| <!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.junit.Assert.assertNull;<a name="line.20"></a> |
| <span class="sourceLineNo">021</span>import static org.junit.Assert.assertTrue;<a name="line.21"></a> |
| <span class="sourceLineNo">022</span><a name="line.22"></a> |
| <span class="sourceLineNo">023</span>import java.io.IOException;<a name="line.23"></a> |
| <span class="sourceLineNo">024</span>import java.util.concurrent.atomic.AtomicBoolean;<a name="line.24"></a> |
| <span class="sourceLineNo">025</span>import java.util.concurrent.atomic.AtomicInteger;<a name="line.25"></a> |
| <span class="sourceLineNo">026</span>import java.util.concurrent.atomic.AtomicReference;<a name="line.26"></a> |
| <span class="sourceLineNo">027</span>import org.apache.hadoop.conf.Configuration;<a name="line.27"></a> |
| <span class="sourceLineNo">028</span>import org.apache.hadoop.hbase.HBaseClassTestRule;<a name="line.28"></a> |
| <span class="sourceLineNo">029</span>import org.apache.hadoop.hbase.HBaseTestingUtil;<a name="line.29"></a> |
| <span class="sourceLineNo">030</span>import org.apache.hadoop.hbase.HConstants;<a name="line.30"></a> |
| <span class="sourceLineNo">031</span>import org.apache.hadoop.hbase.TableName;<a name="line.31"></a> |
| <span class="sourceLineNo">032</span>import org.apache.hadoop.hbase.Waiter.Predicate;<a name="line.32"></a> |
| <span class="sourceLineNo">033</span>import org.apache.hadoop.hbase.client.Admin;<a name="line.33"></a> |
| <span class="sourceLineNo">034</span>import org.apache.hadoop.hbase.client.ColumnFamilyDescriptorBuilder;<a name="line.34"></a> |
| <span class="sourceLineNo">035</span>import org.apache.hadoop.hbase.client.Connection;<a name="line.35"></a> |
| <span class="sourceLineNo">036</span>import org.apache.hadoop.hbase.client.ConnectionFactory;<a name="line.36"></a> |
| <span class="sourceLineNo">037</span>import org.apache.hadoop.hbase.client.Consistency;<a name="line.37"></a> |
| <span class="sourceLineNo">038</span>import org.apache.hadoop.hbase.client.Get;<a name="line.38"></a> |
| <span class="sourceLineNo">039</span>import org.apache.hadoop.hbase.client.Table;<a name="line.39"></a> |
| <span class="sourceLineNo">040</span>import org.apache.hadoop.hbase.client.TableDescriptor;<a name="line.40"></a> |
| <span class="sourceLineNo">041</span>import org.apache.hadoop.hbase.replication.regionserver.TestRegionReplicaReplication;<a name="line.41"></a> |
| <span class="sourceLineNo">042</span>import org.apache.hadoop.hbase.testclassification.LargeTests;<a name="line.42"></a> |
| <span class="sourceLineNo">043</span>import org.apache.hadoop.hbase.util.Bytes;<a name="line.43"></a> |
| <span class="sourceLineNo">044</span>import org.apache.hadoop.hbase.util.JVMClusterUtil.RegionServerThread;<a name="line.44"></a> |
| <span class="sourceLineNo">045</span>import org.apache.hadoop.hbase.util.ServerRegionReplicaUtil;<a name="line.45"></a> |
| <span class="sourceLineNo">046</span>import org.junit.After;<a name="line.46"></a> |
| <span class="sourceLineNo">047</span>import org.junit.Before;<a name="line.47"></a> |
| <span class="sourceLineNo">048</span>import org.junit.ClassRule;<a name="line.48"></a> |
| <span class="sourceLineNo">049</span>import org.junit.Rule;<a name="line.49"></a> |
| <span class="sourceLineNo">050</span>import org.junit.Test;<a name="line.50"></a> |
| <span class="sourceLineNo">051</span>import org.junit.experimental.categories.Category;<a name="line.51"></a> |
| <span class="sourceLineNo">052</span>import org.junit.rules.TestName;<a name="line.52"></a> |
| <span class="sourceLineNo">053</span>import org.slf4j.Logger;<a name="line.53"></a> |
| <span class="sourceLineNo">054</span>import org.slf4j.LoggerFactory;<a name="line.54"></a> |
| <span class="sourceLineNo">055</span><a name="line.55"></a> |
| <span class="sourceLineNo">056</span>/**<a name="line.56"></a> |
| <span class="sourceLineNo">057</span> * Tests failover of secondary region replicas.<a name="line.57"></a> |
| <span class="sourceLineNo">058</span> */<a name="line.58"></a> |
| <span class="sourceLineNo">059</span>@Category(LargeTests.class)<a name="line.59"></a> |
| <span class="sourceLineNo">060</span>public class TestRegionReplicaFailover {<a name="line.60"></a> |
| <span class="sourceLineNo">061</span><a name="line.61"></a> |
| <span class="sourceLineNo">062</span> @ClassRule<a name="line.62"></a> |
| <span class="sourceLineNo">063</span> public static final HBaseClassTestRule CLASS_RULE =<a name="line.63"></a> |
| <span class="sourceLineNo">064</span> HBaseClassTestRule.forClass(TestRegionReplicaFailover.class);<a name="line.64"></a> |
| <span class="sourceLineNo">065</span><a name="line.65"></a> |
| <span class="sourceLineNo">066</span> private static final Logger LOG = LoggerFactory.getLogger(TestRegionReplicaReplication.class);<a name="line.66"></a> |
| <span class="sourceLineNo">067</span><a name="line.67"></a> |
| <span class="sourceLineNo">068</span> private static final HBaseTestingUtil HTU = new HBaseTestingUtil();<a name="line.68"></a> |
| <span class="sourceLineNo">069</span><a name="line.69"></a> |
| <span class="sourceLineNo">070</span> private static final int NB_SERVERS = 3;<a name="line.70"></a> |
| <span class="sourceLineNo">071</span><a name="line.71"></a> |
| <span class="sourceLineNo">072</span> protected final byte[][] families =<a name="line.72"></a> |
| <span class="sourceLineNo">073</span> new byte[][] { HBaseTestingUtil.fam1, HBaseTestingUtil.fam2, HBaseTestingUtil.fam3 };<a name="line.73"></a> |
| <span class="sourceLineNo">074</span> protected final byte[] fam = HBaseTestingUtil.fam1;<a name="line.74"></a> |
| <span class="sourceLineNo">075</span> protected final byte[] qual1 = Bytes.toBytes("qual1");<a name="line.75"></a> |
| <span class="sourceLineNo">076</span> protected final byte[] value1 = Bytes.toBytes("value1");<a name="line.76"></a> |
| <span class="sourceLineNo">077</span> protected final byte[] row = Bytes.toBytes("rowA");<a name="line.77"></a> |
| <span class="sourceLineNo">078</span> protected final byte[] row2 = Bytes.toBytes("rowB");<a name="line.78"></a> |
| <span class="sourceLineNo">079</span><a name="line.79"></a> |
| <span class="sourceLineNo">080</span> @Rule<a name="line.80"></a> |
| <span class="sourceLineNo">081</span> public TestName name = new TestName();<a name="line.81"></a> |
| <span class="sourceLineNo">082</span><a name="line.82"></a> |
| <span class="sourceLineNo">083</span> private TableDescriptor htd;<a name="line.83"></a> |
| <span class="sourceLineNo">084</span><a name="line.84"></a> |
| <span class="sourceLineNo">085</span> @Before<a name="line.85"></a> |
| <span class="sourceLineNo">086</span> public void before() throws Exception {<a name="line.86"></a> |
| <span class="sourceLineNo">087</span> Configuration conf = HTU.getConfiguration();<a name="line.87"></a> |
| <span class="sourceLineNo">088</span> // Up the handlers; this test needs more than usual.<a name="line.88"></a> |
| <span class="sourceLineNo">089</span> conf.setInt(HConstants.REGION_SERVER_HIGH_PRIORITY_HANDLER_COUNT, 10);<a name="line.89"></a> |
| <span class="sourceLineNo">090</span> conf.setBoolean(ServerRegionReplicaUtil.REGION_REPLICA_REPLICATION_CONF_KEY, true);<a name="line.90"></a> |
| <span class="sourceLineNo">091</span> conf.setBoolean(ServerRegionReplicaUtil.REGION_REPLICA_WAIT_FOR_PRIMARY_FLUSH_CONF_KEY, true);<a name="line.91"></a> |
| <span class="sourceLineNo">092</span> conf.setInt("replication.stats.thread.period.seconds", 5);<a name="line.92"></a> |
| <span class="sourceLineNo">093</span> conf.setBoolean("hbase.tests.use.shortcircuit.reads", false);<a name="line.93"></a> |
| <span class="sourceLineNo">094</span><a name="line.94"></a> |
| <span class="sourceLineNo">095</span> HTU.startMiniCluster(NB_SERVERS);<a name="line.95"></a> |
| <span class="sourceLineNo">096</span> htd = HTU.createModifyableTableDescriptor(<a name="line.96"></a> |
| <span class="sourceLineNo">097</span> TableName.valueOf(name.getMethodName().substring(0, name.getMethodName().length() - 3)),<a name="line.97"></a> |
| <span class="sourceLineNo">098</span> ColumnFamilyDescriptorBuilder.DEFAULT_MIN_VERSIONS, 3, HConstants.FOREVER,<a name="line.98"></a> |
| <span class="sourceLineNo">099</span> ColumnFamilyDescriptorBuilder.DEFAULT_KEEP_DELETED).setRegionReplication(3).build();<a name="line.99"></a> |
| <span class="sourceLineNo">100</span> HTU.getAdmin().createTable(htd);<a name="line.100"></a> |
| <span class="sourceLineNo">101</span> }<a name="line.101"></a> |
| <span class="sourceLineNo">102</span><a name="line.102"></a> |
| <span class="sourceLineNo">103</span> @After<a name="line.103"></a> |
| <span class="sourceLineNo">104</span> public void after() throws Exception {<a name="line.104"></a> |
| <span class="sourceLineNo">105</span> HTU.deleteTableIfAny(htd.getTableName());<a name="line.105"></a> |
| <span class="sourceLineNo">106</span> HTU.shutdownMiniCluster();<a name="line.106"></a> |
| <span class="sourceLineNo">107</span> }<a name="line.107"></a> |
| <span class="sourceLineNo">108</span><a name="line.108"></a> |
| <span class="sourceLineNo">109</span> /**<a name="line.109"></a> |
| <span class="sourceLineNo">110</span> * Tests the case where a newly created table with region replicas and no data, the secondary<a name="line.110"></a> |
| <span class="sourceLineNo">111</span> * region replicas are available to read immediately.<a name="line.111"></a> |
| <span class="sourceLineNo">112</span> */<a name="line.112"></a> |
| <span class="sourceLineNo">113</span> @Test<a name="line.113"></a> |
| <span class="sourceLineNo">114</span> public void testSecondaryRegionWithEmptyRegion() throws IOException {<a name="line.114"></a> |
| <span class="sourceLineNo">115</span> // Create a new table with region replication, don't put any data. Test that the secondary<a name="line.115"></a> |
| <span class="sourceLineNo">116</span> // region replica is available to read.<a name="line.116"></a> |
| <span class="sourceLineNo">117</span> try (Connection connection = ConnectionFactory.createConnection(HTU.getConfiguration());<a name="line.117"></a> |
| <span class="sourceLineNo">118</span> Table table = connection.getTable(htd.getTableName())) {<a name="line.118"></a> |
| <span class="sourceLineNo">119</span><a name="line.119"></a> |
| <span class="sourceLineNo">120</span> Get get = new Get(row);<a name="line.120"></a> |
| <span class="sourceLineNo">121</span> get.setConsistency(Consistency.TIMELINE);<a name="line.121"></a> |
| <span class="sourceLineNo">122</span> get.setReplicaId(1);<a name="line.122"></a> |
| <span class="sourceLineNo">123</span> table.get(get); // this should not block<a name="line.123"></a> |
| <span class="sourceLineNo">124</span> }<a name="line.124"></a> |
| <span class="sourceLineNo">125</span> }<a name="line.125"></a> |
| <span class="sourceLineNo">126</span><a name="line.126"></a> |
| <span class="sourceLineNo">127</span> /**<a name="line.127"></a> |
| <span class="sourceLineNo">128</span> * Tests the case where if there is some data in the primary region, reopening the region replicas<a name="line.128"></a> |
| <span class="sourceLineNo">129</span> * (enable/disable table, etc) makes the region replicas readable.<a name="line.129"></a> |
| <span class="sourceLineNo">130</span> */<a name="line.130"></a> |
| <span class="sourceLineNo">131</span> @Test<a name="line.131"></a> |
| <span class="sourceLineNo">132</span> public void testSecondaryRegionWithNonEmptyRegion() throws IOException {<a name="line.132"></a> |
| <span class="sourceLineNo">133</span> // Create a new table with region replication and load some data<a name="line.133"></a> |
| <span class="sourceLineNo">134</span> // than disable and enable the table again and verify the data from secondary<a name="line.134"></a> |
| <span class="sourceLineNo">135</span> try (Connection connection = ConnectionFactory.createConnection(HTU.getConfiguration());<a name="line.135"></a> |
| <span class="sourceLineNo">136</span> Table table = connection.getTable(htd.getTableName())) {<a name="line.136"></a> |
| <span class="sourceLineNo">137</span><a name="line.137"></a> |
| <span class="sourceLineNo">138</span> HTU.loadNumericRows(table, fam, 0, 1000);<a name="line.138"></a> |
| <span class="sourceLineNo">139</span><a name="line.139"></a> |
| <span class="sourceLineNo">140</span> HTU.getAdmin().disableTable(htd.getTableName());<a name="line.140"></a> |
| <span class="sourceLineNo">141</span> HTU.getAdmin().enableTable(htd.getTableName());<a name="line.141"></a> |
| <span class="sourceLineNo">142</span><a name="line.142"></a> |
| <span class="sourceLineNo">143</span> HTU.verifyNumericRows(table, fam, 0, 1000, 1);<a name="line.143"></a> |
| <span class="sourceLineNo">144</span> }<a name="line.144"></a> |
| <span class="sourceLineNo">145</span> }<a name="line.145"></a> |
| <span class="sourceLineNo">146</span><a name="line.146"></a> |
| <span class="sourceLineNo">147</span> /**<a name="line.147"></a> |
| <span class="sourceLineNo">148</span> * Tests the case where killing a primary region with unflushed data recovers<a name="line.148"></a> |
| <span class="sourceLineNo">149</span> */<a name="line.149"></a> |
| <span class="sourceLineNo">150</span> @Test<a name="line.150"></a> |
| <span class="sourceLineNo">151</span> public void testPrimaryRegionKill() throws Exception {<a name="line.151"></a> |
| <span class="sourceLineNo">152</span> try (Connection connection = ConnectionFactory.createConnection(HTU.getConfiguration());<a name="line.152"></a> |
| <span class="sourceLineNo">153</span> Table table = connection.getTable(htd.getTableName())) {<a name="line.153"></a> |
| <span class="sourceLineNo">154</span><a name="line.154"></a> |
| <span class="sourceLineNo">155</span> HTU.loadNumericRows(table, fam, 0, 1000);<a name="line.155"></a> |
| <span class="sourceLineNo">156</span><a name="line.156"></a> |
| <span class="sourceLineNo">157</span> // wal replication is async, we have to wait until the replication catches up, or we timeout<a name="line.157"></a> |
| <span class="sourceLineNo">158</span> verifyNumericRowsWithTimeout(table, fam, 0, 1000, 1, 30000);<a name="line.158"></a> |
| <span class="sourceLineNo">159</span> verifyNumericRowsWithTimeout(table, fam, 0, 1000, 2, 30000);<a name="line.159"></a> |
| <span class="sourceLineNo">160</span><a name="line.160"></a> |
| <span class="sourceLineNo">161</span> // we should not have flushed files now, but data in memstores of primary and secondary<a name="line.161"></a> |
| <span class="sourceLineNo">162</span> // kill the primary region replica now, and ensure that when it comes back up, we can still<a name="line.162"></a> |
| <span class="sourceLineNo">163</span> // read from it the same data from primary and secondaries<a name="line.163"></a> |
| <span class="sourceLineNo">164</span> boolean aborted = false;<a name="line.164"></a> |
| <span class="sourceLineNo">165</span> for (RegionServerThread rs : HTU.getMiniHBaseCluster().getRegionServerThreads()) {<a name="line.165"></a> |
| <span class="sourceLineNo">166</span> for (Region r : rs.getRegionServer().getRegions(htd.getTableName())) {<a name="line.166"></a> |
| <span class="sourceLineNo">167</span> if (r.getRegionInfo().getReplicaId() == 0) {<a name="line.167"></a> |
| <span class="sourceLineNo">168</span> LOG.info("Aborting region server hosting primary region replica");<a name="line.168"></a> |
| <span class="sourceLineNo">169</span> rs.getRegionServer().abort("for test");<a name="line.169"></a> |
| <span class="sourceLineNo">170</span> aborted = true;<a name="line.170"></a> |
| <span class="sourceLineNo">171</span> break;<a name="line.171"></a> |
| <span class="sourceLineNo">172</span> }<a name="line.172"></a> |
| <span class="sourceLineNo">173</span> }<a name="line.173"></a> |
| <span class="sourceLineNo">174</span> }<a name="line.174"></a> |
| <span class="sourceLineNo">175</span> assertTrue(aborted);<a name="line.175"></a> |
| <span class="sourceLineNo">176</span><a name="line.176"></a> |
| <span class="sourceLineNo">177</span> // wal replication is async, we have to wait until the replication catches up, or we timeout<a name="line.177"></a> |
| <span class="sourceLineNo">178</span> verifyNumericRowsWithTimeout(table, fam, 0, 1000, 0, 30000);<a name="line.178"></a> |
| <span class="sourceLineNo">179</span> verifyNumericRowsWithTimeout(table, fam, 0, 1000, 1, 30000);<a name="line.179"></a> |
| <span class="sourceLineNo">180</span> verifyNumericRowsWithTimeout(table, fam, 0, 1000, 2, 30000);<a name="line.180"></a> |
| <span class="sourceLineNo">181</span> }<a name="line.181"></a> |
| <span class="sourceLineNo">182</span><a name="line.182"></a> |
| <span class="sourceLineNo">183</span> // restart the region server<a name="line.183"></a> |
| <span class="sourceLineNo">184</span> HTU.getMiniHBaseCluster().startRegionServer();<a name="line.184"></a> |
| <span class="sourceLineNo">185</span> }<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> * wal replication is async, we have to wait until the replication catches up, or we timeout<a name="line.188"></a> |
| <span class="sourceLineNo">189</span> */<a name="line.189"></a> |
| <span class="sourceLineNo">190</span> private void verifyNumericRowsWithTimeout(final Table table, final byte[] f, final int startRow,<a name="line.190"></a> |
| <span class="sourceLineNo">191</span> final int endRow, final int replicaId, final long timeout) throws Exception {<a name="line.191"></a> |
| <span class="sourceLineNo">192</span> try {<a name="line.192"></a> |
| <span class="sourceLineNo">193</span> HTU.waitFor(timeout, new Predicate<Exception>() {<a name="line.193"></a> |
| <span class="sourceLineNo">194</span> @Override<a name="line.194"></a> |
| <span class="sourceLineNo">195</span> public boolean evaluate() throws Exception {<a name="line.195"></a> |
| <span class="sourceLineNo">196</span> try {<a name="line.196"></a> |
| <span class="sourceLineNo">197</span> HTU.verifyNumericRows(table, f, startRow, endRow, replicaId);<a name="line.197"></a> |
| <span class="sourceLineNo">198</span> return true;<a name="line.198"></a> |
| <span class="sourceLineNo">199</span> } catch (AssertionError ae) {<a name="line.199"></a> |
| <span class="sourceLineNo">200</span> return false;<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> } catch (Throwable t) {<a name="line.204"></a> |
| <span class="sourceLineNo">205</span> // ignore this, but redo the verify do get the actual exception<a name="line.205"></a> |
| <span class="sourceLineNo">206</span> HTU.verifyNumericRows(table, f, startRow, endRow, replicaId);<a name="line.206"></a> |
| <span class="sourceLineNo">207</span> }<a name="line.207"></a> |
| <span class="sourceLineNo">208</span> }<a name="line.208"></a> |
| <span class="sourceLineNo">209</span><a name="line.209"></a> |
| <span class="sourceLineNo">210</span> /**<a name="line.210"></a> |
| <span class="sourceLineNo">211</span> * Tests the case where killing a secondary region with unflushed data recovers, and the replica<a name="line.211"></a> |
| <span class="sourceLineNo">212</span> * becomes available to read again shortly.<a name="line.212"></a> |
| <span class="sourceLineNo">213</span> */<a name="line.213"></a> |
| <span class="sourceLineNo">214</span> @Test<a name="line.214"></a> |
| <span class="sourceLineNo">215</span> public void testSecondaryRegionKill() throws Exception {<a name="line.215"></a> |
| <span class="sourceLineNo">216</span> try (Connection connection = ConnectionFactory.createConnection(HTU.getConfiguration());<a name="line.216"></a> |
| <span class="sourceLineNo">217</span> Table table = connection.getTable(htd.getTableName())) {<a name="line.217"></a> |
| <span class="sourceLineNo">218</span> HTU.loadNumericRows(table, fam, 0, 1000);<a name="line.218"></a> |
| <span class="sourceLineNo">219</span><a name="line.219"></a> |
| <span class="sourceLineNo">220</span> // wait for some time to ensure that async wal replication does it's magic<a name="line.220"></a> |
| <span class="sourceLineNo">221</span> verifyNumericRowsWithTimeout(table, fam, 0, 1000, 1, 30000);<a name="line.221"></a> |
| <span class="sourceLineNo">222</span> verifyNumericRowsWithTimeout(table, fam, 0, 1000, 2, 30000);<a name="line.222"></a> |
| <span class="sourceLineNo">223</span><a name="line.223"></a> |
| <span class="sourceLineNo">224</span> // we should not have flushed files now, but data in memstores of primary and secondary<a name="line.224"></a> |
| <span class="sourceLineNo">225</span> // kill the secondary region replica now, and ensure that when it comes back up, we can still<a name="line.225"></a> |
| <span class="sourceLineNo">226</span> // read from it the same data<a name="line.226"></a> |
| <span class="sourceLineNo">227</span> boolean aborted = false;<a name="line.227"></a> |
| <span class="sourceLineNo">228</span> for (RegionServerThread rs : HTU.getMiniHBaseCluster().getRegionServerThreads()) {<a name="line.228"></a> |
| <span class="sourceLineNo">229</span> for (Region r : rs.getRegionServer().getRegions(htd.getTableName())) {<a name="line.229"></a> |
| <span class="sourceLineNo">230</span> if (r.getRegionInfo().getReplicaId() == 1) {<a name="line.230"></a> |
| <span class="sourceLineNo">231</span> LOG.info("Aborting region server hosting secondary region replica");<a name="line.231"></a> |
| <span class="sourceLineNo">232</span> rs.getRegionServer().abort("for test");<a name="line.232"></a> |
| <span class="sourceLineNo">233</span> aborted = true;<a name="line.233"></a> |
| <span class="sourceLineNo">234</span> break;<a name="line.234"></a> |
| <span class="sourceLineNo">235</span> }<a name="line.235"></a> |
| <span class="sourceLineNo">236</span> }<a name="line.236"></a> |
| <span class="sourceLineNo">237</span> }<a name="line.237"></a> |
| <span class="sourceLineNo">238</span> assertTrue(aborted);<a name="line.238"></a> |
| <span class="sourceLineNo">239</span><a name="line.239"></a> |
| <span class="sourceLineNo">240</span> // It takes extra time for replica region is ready for read as during<a name="line.240"></a> |
| <span class="sourceLineNo">241</span> // region open process, it needs to ask primary region to do a flush and replica region<a name="line.241"></a> |
| <span class="sourceLineNo">242</span> // can open newly flushed hfiles to avoid data out-of-sync.<a name="line.242"></a> |
| <span class="sourceLineNo">243</span> verifyNumericRowsWithTimeout(table, fam, 0, 1000, 1, 30000);<a name="line.243"></a> |
| <span class="sourceLineNo">244</span> HTU.verifyNumericRows(table, fam, 0, 1000, 2);<a name="line.244"></a> |
| <span class="sourceLineNo">245</span> }<a name="line.245"></a> |
| <span class="sourceLineNo">246</span><a name="line.246"></a> |
| <span class="sourceLineNo">247</span> // restart the region server<a name="line.247"></a> |
| <span class="sourceLineNo">248</span> HTU.getMiniHBaseCluster().startRegionServer();<a name="line.248"></a> |
| <span class="sourceLineNo">249</span> }<a name="line.249"></a> |
| <span class="sourceLineNo">250</span><a name="line.250"></a> |
| <span class="sourceLineNo">251</span> /**<a name="line.251"></a> |
| <span class="sourceLineNo">252</span> * Tests the case where there are 3 region replicas and the primary is continuously accepting new<a name="line.252"></a> |
| <span class="sourceLineNo">253</span> * writes while one of the secondaries is killed. Verification is done for both of the secondary<a name="line.253"></a> |
| <span class="sourceLineNo">254</span> * replicas.<a name="line.254"></a> |
| <span class="sourceLineNo">255</span> */<a name="line.255"></a> |
| <span class="sourceLineNo">256</span> @Test<a name="line.256"></a> |
| <span class="sourceLineNo">257</span> public void testSecondaryRegionKillWhilePrimaryIsAcceptingWrites() throws Exception {<a name="line.257"></a> |
| <span class="sourceLineNo">258</span> try (Connection connection = ConnectionFactory.createConnection(HTU.getConfiguration());<a name="line.258"></a> |
| <span class="sourceLineNo">259</span> Table table = connection.getTable(htd.getTableName()); Admin admin = connection.getAdmin()) {<a name="line.259"></a> |
| <span class="sourceLineNo">260</span> // start a thread to do the loading of primary<a name="line.260"></a> |
| <span class="sourceLineNo">261</span> HTU.loadNumericRows(table, fam, 0, 1000); // start with some base<a name="line.261"></a> |
| <span class="sourceLineNo">262</span> admin.flush(table.getName());<a name="line.262"></a> |
| <span class="sourceLineNo">263</span> HTU.loadNumericRows(table, fam, 1000, 2000);<a name="line.263"></a> |
| <span class="sourceLineNo">264</span><a name="line.264"></a> |
| <span class="sourceLineNo">265</span> final AtomicReference<Throwable> ex = new AtomicReference<>(null);<a name="line.265"></a> |
| <span class="sourceLineNo">266</span> final AtomicBoolean done = new AtomicBoolean(false);<a name="line.266"></a> |
| <span class="sourceLineNo">267</span> final AtomicInteger key = new AtomicInteger(2000);<a name="line.267"></a> |
| <span class="sourceLineNo">268</span><a name="line.268"></a> |
| <span class="sourceLineNo">269</span> Thread loader = new Thread() {<a name="line.269"></a> |
| <span class="sourceLineNo">270</span> @Override<a name="line.270"></a> |
| <span class="sourceLineNo">271</span> public void run() {<a name="line.271"></a> |
| <span class="sourceLineNo">272</span> while (!done.get()) {<a name="line.272"></a> |
| <span class="sourceLineNo">273</span> try {<a name="line.273"></a> |
| <span class="sourceLineNo">274</span> HTU.loadNumericRows(table, fam, key.get(), key.get() + 1000);<a name="line.274"></a> |
| <span class="sourceLineNo">275</span> key.addAndGet(1000);<a name="line.275"></a> |
| <span class="sourceLineNo">276</span> } catch (Throwable e) {<a name="line.276"></a> |
| <span class="sourceLineNo">277</span> ex.compareAndSet(null, e);<a name="line.277"></a> |
| <span class="sourceLineNo">278</span> }<a name="line.278"></a> |
| <span class="sourceLineNo">279</span> }<a name="line.279"></a> |
| <span class="sourceLineNo">280</span> }<a name="line.280"></a> |
| <span class="sourceLineNo">281</span> };<a name="line.281"></a> |
| <span class="sourceLineNo">282</span> loader.start();<a name="line.282"></a> |
| <span class="sourceLineNo">283</span><a name="line.283"></a> |
| <span class="sourceLineNo">284</span> Thread aborter = new Thread() {<a name="line.284"></a> |
| <span class="sourceLineNo">285</span> @Override<a name="line.285"></a> |
| <span class="sourceLineNo">286</span> public void run() {<a name="line.286"></a> |
| <span class="sourceLineNo">287</span> try {<a name="line.287"></a> |
| <span class="sourceLineNo">288</span> boolean aborted = false;<a name="line.288"></a> |
| <span class="sourceLineNo">289</span> for (RegionServerThread rs : HTU.getMiniHBaseCluster().getRegionServerThreads()) {<a name="line.289"></a> |
| <span class="sourceLineNo">290</span> for (Region r : rs.getRegionServer().getRegions(htd.getTableName())) {<a name="line.290"></a> |
| <span class="sourceLineNo">291</span> if (r.getRegionInfo().getReplicaId() == 1) {<a name="line.291"></a> |
| <span class="sourceLineNo">292</span> LOG.info("Aborting region server hosting secondary region replica");<a name="line.292"></a> |
| <span class="sourceLineNo">293</span> rs.getRegionServer().abort("for test");<a name="line.293"></a> |
| <span class="sourceLineNo">294</span> aborted = true;<a name="line.294"></a> |
| <span class="sourceLineNo">295</span> }<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> assertTrue(aborted);<a name="line.298"></a> |
| <span class="sourceLineNo">299</span> } catch (Throwable e) {<a name="line.299"></a> |
| <span class="sourceLineNo">300</span> ex.compareAndSet(null, e);<a name="line.300"></a> |
| <span class="sourceLineNo">301</span> }<a name="line.301"></a> |
| <span class="sourceLineNo">302</span> }<a name="line.302"></a> |
| <span class="sourceLineNo">303</span> };<a name="line.303"></a> |
| <span class="sourceLineNo">304</span><a name="line.304"></a> |
| <span class="sourceLineNo">305</span> aborter.start();<a name="line.305"></a> |
| <span class="sourceLineNo">306</span> aborter.join();<a name="line.306"></a> |
| <span class="sourceLineNo">307</span> done.set(true);<a name="line.307"></a> |
| <span class="sourceLineNo">308</span> loader.join();<a name="line.308"></a> |
| <span class="sourceLineNo">309</span><a name="line.309"></a> |
| <span class="sourceLineNo">310</span> assertNull(ex.get());<a name="line.310"></a> |
| <span class="sourceLineNo">311</span><a name="line.311"></a> |
| <span class="sourceLineNo">312</span> assertTrue(key.get() > 1000); // assert that the test is working as designed<a name="line.312"></a> |
| <span class="sourceLineNo">313</span> LOG.info("Loaded up to key :" + key.get());<a name="line.313"></a> |
| <span class="sourceLineNo">314</span> verifyNumericRowsWithTimeout(table, fam, 0, key.get(), 0, 30000);<a name="line.314"></a> |
| <span class="sourceLineNo">315</span> verifyNumericRowsWithTimeout(table, fam, 0, key.get(), 1, 30000);<a name="line.315"></a> |
| <span class="sourceLineNo">316</span> verifyNumericRowsWithTimeout(table, fam, 0, key.get(), 2, 30000);<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> // restart the region server<a name="line.319"></a> |
| <span class="sourceLineNo">320</span> HTU.getMiniHBaseCluster().startRegionServer();<a name="line.320"></a> |
| <span class="sourceLineNo">321</span> }<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> * Tests the case where we are creating a table with a lot of regions and replicas. Opening region<a name="line.324"></a> |
| <span class="sourceLineNo">325</span> * replicas should not block handlers on RS indefinitely.<a name="line.325"></a> |
| <span class="sourceLineNo">326</span> */<a name="line.326"></a> |
| <span class="sourceLineNo">327</span> @Test<a name="line.327"></a> |
| <span class="sourceLineNo">328</span> public void testLotsOfRegionReplicas() throws IOException {<a name="line.328"></a> |
| <span class="sourceLineNo">329</span> int numRegions = NB_SERVERS * 20;<a name="line.329"></a> |
| <span class="sourceLineNo">330</span> int regionReplication = 10;<a name="line.330"></a> |
| <span class="sourceLineNo">331</span> String tableName = htd.getTableName().getNameAsString() + "2";<a name="line.331"></a> |
| <span class="sourceLineNo">332</span> htd = HTU<a name="line.332"></a> |
| <span class="sourceLineNo">333</span> .createModifyableTableDescriptor(TableName.valueOf(tableName),<a name="line.333"></a> |
| <span class="sourceLineNo">334</span> ColumnFamilyDescriptorBuilder.DEFAULT_MIN_VERSIONS, 3, HConstants.FOREVER,<a name="line.334"></a> |
| <span class="sourceLineNo">335</span> ColumnFamilyDescriptorBuilder.DEFAULT_KEEP_DELETED)<a name="line.335"></a> |
| <span class="sourceLineNo">336</span> .setRegionReplication(regionReplication).build();<a name="line.336"></a> |
| <span class="sourceLineNo">337</span><a name="line.337"></a> |
| <span class="sourceLineNo">338</span> // dont care about splits themselves too much<a name="line.338"></a> |
| <span class="sourceLineNo">339</span> byte[] startKey = Bytes.toBytes("aaa");<a name="line.339"></a> |
| <span class="sourceLineNo">340</span> byte[] endKey = Bytes.toBytes("zzz");<a name="line.340"></a> |
| <span class="sourceLineNo">341</span> byte[][] splits = HTU.getRegionSplitStartKeys(startKey, endKey, numRegions);<a name="line.341"></a> |
| <span class="sourceLineNo">342</span> HTU.getAdmin().createTable(htd, startKey, endKey, numRegions);<a name="line.342"></a> |
| <span class="sourceLineNo">343</span><a name="line.343"></a> |
| <span class="sourceLineNo">344</span> try (Connection connection = ConnectionFactory.createConnection(HTU.getConfiguration());<a name="line.344"></a> |
| <span class="sourceLineNo">345</span> Table table = connection.getTable(htd.getTableName())) {<a name="line.345"></a> |
| <span class="sourceLineNo">346</span><a name="line.346"></a> |
| <span class="sourceLineNo">347</span> for (int i = 1; i < splits.length; i++) {<a name="line.347"></a> |
| <span class="sourceLineNo">348</span> for (int j = 0; j < regionReplication; j++) {<a name="line.348"></a> |
| <span class="sourceLineNo">349</span> Get get = new Get(splits[i]);<a name="line.349"></a> |
| <span class="sourceLineNo">350</span> get.setConsistency(Consistency.TIMELINE);<a name="line.350"></a> |
| <span class="sourceLineNo">351</span> get.setReplicaId(j);<a name="line.351"></a> |
| <span class="sourceLineNo">352</span> table.get(get); // this should not block. Regions should be coming online<a name="line.352"></a> |
| <span class="sourceLineNo">353</span> }<a name="line.353"></a> |
| <span class="sourceLineNo">354</span> }<a name="line.354"></a> |
| <span class="sourceLineNo">355</span> }<a name="line.355"></a> |
| <span class="sourceLineNo">356</span><a name="line.356"></a> |
| <span class="sourceLineNo">357</span> HTU.deleteTableIfAny(TableName.valueOf(tableName));<a name="line.357"></a> |
| <span class="sourceLineNo">358</span> }<a name="line.358"></a> |
| <span class="sourceLineNo">359</span>}<a name="line.359"></a> |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| </pre> |
| </div> |
| </body> |
| </html> |