blob: c2fee7b3d94338c97b53fd77f38a4e62ebd2beb5 [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.master;<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.HConstants.HBASE_SPLIT_WAL_MAX_SPLITTER;<a name="line.20"></a>
<span class="sourceLineNo">021</span>import static org.apache.hadoop.hbase.SplitLogCounters.tot_mgr_wait_for_zk_delete;<a name="line.21"></a>
<span class="sourceLineNo">022</span>import static org.junit.Assert.assertEquals;<a name="line.22"></a>
<span class="sourceLineNo">023</span>import static org.junit.Assert.assertTrue;<a name="line.23"></a>
<span class="sourceLineNo">024</span>import static org.junit.Assert.fail;<a name="line.24"></a>
<span class="sourceLineNo">025</span><a name="line.25"></a>
<span class="sourceLineNo">026</span>import java.io.IOException;<a name="line.26"></a>
<span class="sourceLineNo">027</span>import java.util.ArrayList;<a name="line.27"></a>
<span class="sourceLineNo">028</span>import java.util.Arrays;<a name="line.28"></a>
<span class="sourceLineNo">029</span>import java.util.Iterator;<a name="line.29"></a>
<span class="sourceLineNo">030</span>import java.util.List;<a name="line.30"></a>
<span class="sourceLineNo">031</span>import java.util.NavigableSet;<a name="line.31"></a>
<span class="sourceLineNo">032</span>import java.util.concurrent.ExecutorService;<a name="line.32"></a>
<span class="sourceLineNo">033</span>import java.util.concurrent.Executors;<a name="line.33"></a>
<span class="sourceLineNo">034</span>import java.util.concurrent.Future;<a name="line.34"></a>
<span class="sourceLineNo">035</span>import java.util.concurrent.TimeUnit;<a name="line.35"></a>
<span class="sourceLineNo">036</span>import java.util.concurrent.TimeoutException;<a name="line.36"></a>
<span class="sourceLineNo">037</span>import java.util.concurrent.atomic.LongAdder;<a name="line.37"></a>
<span class="sourceLineNo">038</span>import org.apache.hadoop.conf.Configuration;<a name="line.38"></a>
<span class="sourceLineNo">039</span>import org.apache.hadoop.fs.FSDataOutputStream;<a name="line.39"></a>
<span class="sourceLineNo">040</span>import org.apache.hadoop.fs.FileSystem;<a name="line.40"></a>
<span class="sourceLineNo">041</span>import org.apache.hadoop.fs.Path;<a name="line.41"></a>
<span class="sourceLineNo">042</span>import org.apache.hadoop.hbase.HBaseTestingUtil;<a name="line.42"></a>
<span class="sourceLineNo">043</span>import org.apache.hadoop.hbase.HConstants;<a name="line.43"></a>
<span class="sourceLineNo">044</span>import org.apache.hadoop.hbase.KeyValue;<a name="line.44"></a>
<span class="sourceLineNo">045</span>import org.apache.hadoop.hbase.ServerName;<a name="line.45"></a>
<span class="sourceLineNo">046</span>import org.apache.hadoop.hbase.SingleProcessHBaseCluster;<a name="line.46"></a>
<span class="sourceLineNo">047</span>import org.apache.hadoop.hbase.SplitLogCounters;<a name="line.47"></a>
<span class="sourceLineNo">048</span>import org.apache.hadoop.hbase.StartTestingClusterOption;<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.Waiter;<a name="line.50"></a>
<span class="sourceLineNo">051</span>import org.apache.hadoop.hbase.client.Put;<a name="line.51"></a>
<span class="sourceLineNo">052</span>import org.apache.hadoop.hbase.client.RegionInfo;<a name="line.52"></a>
<span class="sourceLineNo">053</span>import org.apache.hadoop.hbase.client.RegionInfoBuilder;<a name="line.53"></a>
<span class="sourceLineNo">054</span>import org.apache.hadoop.hbase.client.RegionLocator;<a name="line.54"></a>
<span class="sourceLineNo">055</span>import org.apache.hadoop.hbase.client.Table;<a name="line.55"></a>
<span class="sourceLineNo">056</span>import org.apache.hadoop.hbase.coordination.ZKSplitLogManagerCoordination;<a name="line.56"></a>
<span class="sourceLineNo">057</span>import org.apache.hadoop.hbase.master.assignment.RegionStates;<a name="line.57"></a>
<span class="sourceLineNo">058</span>import org.apache.hadoop.hbase.regionserver.HRegionServer;<a name="line.58"></a>
<span class="sourceLineNo">059</span>import org.apache.hadoop.hbase.regionserver.MultiVersionConcurrencyControl;<a name="line.59"></a>
<span class="sourceLineNo">060</span>import org.apache.hadoop.hbase.regionserver.Region;<a name="line.60"></a>
<span class="sourceLineNo">061</span>import org.apache.hadoop.hbase.util.Bytes;<a name="line.61"></a>
<span class="sourceLineNo">062</span>import org.apache.hadoop.hbase.util.CommonFSUtils;<a name="line.62"></a>
<span class="sourceLineNo">063</span>import org.apache.hadoop.hbase.util.EnvironmentEdgeManager;<a name="line.63"></a>
<span class="sourceLineNo">064</span>import org.apache.hadoop.hbase.util.JVMClusterUtil.MasterThread;<a name="line.64"></a>
<span class="sourceLineNo">065</span>import org.apache.hadoop.hbase.util.JVMClusterUtil.RegionServerThread;<a name="line.65"></a>
<span class="sourceLineNo">066</span>import org.apache.hadoop.hbase.util.Threads;<a name="line.66"></a>
<span class="sourceLineNo">067</span>import org.apache.hadoop.hbase.wal.WAL;<a name="line.67"></a>
<span class="sourceLineNo">068</span>import org.apache.hadoop.hbase.wal.WALEdit;<a name="line.68"></a>
<span class="sourceLineNo">069</span>import org.apache.hadoop.hbase.wal.WALKeyImpl;<a name="line.69"></a>
<span class="sourceLineNo">070</span>import org.apache.hadoop.hbase.zookeeper.ZKUtil;<a name="line.70"></a>
<span class="sourceLineNo">071</span>import org.junit.After;<a name="line.71"></a>
<span class="sourceLineNo">072</span>import org.junit.AfterClass;<a name="line.72"></a>
<span class="sourceLineNo">073</span>import org.junit.Before;<a name="line.73"></a>
<span class="sourceLineNo">074</span>import org.junit.BeforeClass;<a name="line.74"></a>
<span class="sourceLineNo">075</span>import org.junit.Rule;<a name="line.75"></a>
<span class="sourceLineNo">076</span>import org.junit.Test;<a name="line.76"></a>
<span class="sourceLineNo">077</span>import org.junit.rules.TestName;<a name="line.77"></a>
<span class="sourceLineNo">078</span>import org.slf4j.Logger;<a name="line.78"></a>
<span class="sourceLineNo">079</span>import org.slf4j.LoggerFactory;<a name="line.79"></a>
<span class="sourceLineNo">080</span><a name="line.80"></a>
<span class="sourceLineNo">081</span>import org.apache.hadoop.hbase.shaded.protobuf.ProtobufUtil;<a name="line.81"></a>
<span class="sourceLineNo">082</span><a name="line.82"></a>
<span class="sourceLineNo">083</span>/**<a name="line.83"></a>
<span class="sourceLineNo">084</span> * Base class for testing distributed log splitting.<a name="line.84"></a>
<span class="sourceLineNo">085</span> */<a name="line.85"></a>
<span class="sourceLineNo">086</span>public abstract class AbstractTestDLS {<a name="line.86"></a>
<span class="sourceLineNo">087</span> private static final Logger LOG = LoggerFactory.getLogger(TestSplitLogManager.class);<a name="line.87"></a>
<span class="sourceLineNo">088</span><a name="line.88"></a>
<span class="sourceLineNo">089</span> private static final HBaseTestingUtil TEST_UTIL = new HBaseTestingUtil();<a name="line.89"></a>
<span class="sourceLineNo">090</span><a name="line.90"></a>
<span class="sourceLineNo">091</span> // Start a cluster with 2 masters and 5 regionservers<a name="line.91"></a>
<span class="sourceLineNo">092</span> private static final int NUM_MASTERS = 2;<a name="line.92"></a>
<span class="sourceLineNo">093</span> private static final int NUM_RS = 5;<a name="line.93"></a>
<span class="sourceLineNo">094</span> private static byte[] COLUMN_FAMILY = Bytes.toBytes("family");<a name="line.94"></a>
<span class="sourceLineNo">095</span><a name="line.95"></a>
<span class="sourceLineNo">096</span> @Rule<a name="line.96"></a>
<span class="sourceLineNo">097</span> public TestName testName = new TestName();<a name="line.97"></a>
<span class="sourceLineNo">098</span><a name="line.98"></a>
<span class="sourceLineNo">099</span> private TableName tableName;<a name="line.99"></a>
<span class="sourceLineNo">100</span> private SingleProcessHBaseCluster cluster;<a name="line.100"></a>
<span class="sourceLineNo">101</span> private HMaster master;<a name="line.101"></a>
<span class="sourceLineNo">102</span> private Configuration conf;<a name="line.102"></a>
<span class="sourceLineNo">103</span><a name="line.103"></a>
<span class="sourceLineNo">104</span> @Rule<a name="line.104"></a>
<span class="sourceLineNo">105</span> public TestName name = new TestName();<a name="line.105"></a>
<span class="sourceLineNo">106</span><a name="line.106"></a>
<span class="sourceLineNo">107</span> @BeforeClass<a name="line.107"></a>
<span class="sourceLineNo">108</span> public static void setup() throws Exception {<a name="line.108"></a>
<span class="sourceLineNo">109</span> // Uncomment the following line if more verbosity is needed for<a name="line.109"></a>
<span class="sourceLineNo">110</span> // debugging (see HBASE-12285 for details).<a name="line.110"></a>
<span class="sourceLineNo">111</span> // Logger.getLogger("org.apache.hadoop.hbase").setLevel(Level.DEBUG);<a name="line.111"></a>
<span class="sourceLineNo">112</span> TEST_UTIL.startMiniZKCluster();<a name="line.112"></a>
<span class="sourceLineNo">113</span> TEST_UTIL.startMiniDFSCluster(3);<a name="line.113"></a>
<span class="sourceLineNo">114</span> }<a name="line.114"></a>
<span class="sourceLineNo">115</span><a name="line.115"></a>
<span class="sourceLineNo">116</span> @AfterClass<a name="line.116"></a>
<span class="sourceLineNo">117</span> public static void tearDown() throws Exception {<a name="line.117"></a>
<span class="sourceLineNo">118</span> TEST_UTIL.shutdownMiniCluster();<a name="line.118"></a>
<span class="sourceLineNo">119</span> }<a name="line.119"></a>
<span class="sourceLineNo">120</span><a name="line.120"></a>
<span class="sourceLineNo">121</span> protected abstract String getWalProvider();<a name="line.121"></a>
<span class="sourceLineNo">122</span><a name="line.122"></a>
<span class="sourceLineNo">123</span> private void startCluster(int numRS) throws Exception {<a name="line.123"></a>
<span class="sourceLineNo">124</span> SplitLogCounters.resetCounters();<a name="line.124"></a>
<span class="sourceLineNo">125</span> LOG.info("Starting cluster");<a name="line.125"></a>
<span class="sourceLineNo">126</span> conf.setLong("hbase.splitlog.max.resubmit", 0);<a name="line.126"></a>
<span class="sourceLineNo">127</span> // Make the failure test faster<a name="line.127"></a>
<span class="sourceLineNo">128</span> conf.setInt("zookeeper.recovery.retry", 0);<a name="line.128"></a>
<span class="sourceLineNo">129</span> conf.setInt(HConstants.REGIONSERVER_INFO_PORT, -1);<a name="line.129"></a>
<span class="sourceLineNo">130</span> conf.setFloat(HConstants.LOAD_BALANCER_SLOP_KEY, (float) 100.0); // no load balancing<a name="line.130"></a>
<span class="sourceLineNo">131</span> conf.setInt(HBASE_SPLIT_WAL_MAX_SPLITTER, 3);<a name="line.131"></a>
<span class="sourceLineNo">132</span> conf.setInt(HConstants.REGION_SERVER_HIGH_PRIORITY_HANDLER_COUNT, 10);<a name="line.132"></a>
<span class="sourceLineNo">133</span> conf.set("hbase.wal.provider", getWalProvider());<a name="line.133"></a>
<span class="sourceLineNo">134</span> StartTestingClusterOption option =<a name="line.134"></a>
<span class="sourceLineNo">135</span> StartTestingClusterOption.builder().numMasters(NUM_MASTERS).numRegionServers(numRS).build();<a name="line.135"></a>
<span class="sourceLineNo">136</span> TEST_UTIL.startMiniHBaseCluster(option);<a name="line.136"></a>
<span class="sourceLineNo">137</span> cluster = TEST_UTIL.getHBaseCluster();<a name="line.137"></a>
<span class="sourceLineNo">138</span> LOG.info("Waiting for active/ready master");<a name="line.138"></a>
<span class="sourceLineNo">139</span> cluster.waitForActiveAndReadyMaster();<a name="line.139"></a>
<span class="sourceLineNo">140</span> master = cluster.getMaster();<a name="line.140"></a>
<span class="sourceLineNo">141</span> TEST_UTIL.waitFor(120000, 200, new Waiter.Predicate&lt;Exception&gt;() {<a name="line.141"></a>
<span class="sourceLineNo">142</span> @Override<a name="line.142"></a>
<span class="sourceLineNo">143</span> public boolean evaluate() throws Exception {<a name="line.143"></a>
<span class="sourceLineNo">144</span> return cluster.getLiveRegionServerThreads().size() &gt;= numRS;<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><a name="line.148"></a>
<span class="sourceLineNo">149</span> @Before<a name="line.149"></a>
<span class="sourceLineNo">150</span> public void before() throws Exception {<a name="line.150"></a>
<span class="sourceLineNo">151</span> conf = TEST_UTIL.getConfiguration();<a name="line.151"></a>
<span class="sourceLineNo">152</span> tableName = TableName.valueOf(testName.getMethodName());<a name="line.152"></a>
<span class="sourceLineNo">153</span> }<a name="line.153"></a>
<span class="sourceLineNo">154</span><a name="line.154"></a>
<span class="sourceLineNo">155</span> @After<a name="line.155"></a>
<span class="sourceLineNo">156</span> public void after() throws Exception {<a name="line.156"></a>
<span class="sourceLineNo">157</span> TEST_UTIL.shutdownMiniHBaseCluster();<a name="line.157"></a>
<span class="sourceLineNo">158</span> TEST_UTIL.getTestFileSystem().delete(CommonFSUtils.getRootDir(TEST_UTIL.getConfiguration()),<a name="line.158"></a>
<span class="sourceLineNo">159</span> true);<a name="line.159"></a>
<span class="sourceLineNo">160</span> ZKUtil.deleteNodeRecursively(TEST_UTIL.getZooKeeperWatcher(), "/hbase");<a name="line.160"></a>
<span class="sourceLineNo">161</span> }<a name="line.161"></a>
<span class="sourceLineNo">162</span><a name="line.162"></a>
<span class="sourceLineNo">163</span> @Test<a name="line.163"></a>
<span class="sourceLineNo">164</span> public void testMasterStartsUpWithLogSplittingWork() throws Exception {<a name="line.164"></a>
<span class="sourceLineNo">165</span> conf.setInt(ServerManager.WAIT_ON_REGIONSERVERS_MINTOSTART, NUM_RS - 1);<a name="line.165"></a>
<span class="sourceLineNo">166</span> startCluster(NUM_RS);<a name="line.166"></a>
<span class="sourceLineNo">167</span><a name="line.167"></a>
<span class="sourceLineNo">168</span> int numRegionsToCreate = 40;<a name="line.168"></a>
<span class="sourceLineNo">169</span> int numLogLines = 1000;<a name="line.169"></a>
<span class="sourceLineNo">170</span> // turn off load balancing to prevent regions from moving around otherwise<a name="line.170"></a>
<span class="sourceLineNo">171</span> // they will consume recovered.edits<a name="line.171"></a>
<span class="sourceLineNo">172</span> master.balanceSwitch(false);<a name="line.172"></a>
<span class="sourceLineNo">173</span><a name="line.173"></a>
<span class="sourceLineNo">174</span> try (Table ht = installTable(numRegionsToCreate)) {<a name="line.174"></a>
<span class="sourceLineNo">175</span> HRegionServer hrs = findRSToKill(false);<a name="line.175"></a>
<span class="sourceLineNo">176</span> List&lt;RegionInfo&gt; regions = ProtobufUtil.getOnlineRegions(hrs.getRSRpcServices());<a name="line.176"></a>
<span class="sourceLineNo">177</span> makeWAL(hrs, regions, numLogLines, 100);<a name="line.177"></a>
<span class="sourceLineNo">178</span><a name="line.178"></a>
<span class="sourceLineNo">179</span> // abort master<a name="line.179"></a>
<span class="sourceLineNo">180</span> abortMaster(cluster);<a name="line.180"></a>
<span class="sourceLineNo">181</span><a name="line.181"></a>
<span class="sourceLineNo">182</span> // abort RS<a name="line.182"></a>
<span class="sourceLineNo">183</span> LOG.info("Aborting region server: " + hrs.getServerName());<a name="line.183"></a>
<span class="sourceLineNo">184</span> hrs.abort("testing");<a name="line.184"></a>
<span class="sourceLineNo">185</span><a name="line.185"></a>
<span class="sourceLineNo">186</span> // wait for abort completes<a name="line.186"></a>
<span class="sourceLineNo">187</span> TEST_UTIL.waitFor(120000, 200, new Waiter.Predicate&lt;Exception&gt;() {<a name="line.187"></a>
<span class="sourceLineNo">188</span> @Override<a name="line.188"></a>
<span class="sourceLineNo">189</span> public boolean evaluate() throws Exception {<a name="line.189"></a>
<span class="sourceLineNo">190</span> return cluster.getLiveRegionServerThreads().size() &lt;= NUM_RS - 1;<a name="line.190"></a>
<span class="sourceLineNo">191</span> }<a name="line.191"></a>
<span class="sourceLineNo">192</span> });<a name="line.192"></a>
<span class="sourceLineNo">193</span><a name="line.193"></a>
<span class="sourceLineNo">194</span> Thread.sleep(2000);<a name="line.194"></a>
<span class="sourceLineNo">195</span> LOG.info("Current Open Regions:" + HBaseTestingUtil.getAllOnlineRegions(cluster).size());<a name="line.195"></a>
<span class="sourceLineNo">196</span><a name="line.196"></a>
<span class="sourceLineNo">197</span> // wait for abort completes<a name="line.197"></a>
<span class="sourceLineNo">198</span> TEST_UTIL.waitFor(120000, 200, new Waiter.Predicate&lt;Exception&gt;() {<a name="line.198"></a>
<span class="sourceLineNo">199</span> @Override<a name="line.199"></a>
<span class="sourceLineNo">200</span> public boolean evaluate() throws Exception {<a name="line.200"></a>
<span class="sourceLineNo">201</span> return (HBaseTestingUtil.getAllOnlineRegions(cluster).size() &gt;= (numRegionsToCreate + 1));<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><a name="line.204"></a>
<span class="sourceLineNo">205</span> LOG.info("Current Open Regions After Master Node Starts Up:"<a name="line.205"></a>
<span class="sourceLineNo">206</span> + HBaseTestingUtil.getAllOnlineRegions(cluster).size());<a name="line.206"></a>
<span class="sourceLineNo">207</span><a name="line.207"></a>
<span class="sourceLineNo">208</span> assertEquals(numLogLines, TEST_UTIL.countRows(ht));<a name="line.208"></a>
<span class="sourceLineNo">209</span> }<a name="line.209"></a>
<span class="sourceLineNo">210</span> }<a name="line.210"></a>
<span class="sourceLineNo">211</span><a name="line.211"></a>
<span class="sourceLineNo">212</span> @Test<a name="line.212"></a>
<span class="sourceLineNo">213</span> public void testThreeRSAbort() throws Exception {<a name="line.213"></a>
<span class="sourceLineNo">214</span> LOG.info("testThreeRSAbort");<a name="line.214"></a>
<span class="sourceLineNo">215</span> int numRegionsToCreate = 40;<a name="line.215"></a>
<span class="sourceLineNo">216</span> int numRowsPerRegion = 100;<a name="line.216"></a>
<span class="sourceLineNo">217</span><a name="line.217"></a>
<span class="sourceLineNo">218</span> startCluster(NUM_RS); // NUM_RS=6.<a name="line.218"></a>
<span class="sourceLineNo">219</span><a name="line.219"></a>
<span class="sourceLineNo">220</span> try (Table table = installTable(numRegionsToCreate)) {<a name="line.220"></a>
<span class="sourceLineNo">221</span> populateDataInTable(numRowsPerRegion);<a name="line.221"></a>
<span class="sourceLineNo">222</span><a name="line.222"></a>
<span class="sourceLineNo">223</span> List&lt;RegionServerThread&gt; rsts = cluster.getLiveRegionServerThreads();<a name="line.223"></a>
<span class="sourceLineNo">224</span> assertEquals(NUM_RS, rsts.size());<a name="line.224"></a>
<span class="sourceLineNo">225</span> cluster.killRegionServer(rsts.get(0).getRegionServer().getServerName());<a name="line.225"></a>
<span class="sourceLineNo">226</span> cluster.killRegionServer(rsts.get(1).getRegionServer().getServerName());<a name="line.226"></a>
<span class="sourceLineNo">227</span> cluster.killRegionServer(rsts.get(2).getRegionServer().getServerName());<a name="line.227"></a>
<span class="sourceLineNo">228</span><a name="line.228"></a>
<span class="sourceLineNo">229</span> TEST_UTIL.waitFor(60000, new Waiter.ExplainingPredicate&lt;Exception&gt;() {<a name="line.229"></a>
<span class="sourceLineNo">230</span><a name="line.230"></a>
<span class="sourceLineNo">231</span> @Override<a name="line.231"></a>
<span class="sourceLineNo">232</span> public boolean evaluate() throws Exception {<a name="line.232"></a>
<span class="sourceLineNo">233</span> return cluster.getLiveRegionServerThreads().size() &lt;= NUM_RS - 3;<a name="line.233"></a>
<span class="sourceLineNo">234</span> }<a name="line.234"></a>
<span class="sourceLineNo">235</span><a name="line.235"></a>
<span class="sourceLineNo">236</span> @Override<a name="line.236"></a>
<span class="sourceLineNo">237</span> public String explainFailure() throws Exception {<a name="line.237"></a>
<span class="sourceLineNo">238</span> return "Timed out waiting for server aborts.";<a name="line.238"></a>
<span class="sourceLineNo">239</span> }<a name="line.239"></a>
<span class="sourceLineNo">240</span> });<a name="line.240"></a>
<span class="sourceLineNo">241</span> TEST_UTIL.waitUntilAllRegionsAssigned(tableName);<a name="line.241"></a>
<span class="sourceLineNo">242</span> int rows;<a name="line.242"></a>
<span class="sourceLineNo">243</span> try {<a name="line.243"></a>
<span class="sourceLineNo">244</span> rows = TEST_UTIL.countRows(table);<a name="line.244"></a>
<span class="sourceLineNo">245</span> } catch (Exception e) {<a name="line.245"></a>
<span class="sourceLineNo">246</span> Threads.printThreadInfo(System.out, "Thread dump before fail");<a name="line.246"></a>
<span class="sourceLineNo">247</span> throw e;<a name="line.247"></a>
<span class="sourceLineNo">248</span> }<a name="line.248"></a>
<span class="sourceLineNo">249</span> assertEquals(numRegionsToCreate * numRowsPerRegion, rows);<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><a name="line.252"></a>
<span class="sourceLineNo">253</span> @Test<a name="line.253"></a>
<span class="sourceLineNo">254</span> public void testDelayedDeleteOnFailure() throws Exception {<a name="line.254"></a>
<span class="sourceLineNo">255</span> if (<a name="line.255"></a>
<span class="sourceLineNo">256</span> !this.conf.getBoolean(HConstants.HBASE_SPLIT_WAL_COORDINATED_BY_ZK,<a name="line.256"></a>
<span class="sourceLineNo">257</span> HConstants.DEFAULT_HBASE_SPLIT_COORDINATED_BY_ZK)<a name="line.257"></a>
<span class="sourceLineNo">258</span> ) {<a name="line.258"></a>
<span class="sourceLineNo">259</span> // This test depends on zk coordination....<a name="line.259"></a>
<span class="sourceLineNo">260</span> return;<a name="line.260"></a>
<span class="sourceLineNo">261</span> }<a name="line.261"></a>
<span class="sourceLineNo">262</span> LOG.info("testDelayedDeleteOnFailure");<a name="line.262"></a>
<span class="sourceLineNo">263</span> startCluster(1);<a name="line.263"></a>
<span class="sourceLineNo">264</span> final SplitLogManager slm = master.getMasterWalManager().getSplitLogManager();<a name="line.264"></a>
<span class="sourceLineNo">265</span> final FileSystem fs = master.getMasterFileSystem().getFileSystem();<a name="line.265"></a>
<span class="sourceLineNo">266</span> final Path rootLogDir =<a name="line.266"></a>
<span class="sourceLineNo">267</span> new Path(CommonFSUtils.getWALRootDir(conf), HConstants.HREGION_LOGDIR_NAME);<a name="line.267"></a>
<span class="sourceLineNo">268</span> final Path logDir = new Path(rootLogDir, ServerName.valueOf("x", 1, 1).toString());<a name="line.268"></a>
<span class="sourceLineNo">269</span> fs.mkdirs(logDir);<a name="line.269"></a>
<span class="sourceLineNo">270</span> ExecutorService executor = null;<a name="line.270"></a>
<span class="sourceLineNo">271</span> try {<a name="line.271"></a>
<span class="sourceLineNo">272</span> final Path corruptedLogFile = new Path(logDir, "x");<a name="line.272"></a>
<span class="sourceLineNo">273</span> FSDataOutputStream out;<a name="line.273"></a>
<span class="sourceLineNo">274</span> out = fs.create(corruptedLogFile);<a name="line.274"></a>
<span class="sourceLineNo">275</span> out.write(0);<a name="line.275"></a>
<span class="sourceLineNo">276</span> out.write(Bytes.toBytes("corrupted bytes"));<a name="line.276"></a>
<span class="sourceLineNo">277</span> out.close();<a name="line.277"></a>
<span class="sourceLineNo">278</span> ZKSplitLogManagerCoordination coordination =<a name="line.278"></a>
<span class="sourceLineNo">279</span> (ZKSplitLogManagerCoordination) (master.getCoordinatedStateManager())<a name="line.279"></a>
<span class="sourceLineNo">280</span> .getSplitLogManagerCoordination();<a name="line.280"></a>
<span class="sourceLineNo">281</span> coordination.setIgnoreDeleteForTesting(true);<a name="line.281"></a>
<span class="sourceLineNo">282</span> executor = Executors.newSingleThreadExecutor();<a name="line.282"></a>
<span class="sourceLineNo">283</span> Runnable runnable = new Runnable() {<a name="line.283"></a>
<span class="sourceLineNo">284</span> @Override<a name="line.284"></a>
<span class="sourceLineNo">285</span> public void run() {<a name="line.285"></a>
<span class="sourceLineNo">286</span> try {<a name="line.286"></a>
<span class="sourceLineNo">287</span> // since the logDir is a fake, corrupted one, so the split log worker<a name="line.287"></a>
<span class="sourceLineNo">288</span> // will finish it quickly with error, and this call will fail and throw<a name="line.288"></a>
<span class="sourceLineNo">289</span> // an IOException.<a name="line.289"></a>
<span class="sourceLineNo">290</span> slm.splitLogDistributed(logDir);<a name="line.290"></a>
<span class="sourceLineNo">291</span> } catch (IOException ioe) {<a name="line.291"></a>
<span class="sourceLineNo">292</span> try {<a name="line.292"></a>
<span class="sourceLineNo">293</span> assertTrue(fs.exists(corruptedLogFile));<a name="line.293"></a>
<span class="sourceLineNo">294</span> // this call will block waiting for the task to be removed from the<a name="line.294"></a>
<span class="sourceLineNo">295</span> // tasks map which is not going to happen since ignoreZKDeleteForTesting<a name="line.295"></a>
<span class="sourceLineNo">296</span> // is set to true, until it is interrupted.<a name="line.296"></a>
<span class="sourceLineNo">297</span> slm.splitLogDistributed(logDir);<a name="line.297"></a>
<span class="sourceLineNo">298</span> } catch (IOException e) {<a name="line.298"></a>
<span class="sourceLineNo">299</span> assertTrue(Thread.currentThread().isInterrupted());<a name="line.299"></a>
<span class="sourceLineNo">300</span> return;<a name="line.300"></a>
<span class="sourceLineNo">301</span> }<a name="line.301"></a>
<span class="sourceLineNo">302</span> fail("did not get the expected IOException from the 2nd call");<a name="line.302"></a>
<span class="sourceLineNo">303</span> }<a name="line.303"></a>
<span class="sourceLineNo">304</span> fail("did not get the expected IOException from the 1st call");<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> Future&lt;?&gt; result = executor.submit(runnable);<a name="line.307"></a>
<span class="sourceLineNo">308</span> try {<a name="line.308"></a>
<span class="sourceLineNo">309</span> result.get(2000, TimeUnit.MILLISECONDS);<a name="line.309"></a>
<span class="sourceLineNo">310</span> } catch (TimeoutException te) {<a name="line.310"></a>
<span class="sourceLineNo">311</span> // it is ok, expected.<a name="line.311"></a>
<span class="sourceLineNo">312</span> }<a name="line.312"></a>
<span class="sourceLineNo">313</span> waitForCounter(tot_mgr_wait_for_zk_delete, 0, 1, 10000);<a name="line.313"></a>
<span class="sourceLineNo">314</span> executor.shutdownNow();<a name="line.314"></a>
<span class="sourceLineNo">315</span> executor = null;<a name="line.315"></a>
<span class="sourceLineNo">316</span><a name="line.316"></a>
<span class="sourceLineNo">317</span> // make sure the runnable is finished with no exception thrown.<a name="line.317"></a>
<span class="sourceLineNo">318</span> result.get();<a name="line.318"></a>
<span class="sourceLineNo">319</span> } finally {<a name="line.319"></a>
<span class="sourceLineNo">320</span> if (executor != null) {<a name="line.320"></a>
<span class="sourceLineNo">321</span> // interrupt the thread in case the test fails in the middle.<a name="line.321"></a>
<span class="sourceLineNo">322</span> // it has no effect if the thread is already terminated.<a name="line.322"></a>
<span class="sourceLineNo">323</span> executor.shutdownNow();<a name="line.323"></a>
<span class="sourceLineNo">324</span> }<a name="line.324"></a>
<span class="sourceLineNo">325</span> fs.delete(logDir, true);<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> private Table installTable(int nrs) throws Exception {<a name="line.329"></a>
<span class="sourceLineNo">330</span> return installTable(nrs, 0);<a name="line.330"></a>
<span class="sourceLineNo">331</span> }<a name="line.331"></a>
<span class="sourceLineNo">332</span><a name="line.332"></a>
<span class="sourceLineNo">333</span> private Table installTable(int nrs, int existingRegions) throws Exception {<a name="line.333"></a>
<span class="sourceLineNo">334</span> // Create a table with regions<a name="line.334"></a>
<span class="sourceLineNo">335</span> byte[] family = Bytes.toBytes("family");<a name="line.335"></a>
<span class="sourceLineNo">336</span> LOG.info("Creating table with " + nrs + " regions");<a name="line.336"></a>
<span class="sourceLineNo">337</span> Table table = TEST_UTIL.createMultiRegionTable(tableName, family, nrs);<a name="line.337"></a>
<span class="sourceLineNo">338</span> int numRegions = -1;<a name="line.338"></a>
<span class="sourceLineNo">339</span> try (RegionLocator r = TEST_UTIL.getConnection().getRegionLocator(tableName)) {<a name="line.339"></a>
<span class="sourceLineNo">340</span> numRegions = r.getStartKeys().length;<a name="line.340"></a>
<span class="sourceLineNo">341</span> }<a name="line.341"></a>
<span class="sourceLineNo">342</span> assertEquals(nrs, numRegions);<a name="line.342"></a>
<span class="sourceLineNo">343</span> LOG.info("Waiting for no more RIT\n");<a name="line.343"></a>
<span class="sourceLineNo">344</span> blockUntilNoRIT();<a name="line.344"></a>
<span class="sourceLineNo">345</span> // disable-enable cycle to get rid of table's dead regions left behind<a name="line.345"></a>
<span class="sourceLineNo">346</span> // by createMultiRegions<a name="line.346"></a>
<span class="sourceLineNo">347</span> assertTrue(TEST_UTIL.getAdmin().isTableEnabled(tableName));<a name="line.347"></a>
<span class="sourceLineNo">348</span> LOG.debug("Disabling table\n");<a name="line.348"></a>
<span class="sourceLineNo">349</span> TEST_UTIL.getAdmin().disableTable(tableName);<a name="line.349"></a>
<span class="sourceLineNo">350</span> LOG.debug("Waiting for no more RIT\n");<a name="line.350"></a>
<span class="sourceLineNo">351</span> blockUntilNoRIT();<a name="line.351"></a>
<span class="sourceLineNo">352</span> NavigableSet&lt;String&gt; regions = HBaseTestingUtil.getAllOnlineRegions(cluster);<a name="line.352"></a>
<span class="sourceLineNo">353</span> LOG.debug("Verifying only catalog region is assigned\n");<a name="line.353"></a>
<span class="sourceLineNo">354</span> if (regions.size() != 1) {<a name="line.354"></a>
<span class="sourceLineNo">355</span> for (String oregion : regions) {<a name="line.355"></a>
<span class="sourceLineNo">356</span> LOG.debug("Region still online: " + oregion);<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> assertEquals(1 + existingRegions, regions.size());<a name="line.359"></a>
<span class="sourceLineNo">360</span> LOG.debug("Enabling table\n");<a name="line.360"></a>
<span class="sourceLineNo">361</span> TEST_UTIL.getAdmin().enableTable(tableName);<a name="line.361"></a>
<span class="sourceLineNo">362</span> LOG.debug("Waiting for no more RIT\n");<a name="line.362"></a>
<span class="sourceLineNo">363</span> blockUntilNoRIT();<a name="line.363"></a>
<span class="sourceLineNo">364</span> LOG.debug("Verifying there are " + numRegions + " assigned on cluster\n");<a name="line.364"></a>
<span class="sourceLineNo">365</span> regions = HBaseTestingUtil.getAllOnlineRegions(cluster);<a name="line.365"></a>
<span class="sourceLineNo">366</span> assertEquals(numRegions + 1 + existingRegions, regions.size());<a name="line.366"></a>
<span class="sourceLineNo">367</span> return table;<a name="line.367"></a>
<span class="sourceLineNo">368</span> }<a name="line.368"></a>
<span class="sourceLineNo">369</span><a name="line.369"></a>
<span class="sourceLineNo">370</span> void populateDataInTable(int nrows) throws Exception {<a name="line.370"></a>
<span class="sourceLineNo">371</span> List&lt;RegionServerThread&gt; rsts = cluster.getLiveRegionServerThreads();<a name="line.371"></a>
<span class="sourceLineNo">372</span> assertEquals(NUM_RS, rsts.size());<a name="line.372"></a>
<span class="sourceLineNo">373</span><a name="line.373"></a>
<span class="sourceLineNo">374</span> for (RegionServerThread rst : rsts) {<a name="line.374"></a>
<span class="sourceLineNo">375</span> HRegionServer hrs = rst.getRegionServer();<a name="line.375"></a>
<span class="sourceLineNo">376</span> List&lt;RegionInfo&gt; hris = ProtobufUtil.getOnlineRegions(hrs.getRSRpcServices());<a name="line.376"></a>
<span class="sourceLineNo">377</span> for (RegionInfo hri : hris) {<a name="line.377"></a>
<span class="sourceLineNo">378</span> if (hri.getTable().isSystemTable()) {<a name="line.378"></a>
<span class="sourceLineNo">379</span> continue;<a name="line.379"></a>
<span class="sourceLineNo">380</span> }<a name="line.380"></a>
<span class="sourceLineNo">381</span> LOG.debug(<a name="line.381"></a>
<span class="sourceLineNo">382</span> "adding data to rs = " + rst.getName() + " region = " + hri.getRegionNameAsString());<a name="line.382"></a>
<span class="sourceLineNo">383</span> Region region = hrs.getOnlineRegion(hri.getRegionName());<a name="line.383"></a>
<span class="sourceLineNo">384</span> assertTrue(region != null);<a name="line.384"></a>
<span class="sourceLineNo">385</span> putData(region, hri.getStartKey(), nrows, Bytes.toBytes("q"), COLUMN_FAMILY);<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 makeWAL(HRegionServer hrs, List&lt;RegionInfo&gt; regions, int num_edits, int edit_size)<a name="line.390"></a>
<span class="sourceLineNo">391</span> throws IOException {<a name="line.391"></a>
<span class="sourceLineNo">392</span> makeWAL(hrs, regions, num_edits, edit_size, true);<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> public void makeWAL(HRegionServer hrs, List&lt;RegionInfo&gt; regions, int numEdits, int editSize,<a name="line.395"></a>
<span class="sourceLineNo">396</span> boolean cleanShutdown) throws IOException {<a name="line.396"></a>
<span class="sourceLineNo">397</span> // remove root and meta region<a name="line.397"></a>
<span class="sourceLineNo">398</span> regions.remove(RegionInfoBuilder.FIRST_META_REGIONINFO);<a name="line.398"></a>
<span class="sourceLineNo">399</span><a name="line.399"></a>
<span class="sourceLineNo">400</span> for (Iterator&lt;RegionInfo&gt; iter = regions.iterator(); iter.hasNext();) {<a name="line.400"></a>
<span class="sourceLineNo">401</span> RegionInfo regionInfo = iter.next();<a name="line.401"></a>
<span class="sourceLineNo">402</span> if (regionInfo.getTable().isSystemTable()) {<a name="line.402"></a>
<span class="sourceLineNo">403</span> iter.remove();<a name="line.403"></a>
<span class="sourceLineNo">404</span> }<a name="line.404"></a>
<span class="sourceLineNo">405</span> }<a name="line.405"></a>
<span class="sourceLineNo">406</span> byte[] value = new byte[editSize];<a name="line.406"></a>
<span class="sourceLineNo">407</span><a name="line.407"></a>
<span class="sourceLineNo">408</span> List&lt;RegionInfo&gt; hris = new ArrayList&lt;&gt;();<a name="line.408"></a>
<span class="sourceLineNo">409</span> for (RegionInfo region : regions) {<a name="line.409"></a>
<span class="sourceLineNo">410</span> if (region.getTable() != tableName) {<a name="line.410"></a>
<span class="sourceLineNo">411</span> continue;<a name="line.411"></a>
<span class="sourceLineNo">412</span> }<a name="line.412"></a>
<span class="sourceLineNo">413</span> hris.add(region);<a name="line.413"></a>
<span class="sourceLineNo">414</span> }<a name="line.414"></a>
<span class="sourceLineNo">415</span> LOG.info("Creating wal edits across " + hris.size() + " regions.");<a name="line.415"></a>
<span class="sourceLineNo">416</span> for (int i = 0; i &lt; editSize; i++) {<a name="line.416"></a>
<span class="sourceLineNo">417</span> value[i] = (byte) ('a' + (i % 26));<a name="line.417"></a>
<span class="sourceLineNo">418</span> }<a name="line.418"></a>
<span class="sourceLineNo">419</span> int n = hris.size();<a name="line.419"></a>
<span class="sourceLineNo">420</span> int[] counts = new int[n];<a name="line.420"></a>
<span class="sourceLineNo">421</span> // sync every ~30k to line up with desired wal rolls<a name="line.421"></a>
<span class="sourceLineNo">422</span> final int syncEvery = 30 * 1024 / editSize;<a name="line.422"></a>
<span class="sourceLineNo">423</span> MultiVersionConcurrencyControl mvcc = new MultiVersionConcurrencyControl();<a name="line.423"></a>
<span class="sourceLineNo">424</span> if (n &gt; 0) {<a name="line.424"></a>
<span class="sourceLineNo">425</span> for (int i = 0; i &lt; numEdits; i += 1) {<a name="line.425"></a>
<span class="sourceLineNo">426</span> WALEdit e = new WALEdit();<a name="line.426"></a>
<span class="sourceLineNo">427</span> RegionInfo curRegionInfo = hris.get(i % n);<a name="line.427"></a>
<span class="sourceLineNo">428</span> WAL log = hrs.getWAL(curRegionInfo);<a name="line.428"></a>
<span class="sourceLineNo">429</span> byte[] startRow = curRegionInfo.getStartKey();<a name="line.429"></a>
<span class="sourceLineNo">430</span> if (startRow == null || startRow.length == 0) {<a name="line.430"></a>
<span class="sourceLineNo">431</span> startRow = new byte[] { 0, 0, 0, 0, 1 };<a name="line.431"></a>
<span class="sourceLineNo">432</span> }<a name="line.432"></a>
<span class="sourceLineNo">433</span> byte[] row = Bytes.incrementBytes(startRow, counts[i % n]);<a name="line.433"></a>
<span class="sourceLineNo">434</span> row = Arrays.copyOfRange(row, 3, 8); // use last 5 bytes because<a name="line.434"></a>
<span class="sourceLineNo">435</span> // HBaseTestingUtility.createMultiRegions use 5 bytes key<a name="line.435"></a>
<span class="sourceLineNo">436</span> byte[] qualifier = Bytes.toBytes("c" + Integer.toString(i));<a name="line.436"></a>
<span class="sourceLineNo">437</span> e.add(<a name="line.437"></a>
<span class="sourceLineNo">438</span> new KeyValue(row, COLUMN_FAMILY, qualifier, EnvironmentEdgeManager.currentTime(), value));<a name="line.438"></a>
<span class="sourceLineNo">439</span> log.appendData(curRegionInfo, new WALKeyImpl(curRegionInfo.getEncodedNameAsBytes(),<a name="line.439"></a>
<span class="sourceLineNo">440</span> tableName, EnvironmentEdgeManager.currentTime(), mvcc), e);<a name="line.440"></a>
<span class="sourceLineNo">441</span> if (0 == i % syncEvery) {<a name="line.441"></a>
<span class="sourceLineNo">442</span> log.sync();<a name="line.442"></a>
<span class="sourceLineNo">443</span> }<a name="line.443"></a>
<span class="sourceLineNo">444</span> counts[i % n] += 1;<a name="line.444"></a>
<span class="sourceLineNo">445</span> }<a name="line.445"></a>
<span class="sourceLineNo">446</span> }<a name="line.446"></a>
<span class="sourceLineNo">447</span> // done as two passes because the regions might share logs. shutdown is idempotent, but sync<a name="line.447"></a>
<span class="sourceLineNo">448</span> // will cause errors if done after.<a name="line.448"></a>
<span class="sourceLineNo">449</span> for (RegionInfo info : hris) {<a name="line.449"></a>
<span class="sourceLineNo">450</span> WAL log = hrs.getWAL(info);<a name="line.450"></a>
<span class="sourceLineNo">451</span> log.sync();<a name="line.451"></a>
<span class="sourceLineNo">452</span> }<a name="line.452"></a>
<span class="sourceLineNo">453</span> if (cleanShutdown) {<a name="line.453"></a>
<span class="sourceLineNo">454</span> for (RegionInfo info : hris) {<a name="line.454"></a>
<span class="sourceLineNo">455</span> WAL log = hrs.getWAL(info);<a name="line.455"></a>
<span class="sourceLineNo">456</span> log.shutdown();<a name="line.456"></a>
<span class="sourceLineNo">457</span> }<a name="line.457"></a>
<span class="sourceLineNo">458</span> }<a name="line.458"></a>
<span class="sourceLineNo">459</span> for (int i = 0; i &lt; n; i++) {<a name="line.459"></a>
<span class="sourceLineNo">460</span> LOG.info("region " + hris.get(i).getRegionNameAsString() + " has " + counts[i] + " edits");<a name="line.460"></a>
<span class="sourceLineNo">461</span> }<a name="line.461"></a>
<span class="sourceLineNo">462</span> return;<a name="line.462"></a>
<span class="sourceLineNo">463</span> }<a name="line.463"></a>
<span class="sourceLineNo">464</span><a name="line.464"></a>
<span class="sourceLineNo">465</span> private void blockUntilNoRIT() throws Exception {<a name="line.465"></a>
<span class="sourceLineNo">466</span> TEST_UTIL.waitUntilNoRegionsInTransition(60000);<a name="line.466"></a>
<span class="sourceLineNo">467</span> }<a name="line.467"></a>
<span class="sourceLineNo">468</span><a name="line.468"></a>
<span class="sourceLineNo">469</span> private void putData(Region region, byte[] startRow, int numRows, byte[] qf, byte[]... families)<a name="line.469"></a>
<span class="sourceLineNo">470</span> throws IOException {<a name="line.470"></a>
<span class="sourceLineNo">471</span> for (int i = 0; i &lt; numRows; i++) {<a name="line.471"></a>
<span class="sourceLineNo">472</span> Put put = new Put(Bytes.add(startRow, Bytes.toBytes(i)));<a name="line.472"></a>
<span class="sourceLineNo">473</span> for (byte[] family : families) {<a name="line.473"></a>
<span class="sourceLineNo">474</span> put.addColumn(family, qf, null);<a name="line.474"></a>
<span class="sourceLineNo">475</span> }<a name="line.475"></a>
<span class="sourceLineNo">476</span> region.put(put);<a name="line.476"></a>
<span class="sourceLineNo">477</span> }<a name="line.477"></a>
<span class="sourceLineNo">478</span> }<a name="line.478"></a>
<span class="sourceLineNo">479</span><a name="line.479"></a>
<span class="sourceLineNo">480</span> private void waitForCounter(LongAdder ctr, long oldval, long newval, long timems)<a name="line.480"></a>
<span class="sourceLineNo">481</span> throws InterruptedException {<a name="line.481"></a>
<span class="sourceLineNo">482</span> long curt = EnvironmentEdgeManager.currentTime();<a name="line.482"></a>
<span class="sourceLineNo">483</span> long endt = curt + timems;<a name="line.483"></a>
<span class="sourceLineNo">484</span> while (curt &lt; endt) {<a name="line.484"></a>
<span class="sourceLineNo">485</span> if (ctr.sum() == oldval) {<a name="line.485"></a>
<span class="sourceLineNo">486</span> Thread.sleep(100);<a name="line.486"></a>
<span class="sourceLineNo">487</span> curt = EnvironmentEdgeManager.currentTime();<a name="line.487"></a>
<span class="sourceLineNo">488</span> } else {<a name="line.488"></a>
<span class="sourceLineNo">489</span> assertEquals(newval, ctr.sum());<a name="line.489"></a>
<span class="sourceLineNo">490</span> return;<a name="line.490"></a>
<span class="sourceLineNo">491</span> }<a name="line.491"></a>
<span class="sourceLineNo">492</span> }<a name="line.492"></a>
<span class="sourceLineNo">493</span> fail();<a name="line.493"></a>
<span class="sourceLineNo">494</span> }<a name="line.494"></a>
<span class="sourceLineNo">495</span><a name="line.495"></a>
<span class="sourceLineNo">496</span> private void abortMaster(SingleProcessHBaseCluster cluster) throws InterruptedException {<a name="line.496"></a>
<span class="sourceLineNo">497</span> for (MasterThread mt : cluster.getLiveMasterThreads()) {<a name="line.497"></a>
<span class="sourceLineNo">498</span> if (mt.getMaster().isActiveMaster()) {<a name="line.498"></a>
<span class="sourceLineNo">499</span> mt.getMaster().abort("Aborting for tests", new Exception("Trace info"));<a name="line.499"></a>
<span class="sourceLineNo">500</span> mt.join();<a name="line.500"></a>
<span class="sourceLineNo">501</span> break;<a name="line.501"></a>
<span class="sourceLineNo">502</span> }<a name="line.502"></a>
<span class="sourceLineNo">503</span> }<a name="line.503"></a>
<span class="sourceLineNo">504</span> LOG.debug("Master is aborted");<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> /**<a name="line.507"></a>
<span class="sourceLineNo">508</span> * Find a RS that has regions of a table.<a name="line.508"></a>
<span class="sourceLineNo">509</span> * @param hasMetaRegion when true, the returned RS has hbase:meta region as well<a name="line.509"></a>
<span class="sourceLineNo">510</span> */<a name="line.510"></a>
<span class="sourceLineNo">511</span> private HRegionServer findRSToKill(boolean hasMetaRegion) throws Exception {<a name="line.511"></a>
<span class="sourceLineNo">512</span> List&lt;RegionServerThread&gt; rsts = cluster.getLiveRegionServerThreads();<a name="line.512"></a>
<span class="sourceLineNo">513</span> List&lt;RegionInfo&gt; regions = null;<a name="line.513"></a>
<span class="sourceLineNo">514</span> HRegionServer hrs = null;<a name="line.514"></a>
<span class="sourceLineNo">515</span><a name="line.515"></a>
<span class="sourceLineNo">516</span> for (RegionServerThread rst : rsts) {<a name="line.516"></a>
<span class="sourceLineNo">517</span> hrs = rst.getRegionServer();<a name="line.517"></a>
<span class="sourceLineNo">518</span> while (rst.isAlive() &amp;&amp; !hrs.isOnline()) {<a name="line.518"></a>
<span class="sourceLineNo">519</span> Thread.sleep(100);<a name="line.519"></a>
<span class="sourceLineNo">520</span> }<a name="line.520"></a>
<span class="sourceLineNo">521</span> if (!rst.isAlive()) {<a name="line.521"></a>
<span class="sourceLineNo">522</span> continue;<a name="line.522"></a>
<span class="sourceLineNo">523</span> }<a name="line.523"></a>
<span class="sourceLineNo">524</span> boolean isCarryingMeta = false;<a name="line.524"></a>
<span class="sourceLineNo">525</span> boolean foundTableRegion = false;<a name="line.525"></a>
<span class="sourceLineNo">526</span> regions = ProtobufUtil.getOnlineRegions(hrs.getRSRpcServices());<a name="line.526"></a>
<span class="sourceLineNo">527</span> for (RegionInfo region : regions) {<a name="line.527"></a>
<span class="sourceLineNo">528</span> if (region.isMetaRegion()) {<a name="line.528"></a>
<span class="sourceLineNo">529</span> isCarryingMeta = true;<a name="line.529"></a>
<span class="sourceLineNo">530</span> }<a name="line.530"></a>
<span class="sourceLineNo">531</span> if (region.getTable() == tableName) {<a name="line.531"></a>
<span class="sourceLineNo">532</span> foundTableRegion = true;<a name="line.532"></a>
<span class="sourceLineNo">533</span> }<a name="line.533"></a>
<span class="sourceLineNo">534</span> if (foundTableRegion &amp;&amp; (isCarryingMeta || !hasMetaRegion)) {<a name="line.534"></a>
<span class="sourceLineNo">535</span> break;<a name="line.535"></a>
<span class="sourceLineNo">536</span> }<a name="line.536"></a>
<span class="sourceLineNo">537</span> }<a name="line.537"></a>
<span class="sourceLineNo">538</span> if (isCarryingMeta &amp;&amp; hasMetaRegion) {<a name="line.538"></a>
<span class="sourceLineNo">539</span> // clients ask for a RS with META<a name="line.539"></a>
<span class="sourceLineNo">540</span> if (!foundTableRegion) {<a name="line.540"></a>
<span class="sourceLineNo">541</span> HRegionServer destRS = hrs;<a name="line.541"></a>
<span class="sourceLineNo">542</span> // the RS doesn't have regions of the specified table so we need move one to this RS<a name="line.542"></a>
<span class="sourceLineNo">543</span> List&lt;RegionInfo&gt; tableRegions = TEST_UTIL.getAdmin().getRegions(tableName);<a name="line.543"></a>
<span class="sourceLineNo">544</span> RegionInfo hri = tableRegions.get(0);<a name="line.544"></a>
<span class="sourceLineNo">545</span> TEST_UTIL.getAdmin().move(hri.getEncodedNameAsBytes(), destRS.getServerName());<a name="line.545"></a>
<span class="sourceLineNo">546</span> // wait for region move completes<a name="line.546"></a>
<span class="sourceLineNo">547</span> RegionStates regionStates =<a name="line.547"></a>
<span class="sourceLineNo">548</span> TEST_UTIL.getHBaseCluster().getMaster().getAssignmentManager().getRegionStates();<a name="line.548"></a>
<span class="sourceLineNo">549</span> TEST_UTIL.waitFor(45000, 200, new Waiter.Predicate&lt;Exception&gt;() {<a name="line.549"></a>
<span class="sourceLineNo">550</span> @Override<a name="line.550"></a>
<span class="sourceLineNo">551</span> public boolean evaluate() throws Exception {<a name="line.551"></a>
<span class="sourceLineNo">552</span> ServerName sn = regionStates.getRegionServerOfRegion(hri);<a name="line.552"></a>
<span class="sourceLineNo">553</span> return (sn != null &amp;&amp; sn.equals(destRS.getServerName()));<a name="line.553"></a>
<span class="sourceLineNo">554</span> }<a name="line.554"></a>
<span class="sourceLineNo">555</span> });<a name="line.555"></a>
<span class="sourceLineNo">556</span> }<a name="line.556"></a>
<span class="sourceLineNo">557</span> return hrs;<a name="line.557"></a>
<span class="sourceLineNo">558</span> } else if (hasMetaRegion || isCarryingMeta) {<a name="line.558"></a>
<span class="sourceLineNo">559</span> continue;<a name="line.559"></a>
<span class="sourceLineNo">560</span> }<a name="line.560"></a>
<span class="sourceLineNo">561</span> if (foundTableRegion) {<a name="line.561"></a>
<span class="sourceLineNo">562</span> break;<a name="line.562"></a>
<span class="sourceLineNo">563</span> }<a name="line.563"></a>
<span class="sourceLineNo">564</span> }<a name="line.564"></a>
<span class="sourceLineNo">565</span><a name="line.565"></a>
<span class="sourceLineNo">566</span> return hrs;<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>
</pre>
</div>
</body>
</html>