineNo">309</span> // For now these writer will always be of type ShipperListener true.<a name="line.309"></a>
<span class="sourceLineNo">310</span> // TODO : Change all writers to be specifically created for compaction context<a name="line.310"></a>
<span class="sourceLineNo">311</span> writer.beforeShipped();<a name="line.311"></a>
<span class="sourceLineNo">312</span> if (generalBloomFilterWriter != null) {<a name="line.312"></a>
<span class="sourceLineNo">313</span> generalBloomFilterWriter.beforeShipped();<a name="line.313"></a>
<span class="sourceLineNo">314</span> }<a name="line.314"></a>
<span class="sourceLineNo">315</span> if (deleteFamilyBloomFilterWriter != null) {<a name="line.315"></a>
<span class="sourceLineNo">316</span> deleteFamilyBloomFilterWriter.beforeShipped();<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><a name="line.319"></a>
<span class="sourceLineNo">320</span> public Path getPath() {<a name="line.320"></a>
<span class="sourceLineNo">321</span> return this.writer.getPath();<a name="line.321"></a>
<span class="sourceLineNo">322</span> }<a name="line.322"></a>
<span class="sourceLineNo">323</span><a name="line.323"></a>
<span class="sourceLineNo">324</span> public boolean hasGeneralBloom() {<a name="line.324"></a>
<span class="sourceLineNo">325</span> return this.generalBloomFilterWriter != null;<a name="line.325"></a>
<span class="sourceLineNo">326</span> }<a name="line.326"></a>
<span class="sourceLineNo">327</span><a name="line.327"></a>
<span class="sourceLineNo">328</span> /**<a name="line.328"></a>
<span class="sourceLineNo">329</span> * For unit testing only.<a name="line.329"></a>
<span class="sourceLineNo">330</span> * @return the Bloom filter used by this writer.<a name="line.330"></a>
<span class="sourceLineNo">331</span> */<a name="line.331"></a>
<span class="sourceLineNo">332</span> BloomFilterWriter getGeneralBloomWriter() {<a name="line.332"></a>
<span class="sourceLineNo">333</span> return generalBloomFilterWriter;<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> private boolean closeBloomFilter(BloomFilterWriter bfw) throws IOException {<a name="line.336"></a>
<span class="sourceLineNo">337</span> boolean haveBloom = (bfw != null &amp;&amp; bfw.getKeyCount() &gt; 0);<a name="line.337"></a>
<span class="sourceLineNo">338</span> if (haveBloom) {<a name="line.338"></a>
<span class="sourceLineNo">339</span> bfw.compactBloom();<a name="line.339"></a>
<span class="sourceLineNo">340</span> }<a name="line.340"></a>
<span class="sourceLineNo">341</span> return haveBloom;<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> private boolean closeGeneralBloomFilter() throws IOException {<a name="line.344"></a>
<span class="sourceLineNo">345</span> boolean hasGeneralBloom = closeBloomFilter(generalBloomFilterWriter);<a name="line.345"></a>
<span class="sourceLineNo">346</span><a name="line.346"></a>
<span class="sourceLineNo">347</span> // add the general Bloom filter writer and append file info<a name="line.347"></a>
<span class="sourceLineNo">348</span> if (hasGeneralBloom) {<a name="line.348"></a>
<span class="sourceLineNo">349</span> writer.addGeneralBloomFilter(generalBloomFilterWriter);<a name="line.349"></a>
<span class="sourceLineNo">350</span> writer.appendFileInfo(BLOOM_FILTER_TYPE_KEY, Bytes.toBytes(bloomType.toString()));<a name="line.350"></a>
<span class="sourceLineNo">351</span> if (bloomParam != null) {<a name="line.351"></a>
<span class="sourceLineNo">352</span> writer.appendFileInfo(BLOOM_FILTER_PARAM_KEY, bloomParam);<a name="line.352"></a>
<span class="sourceLineNo">353</span> }<a name="line.353"></a>
<span class="sourceLineNo">354</span> bloomContext.addLastBloomKey(writer);<a name="line.354"></a>
<span class="sourceLineNo">355</span> }<a name="line.355"></a>
<span class="sourceLineNo">356</span> return hasGeneralBloom;<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> private boolean closeDeleteFamilyBloomFilter() throws IOException {<a name="line.359"></a>
<span class="sourceLineNo">360</span> boolean hasDeleteFamilyBloom = closeBloomFilter(deleteFamilyBloomFilterWriter);<a name="line.360"></a>
<span class="sourceLineNo">361</span><a name="line.361"></a>
<span class="sourceLineNo">362</span> // add the delete family Bloom filter writer<a name="line.362"></a>
<span class="sourceLineNo">363</span> if (hasDeleteFamilyBloom) {<a name="line.363"></a>
<span class="sourceLineNo">364</span> writer.addDeleteFamilyBloomFilter(deleteFamilyBloomFilterWriter);<a name="line.364"></a>
<span class="sourceLineNo">365</span> }<a name="line.365"></a>
<span class="sourceLineNo">366</span><a name="line.366"></a>
<span class="sourceLineNo">367</span> // append file info about the number of delete family kvs<a name="line.367"></a>
<span class="sourceLineNo">368</span> // even if there is no delete family Bloom.<a name="line.368"></a>
<span class="sourceLineNo">369</span> writer.appendFileInfo(DELETE_FAMILY_COUNT, Bytes.toBytes(this.deleteFamilyCnt));<a name="line.369"></a>
<span class="sourceLineNo">370</span><a name="line.370"></a>
<span class="sourceLineNo">371</span> return hasDeleteFamilyBloom;<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> public void close() throws IOException {<a name="line.374"></a>
<span class="sourceLineNo">375</span> boolean hasGeneralBloom = this.closeGeneralBloomFilter();<a name="line.375"></a>
<span class="sourceLineNo">376</span> boolean hasDeleteFamilyBloom = this.closeDeleteFamilyBloomFilter();<a name="line.376"></a>
<span class="sourceLineNo">377</span><a name="line.377"></a>
<span class="sourceLineNo">378</span> writer.close();<a name="line.378"></a>
<span class="sourceLineNo">379</span><a name="line.379"></a>
<span class="sourceLineNo">380</span> // Log final Bloom filter statistics. This needs to be done after close()<a name="line.380"></a>
<span class="sourceLineNo">381</span> // because compound Bloom filters might be finalized as part of closing.<a name="line.381"></a>
<span class="sourceLineNo">382</span> if (LOG.isTraceEnabled()) {<a name="line.382"></a>
<span class="sourceLineNo">383</span> LOG.trace(<a name="line.383"></a>
<span class="sourceLineNo">384</span> (hasGeneralBloom ? "" : "NO ") + "General Bloom and " + (hasDeleteFamilyBloom ? "" : "NO ")<a name="line.384"></a>
<span class="sourceLineNo">385</span> + "DeleteFamily" + " was added to HFile " + getPath());<a name="line.385"></a>
<span class="sourceLineNo">386</span> }<a name="line.386"></a>
<span class="sourceLineNo">387</span><a name="line.387"></a>
<span class="sourceLineNo">388</span> }<a name="line.388"></a>
<span class="sourceLineNo">389</span><a name="line.389"></a>
<span class="sourceLineNo">390</span> public void appendFileInfo(byte[] key, byte[] value) throws IOException {<a name="line.390"></a>
<span class="sourceLineNo">391</span> writer.appendFileInfo(key, value);<a name="line.391"></a>
<span class="sourceLineNo">392</span> }<a name="line.392"></a>
<span class="sourceLineNo">393</span><a name="line.393"></a>
<span class="sourceLineNo">394</span> /**<a name="line.394"></a>
<span class="sourceLineNo">395</span> * For use in testing.<a name="line.395"></a>
<span class="sourceLineNo">396</span> */<a name="line.396"></a>
<span class="sourceLineNo">397</span> HFile.Writer getHFileWriter() {<a name="line.397"></a>
<span class="sourceLineNo">398</span> return writer;<a name="line.398"></a>
<span class="sourceLineNo">399</span> }<a name="line.399"></a>
<span class="sourceLineNo">400</span><a name="line.400"></a>
<span class="sourceLineNo">401</span> /**<a name="line.401"></a>
<span class="sourceLineNo">402</span> * @param dir Directory to create file in.<a name="line.402"></a>
<span class="sourceLineNo">403</span> * @return random filename inside passed &lt;code&gt;dir&lt;/code&gt;<a name="line.403"></a>
<span class="sourceLineNo">404</span> */<a name="line.404"></a>
<span class="sourceLineNo">405</span> static Path getUniqueFile(final FileSystem fs, final Path dir) throws IOException {<a name="line.405"></a>
<span class="sourceLineNo">406</span> if (!fs.getFileStatus(dir).isDirectory()) {<a name="line.406"></a>
<span class="sourceLineNo">407</span> throw new IOException("Expecting " + dir.toString() + " to be a directory");<a name="line.407"></a>
<span class="sourceLineNo">408</span> }<a name="line.408"></a>
<span class="sourceLineNo">409</span> return new Path(dir, dash.matcher(UUID.randomUUID().toString()).replaceAll(""));<a name="line.409"></a>
<span class="sourceLineNo">410</span> }<a name="line.410"></a>
<span class="sourceLineNo">411</span><a name="line.411"></a>
<span class="sourceLineNo">412</span> @edu.umd.cs.findbugs.annotations.SuppressWarnings(value = "ICAST_INTEGER_MULTIPLY_CAST_TO_LONG",<a name="line.412"></a>
<span class="sourceLineNo">413</span> justification = "Will not overflow")<a name="line.413"></a>
<span class="sourceLineNo">414</span> public static class Builder {<a name="line.414"></a>
<span class="sourceLineNo">415</span> private final Configuration conf;<a name="line.415"></a>
<span class="sourceLineNo">416</span> private final CacheConfig cacheConf;<a name="line.416"></a>
<span class="sourceLineNo">417</span> private final FileSystem fs;<a name="line.417"></a>
<span class="sourceLineNo">418</span><a name="line.418"></a>
<span class="sourceLineNo">419</span> private BloomType bloomType = BloomType.NONE;<a name="line.419"></a>
<span class="sourceLineNo">420</span> private long maxKeyCount = 0;<a name="line.420"></a>
<span class="sourceLineNo">421</span> private Path dir;<a name="line.421"></a>
<span class="sourceLineNo">422</span> private Path filePath;<a name="line.422"></a>
<span class="sourceLineNo">423</span> private InetSocketAddress[] favoredNodes;<a name="line.423"></a>
<span class="sourceLineNo">424</span> private HFileContext fileContext;<a name="line.424"></a>
<span class="sourceLineNo">425</span> private boolean shouldDropCacheBehind;<a name="line.425"></a>
<span class="sourceLineNo">426</span> private Supplier&lt;Collection&lt;HStoreFile&gt;&gt; compactedFilesSupplier = () -&gt; Collections.emptySet();<a name="line.426"></a>
<span class="sourceLineNo">427</span> private String fileStoragePolicy;<a name="line.427"></a>
<span class="sourceLineNo">428</span> // this is used to track the creation of the StoreFileWriter, mainly used for the SFT<a name="line.428"></a>
<span class="sourceLineNo">429</span> // implementation where we will write store files directly to the final place, instead of<a name="line.429"></a>
<span class="sourceLineNo">430</span> // writing a tmp file first. Under this scenario, we will have a background task to purge the<a name="line.430"></a>
<span class="sourceLineNo">431</span> // store files which are not recorded in the SFT, but for the newly created store file writer,<a name="line.431"></a>
<span class="sourceLineNo">432</span> // they are not tracked in SFT, so here we need to record them and treat them specially.<a name="line.432"></a>
<span class="sourceLineNo">433</span> private Consumer&lt;Path&gt; writerCreationTracker;<a name="line.433"></a>
<span class="sourceLineNo">434</span><a name="line.434"></a>
<span class="sourceLineNo">435</span> public Builder(Configuration conf, CacheConfig cacheConf, FileSystem fs) {<a name="line.435"></a>
<span class="sourceLineNo">436</span> this.conf = conf;<a name="line.436"></a>
<span class="sourceLineNo">437</span> this.cacheConf = cacheConf;<a name="line.437"></a>
<span class="sourceLineNo">438</span> this.fs = fs;<a name="line.438"></a>
<span class="sourceLineNo">439</span> }<a name="line.439"></a>
<span class="sourceLineNo">440</span><a name="line.440"></a>
<span class="sourceLineNo">441</span> /**<a name="line.441"></a>
<span class="sourceLineNo">442</span> * Creates Builder with cache configuration disabled<a name="line.442"></a>
<span class="sourceLineNo">443</span> */<a name="line.443"></a>
<span class="sourceLineNo">444</span> public Builder(Configuration conf, FileSystem fs) {<a name="line.444"></a>
<span class="sourceLineNo">445</span> this.conf = conf;<a name="line.445"></a>
<span class="sourceLineNo">446</span> this.cacheConf = CacheConfig.DISABLED;<a name="line.446"></a>
<span class="sourceLineNo">447</span> this.fs = fs;<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> * Use either this method or {@link #withFilePath}, but not both.<a name="line.451"></a>
<span class="sourceLineNo">452</span> * @param dir Path to column family directory. The directory is created if does not exist. The<a name="line.452"></a>
<span class="sourceLineNo">453</span> * file is given a unique name within this directory.<a name="line.453"></a>
<span class="sourceLineNo">454</span> * @return this (for chained invocation)<a name="line.454"></a>
<span class="sourceLineNo">455</span> */<a name="line.455"></a>
<span class="sourceLineNo">456</span> public Builder withOutputDir(Path dir) {<a name="line.456"></a>
<span class="sourceLineNo">457</span> Preconditions.checkNotNull(dir);<a name="line.457"></a>
<span class="sourceLineNo">458</span> this.dir = dir;<a name="line.458"></a>
<span class="sourceLineNo">459</span> return this;<a name="line.459"></a>
<span class="sourceLineNo">460</span> }<a name="line.460"></a>
<span class="sourceLineNo">461</span><a name="line.461"></a>
<span class="sourceLineNo">462</span> /**<a name="line.462"></a>
<span class="sourceLineNo">463</span> * Use either this method or {@link #withOutputDir}, but not both.<a name="line.463"></a>
<span class="sourceLineNo">464</span> * @param filePath the StoreFile path to write<a name="line.464"></a>
<span class="sourceLineNo">465</span> * @return this (for chained invocation)<a name="line.465"></a>
<span class="sourceLineNo">466</span> */<a name="line.466"></a>
<span class="sourceLineNo">467</span> public Builder withFilePath(Path filePath) {<a name="line.467"></a>
<span class="sourceLineNo">468</span> Preconditions.checkNotNull(filePath);<a name="line.468"></a>
<span class="sourceLineNo">469</span> this.filePath = filePath;<a name="line.469"></a>
<span class="sourceLineNo">470</span> return this;<a name="line.470"></a>
<span class="sourceLineNo">471</span> }<a name="line.471"></a>
<span class="sourceLineNo">472</span><a name="line.472"></a>
<span class="sourceLineNo">473</span> /**<a name="line.473"></a>
<span class="sourceLineNo">474</span> * @param favoredNodes an array of favored nodes or possibly null<a name="line.474"></a>
<span class="sourceLineNo">475</span> * @return this (for chained invocation)<a name="line.475"></a>
<span class="sourceLineNo">476</span> */<a name="line.476"></a>
<span class="sourceLineNo">477</span> public Builder withFavoredNodes(InetSocketAddress[] favoredNodes) {<a name="line.477"></a>
<span class="sourceLineNo">478</span> this.favoredNodes = favoredNodes;<a name="line.478"></a>
<span class="sourceLineNo">479</span> return this;<a name="line.479"></a>
<span class="sourceLineNo">480</span> }<a name="line.480"></a>
<span class="sourceLineNo">481</span><a name="line.481"></a>
<span class="sourceLineNo">482</span> public Builder withBloomType(BloomType bloomType) {<a name="line.482"></a>
<span class="sourceLineNo">483</span> Preconditions.checkNotNull(bloomType);<a name="line.483"></a>
<span class="sourceLineNo">484</span> this.bloomType = bloomType;<a name="line.484"></a>
<span class="sourceLineNo">485</span> return this;<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> /**<a name="line.488"></a>
<span class="sourceLineNo">489</span> * @param maxKeyCount estimated maximum number of keys we expect to add<a name="line.489"></a>
<span class="sourceLineNo">490</span> * @return this (for chained invocation)<a name="line.490"></a>
<span class="sourceLineNo">491</span> */<a name="line.491"></a>
<span class="sourceLineNo">492</span> public Builder withMaxKeyCount(long maxKeyCount) {<a name="line.492"></a>
<span class="sourceLineNo">493</span> this.maxKeyCount = maxKeyCount;<a name="line.493"></a>
<span class="sourceLineNo">494</span> return this;<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> public Builder withFileContext(HFileContext fileContext) {<a name="line.497"></a>
<span class="sourceLineNo">498</span> this.fileContext = fileContext;<a name="line.498"></a>
<span class="sourceLineNo">499</span> return this;<a name="line.499"></a>
<span class="sourceLineNo">500</span> }<a name="line.500"></a>
<span class="sourceLineNo">501</span><a name="line.501"></a>
<span class="sourceLineNo">502</span> public Builder withShouldDropCacheBehind(boolean shouldDropCacheBehind) {<a name="line.502"></a>
<span class="sourceLineNo">503</span> this.shouldDropCacheBehind = shouldDropCacheBehind;<a name="line.503"></a>
<span class="sourceLineNo">504</span> return this;<a name="line.504"></a>
<span class="sourceLineNo">505</span> }<a name="line.505"></a>
<span class="sourceLineNo">506</span><a name="line.506"></a>
<span class="sourceLineNo">507</span> public Builder<a name="line.507"></a>
<span class="sourceLineNo">508</span> withCompactedFilesSupplier(Supplier&lt;Collection&lt;HStoreFile&gt;&gt; compactedFilesSupplier) {<a name="line.508"></a>
<span class="sourceLineNo">509</span> this.compactedFilesSupplier = compactedFilesSupplier;<a name="line.509"></a>
<span class="sourceLineNo">510</span> return this;<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> public Builder withFileStoragePolicy(String fileStoragePolicy) {<a name="line.513"></a>
<span class="sourceLineNo">514</span> this.fileStoragePolicy = fileStoragePolicy;<a name="line.514"></a>
<span class="sourceLineNo">515</span> return this;<a name="line.515"></a>
<span class="sourceLineNo">516</span> }<a name="line.516"></a>
<span class="sourceLineNo">517</span><a name="line.517"></a>
<span class="sourceLineNo">518</span> public Builder withWriterCreationTracker(Consumer&lt;Path&gt; writerCreationTracker) {<a name="line.518"></a>
<span class="sourceLineNo">519</span> this.writerCreationTracker = writerCreationTracker;<a name="line.519"></a>
<span class="sourceLineNo">520</span> return this;<a name="line.520"></a>
<span class="sourceLineNo">521</span> }<a name="line.521"></a>
<span class="sourceLineNo">522</span><a name="line.522"></a>
<span class="sourceLineNo">523</span> /**<a name="line.523"></a>
<span class="sourceLineNo">524</span> * Create a store file writer. Client is responsible for closing file when done. If metadata,<a name="line.524"></a>
<span class="sourceLineNo">525</span> * add BEFORE closing using {@link StoreFileWriter#appendMetadata}.<a name="line.525"></a>
<span class="sourceLineNo">526</span> */<a name="line.526"></a>
<span class="sourceLineNo">527</span> public StoreFileWriter build() throws IOException {<a name="line.527"></a>
<span class="sourceLineNo">528</span> if ((dir == null ? 0 : 1) + (filePath == null ? 0 : 1) != 1) {<a name="line.528"></a>
<span class="sourceLineNo">529</span> throw new IllegalArgumentException("Either specify parent directory " + "or file path");<a name="line.529"></a>
<span class="sourceLineNo">530</span> }<a name="line.530"></a>
<span class="sourceLineNo">531</span><a name="line.531"></a>
<span class="sourceLineNo">532</span> if (dir == null) {<a name="line.532"></a>
<span class="sourceLineNo">533</span> dir = filePath.getParent();<a name="line.533"></a>
<span class="sourceLineNo">534</span> }<a name="line.534"></a>
<span class="sourceLineNo">535</span><a name="line.535"></a>
<span class="sourceLineNo">536</span> if (!fs.exists(dir)) {<a name="line.536"></a>
<span class="sourceLineNo">537</span> // Handle permission for non-HDFS filesystem properly<a name="line.537"></a>
<span class="sourceLineNo">538</span> // See HBASE-17710<a name="line.538"></a>
<span class="sourceLineNo">539</span> HRegionFileSystem.mkdirs(fs, conf, dir);<a name="line.539"></a>
<span class="sourceLineNo">540</span> }<a name="line.540"></a>
<span class="sourceLineNo">541</span><a name="line.541"></a>
<span class="sourceLineNo">542</span> // set block storage policy for temp path<a name="line.542"></a>
<span class="sourceLineNo">543</span> String policyName = this.conf.get(ColumnFamilyDescriptorBuilder.STORAGE_POLICY);<a name="line.543"></a>
<span class="sourceLineNo">544</span> if (null == policyName) {<a name="line.544"></a>
<span class="sourceLineNo">545</span> policyName = this.conf.get(HStore.BLOCK_STORAGE_POLICY_KEY);<a name="line.545"></a>
<span class="sourceLineNo">546</span> }<a name="line.546"></a>
<span class="sourceLineNo">547</span> CommonFSUtils.setStoragePolicy(this.fs, dir, policyName);<a name="line.547"></a>
<span class="sourceLineNo">548</span><a name="line.548"></a>
<span class="sourceLineNo">549</span> if (filePath == null) {<a name="line.549"></a>
<span class="sourceLineNo">550</span> // The stored file and related blocks will used the directory based StoragePolicy.<a name="line.550"></a>
<span class="sourceLineNo">551</span> // Because HDFS DistributedFileSystem does not support create files with storage policy<a name="line.551"></a>
<span class="sourceLineNo">552</span> // before version 3.3.0 (See HDFS-13209). Use child dir here is to make stored files<a name="line.552"></a>
<span class="sourceLineNo">553</span> // satisfy the specific storage policy when writing. So as to avoid later data movement.<a name="line.553"></a>
<span class="sourceLineNo">554</span> // We don't want to change whole temp dir to 'fileStoragePolicy'.<a name="line.554"></a>
<span class="sourceLineNo">555</span> if (!Strings.isNullOrEmpty(fileStoragePolicy)) {<a name="line.555"></a>
<span class="sourceLineNo">556</span> dir = new Path(dir, HConstants.STORAGE_POLICY_PREFIX + fileStoragePolicy);<a name="line.556"></a>
<span class="sourceLineNo">557</span> if (!fs.exists(dir)) {<a name="line.557"></a>
<span class="sourceLineNo">558</span> HRegionFileSystem.mkdirs(fs, conf, dir);<a name="line.558"></a>
<span class="sourceLineNo">559</span> LOG.info(<a name="line.559"></a>
<span class="sourceLineNo">560</span> "Create tmp dir " + dir.toString() + " with storage policy: " + fileStoragePolicy);<a name="line.560"></a>
<span class="sourceLineNo">561</span> }<a name="line.561"></a>
<span class="sourceLineNo">562</span> CommonFSUtils.setStoragePolicy(this.fs, dir, fileStoragePolicy);<a name="line.562"></a>
<span class="sourceLineNo">563</span> }<a name="line.563"></a>
<span class="sourceLineNo">564</span> filePath = getUniqueFile(fs, dir);<a name="line.564"></a>
<span class="sourceLineNo">565</span> if (!BloomFilterFactory.isGeneralBloomEnabled(conf)) {<a name="line.565"></a>
<span class="sourceLineNo">566</span> bloomType = BloomType.NONE;<a name="line.566"></a>
<span class="sourceLineNo">567</span> }<a name="line.567"></a>
<span class="sourceLineNo">568</span> }<a name="line.568"></a>
<span class="sourceLineNo">569</span> // make sure we call this before actually create the writer<a name="line.569"></a>
<span class="sourceLineNo">570</span> // in fact, it is not a big deal to even add an inexistent file to the track, as we will never<a name="line.570"></a>
<span class="sourceLineNo">571</span> // try to delete it and finally we will clean the tracker up after compaction. But if the file<a name="line.571"></a>
<span class="sourceLineNo">572</span> // cleaner find the file but we haven't recorded it yet, it may accidentally delete the file<a name="line.572"></a>
<span class="sourceLineNo">573</span> // and cause problem.<a name="line.573"></a>
<span class="sourceLineNo">574</span> if (writerCreationTracker != null) {<a name="line.574"></a>
<span class="sourceLineNo">575</span> writerCreationTracker.accept(filePath);<a name="line.575"></a>
<span class="sourceLineNo">576</span> }<a name="line.576"></a>
<span class="sourceLineNo">577</span> return new StoreFileWriter(fs, filePath, conf, cacheConf, bloomType, maxKeyCount,<a name="line.577"></a>
<span class="sourceLineNo">578</span> favoredNodes, fileContext, shouldDropCacheBehind, compactedFilesSupplier);<a name="line.578"></a>
<span class="sourceLineNo">579</span> }<a name="line.579"></a>
<span class="sourceLineNo">580</span> }<a name="line.580"></a>
<span class="sourceLineNo">581</span>}<a name="line.581"></a>
</pre>
</div>
</body>
</html>