blob: 1a1cf4be09f7febdab45e9aff10a55b677212242 [file] [log] [blame]
<!DOCTYPE HTML>
<html lang="en">
<head>
<!-- Generated by javadoc (17) -->
<title>Source code</title>
<meta name="viewport" content="width=device-width, initial-scale=1">
<meta name="description" content="source: package: org.apache.hadoop.hbase.master, class: TestMasterFailover">
<meta name="generator" content="javadoc/SourceToHTMLConverter">
<link rel="stylesheet" type="text/css" href="../../../../../../stylesheet.css" title="Style">
</head>
<body class="source-page">
<main role="main">
<div class="source-container">
<pre><span class="source-line-no">001</span><span id="line-1">/*</span>
<span class="source-line-no">002</span><span id="line-2"> * Licensed to the Apache Software Foundation (ASF) under one</span>
<span class="source-line-no">003</span><span id="line-3"> * or more contributor license agreements. See the NOTICE file</span>
<span class="source-line-no">004</span><span id="line-4"> * distributed with this work for additional information</span>
<span class="source-line-no">005</span><span id="line-5"> * regarding copyright ownership. The ASF licenses this file</span>
<span class="source-line-no">006</span><span id="line-6"> * to you under the Apache License, Version 2.0 (the</span>
<span class="source-line-no">007</span><span id="line-7"> * "License"); you may not use this file except in compliance</span>
<span class="source-line-no">008</span><span id="line-8"> * with the License. You may obtain a copy of the License at</span>
<span class="source-line-no">009</span><span id="line-9"> *</span>
<span class="source-line-no">010</span><span id="line-10"> * http://www.apache.org/licenses/LICENSE-2.0</span>
<span class="source-line-no">011</span><span id="line-11"> *</span>
<span class="source-line-no">012</span><span id="line-12"> * Unless required by applicable law or agreed to in writing, software</span>
<span class="source-line-no">013</span><span id="line-13"> * distributed under the License is distributed on an "AS IS" BASIS,</span>
<span class="source-line-no">014</span><span id="line-14"> * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.</span>
<span class="source-line-no">015</span><span id="line-15"> * See the License for the specific language governing permissions and</span>
<span class="source-line-no">016</span><span id="line-16"> * limitations under the License.</span>
<span class="source-line-no">017</span><span id="line-17"> */</span>
<span class="source-line-no">018</span><span id="line-18">package org.apache.hadoop.hbase.master;</span>
<span class="source-line-no">019</span><span id="line-19"></span>
<span class="source-line-no">020</span><span id="line-20">import static org.junit.Assert.assertEquals;</span>
<span class="source-line-no">021</span><span id="line-21">import static org.junit.Assert.assertNotNull;</span>
<span class="source-line-no">022</span><span id="line-22">import static org.junit.Assert.assertTrue;</span>
<span class="source-line-no">023</span><span id="line-23"></span>
<span class="source-line-no">024</span><span id="line-24">import java.util.List;</span>
<span class="source-line-no">025</span><span id="line-25">import java.util.concurrent.TimeUnit;</span>
<span class="source-line-no">026</span><span id="line-26">import org.apache.hadoop.hbase.ClusterMetrics;</span>
<span class="source-line-no">027</span><span id="line-27">import org.apache.hadoop.hbase.HBaseClassTestRule;</span>
<span class="source-line-no">028</span><span id="line-28">import org.apache.hadoop.hbase.HBaseTestingUtil;</span>
<span class="source-line-no">029</span><span id="line-29">import org.apache.hadoop.hbase.ServerName;</span>
<span class="source-line-no">030</span><span id="line-30">import org.apache.hadoop.hbase.SingleProcessHBaseCluster;</span>
<span class="source-line-no">031</span><span id="line-31">import org.apache.hadoop.hbase.StartTestingClusterOption;</span>
<span class="source-line-no">032</span><span id="line-32">import org.apache.hadoop.hbase.master.RegionState.State;</span>
<span class="source-line-no">033</span><span id="line-33">import org.apache.hadoop.hbase.regionserver.HRegionServer;</span>
<span class="source-line-no">034</span><span id="line-34">import org.apache.hadoop.hbase.testclassification.FlakeyTests;</span>
<span class="source-line-no">035</span><span id="line-35">import org.apache.hadoop.hbase.testclassification.LargeTests;</span>
<span class="source-line-no">036</span><span id="line-36">import org.apache.hadoop.hbase.util.JVMClusterUtil.MasterThread;</span>
<span class="source-line-no">037</span><span id="line-37">import org.apache.hadoop.hbase.zookeeper.MetaTableLocator;</span>
<span class="source-line-no">038</span><span id="line-38">import org.junit.ClassRule;</span>
<span class="source-line-no">039</span><span id="line-39">import org.junit.Rule;</span>
<span class="source-line-no">040</span><span id="line-40">import org.junit.Test;</span>
<span class="source-line-no">041</span><span id="line-41">import org.junit.experimental.categories.Category;</span>
<span class="source-line-no">042</span><span id="line-42">import org.junit.rules.TestName;</span>
<span class="source-line-no">043</span><span id="line-43">import org.slf4j.Logger;</span>
<span class="source-line-no">044</span><span id="line-44">import org.slf4j.LoggerFactory;</span>
<span class="source-line-no">045</span><span id="line-45"></span>
<span class="source-line-no">046</span><span id="line-46">@Category({ FlakeyTests.class, LargeTests.class })</span>
<span class="source-line-no">047</span><span id="line-47">public class TestMasterFailover {</span>
<span class="source-line-no">048</span><span id="line-48"></span>
<span class="source-line-no">049</span><span id="line-49"> @ClassRule</span>
<span class="source-line-no">050</span><span id="line-50"> public static final HBaseClassTestRule CLASS_RULE =</span>
<span class="source-line-no">051</span><span id="line-51"> HBaseClassTestRule.forClass(TestMasterFailover.class);</span>
<span class="source-line-no">052</span><span id="line-52"></span>
<span class="source-line-no">053</span><span id="line-53"> private static final Logger LOG = LoggerFactory.getLogger(TestMasterFailover.class);</span>
<span class="source-line-no">054</span><span id="line-54"> @Rule</span>
<span class="source-line-no">055</span><span id="line-55"> public TestName name = new TestName();</span>
<span class="source-line-no">056</span><span id="line-56"></span>
<span class="source-line-no">057</span><span id="line-57"> /**</span>
<span class="source-line-no">058</span><span id="line-58"> * Simple test of master failover.</span>
<span class="source-line-no">059</span><span id="line-59"> * &lt;p&gt;</span>
<span class="source-line-no">060</span><span id="line-60"> * Starts with three masters. Kills a backup master. Then kills the active master. Ensures the</span>
<span class="source-line-no">061</span><span id="line-61"> * final master becomes active and we can still contact the cluster.</span>
<span class="source-line-no">062</span><span id="line-62"> */</span>
<span class="source-line-no">063</span><span id="line-63"> @Test</span>
<span class="source-line-no">064</span><span id="line-64"> public void testSimpleMasterFailover() throws Exception {</span>
<span class="source-line-no">065</span><span id="line-65"> final int NUM_MASTERS = 3;</span>
<span class="source-line-no">066</span><span id="line-66"> final int NUM_RS = 3;</span>
<span class="source-line-no">067</span><span id="line-67"></span>
<span class="source-line-no">068</span><span id="line-68"> // Start the cluster</span>
<span class="source-line-no">069</span><span id="line-69"> HBaseTestingUtil TEST_UTIL = new HBaseTestingUtil();</span>
<span class="source-line-no">070</span><span id="line-70"> try {</span>
<span class="source-line-no">071</span><span id="line-71"> StartTestingClusterOption option = StartTestingClusterOption.builder().numMasters(NUM_MASTERS)</span>
<span class="source-line-no">072</span><span id="line-72"> .numRegionServers(NUM_RS).numDataNodes(NUM_RS).build();</span>
<span class="source-line-no">073</span><span id="line-73"> TEST_UTIL.startMiniCluster(option);</span>
<span class="source-line-no">074</span><span id="line-74"> SingleProcessHBaseCluster cluster = TEST_UTIL.getHBaseCluster();</span>
<span class="source-line-no">075</span><span id="line-75"></span>
<span class="source-line-no">076</span><span id="line-76"> // get all the master threads</span>
<span class="source-line-no">077</span><span id="line-77"> List&lt;MasterThread&gt; masterThreads = cluster.getMasterThreads();</span>
<span class="source-line-no">078</span><span id="line-78"></span>
<span class="source-line-no">079</span><span id="line-79"> // wait for each to come online</span>
<span class="source-line-no">080</span><span id="line-80"> for (MasterThread mt : masterThreads) {</span>
<span class="source-line-no">081</span><span id="line-81"> assertTrue(mt.isAlive());</span>
<span class="source-line-no">082</span><span id="line-82"> }</span>
<span class="source-line-no">083</span><span id="line-83"></span>
<span class="source-line-no">084</span><span id="line-84"> // verify only one is the active master and we have right number</span>
<span class="source-line-no">085</span><span id="line-85"> int numActive = 0;</span>
<span class="source-line-no">086</span><span id="line-86"> int activeIndex = -1;</span>
<span class="source-line-no">087</span><span id="line-87"> ServerName activeName = null;</span>
<span class="source-line-no">088</span><span id="line-88"> HMaster active = null;</span>
<span class="source-line-no">089</span><span id="line-89"> for (int i = 0; i &lt; masterThreads.size(); i++) {</span>
<span class="source-line-no">090</span><span id="line-90"> if (masterThreads.get(i).getMaster().isActiveMaster()) {</span>
<span class="source-line-no">091</span><span id="line-91"> numActive++;</span>
<span class="source-line-no">092</span><span id="line-92"> activeIndex = i;</span>
<span class="source-line-no">093</span><span id="line-93"> active = masterThreads.get(activeIndex).getMaster();</span>
<span class="source-line-no">094</span><span id="line-94"> activeName = active.getServerName();</span>
<span class="source-line-no">095</span><span id="line-95"> }</span>
<span class="source-line-no">096</span><span id="line-96"> }</span>
<span class="source-line-no">097</span><span id="line-97"> assertEquals(1, numActive);</span>
<span class="source-line-no">098</span><span id="line-98"> assertEquals(NUM_MASTERS, masterThreads.size());</span>
<span class="source-line-no">099</span><span id="line-99"> LOG.info("Active master " + activeName);</span>
<span class="source-line-no">100</span><span id="line-100"></span>
<span class="source-line-no">101</span><span id="line-101"> // Check that ClusterStatus reports the correct active and backup masters</span>
<span class="source-line-no">102</span><span id="line-102"> assertNotNull(active);</span>
<span class="source-line-no">103</span><span id="line-103"> ClusterMetrics status = active.getClusterMetrics();</span>
<span class="source-line-no">104</span><span id="line-104"> assertEquals(activeName, status.getMasterName());</span>
<span class="source-line-no">105</span><span id="line-105"> assertEquals(2, status.getBackupMasterNames().size());</span>
<span class="source-line-no">106</span><span id="line-106"></span>
<span class="source-line-no">107</span><span id="line-107"> // attempt to stop one of the inactive masters</span>
<span class="source-line-no">108</span><span id="line-108"> int backupIndex = (activeIndex == 0 ? 1 : activeIndex - 1);</span>
<span class="source-line-no">109</span><span id="line-109"> HMaster master = cluster.getMaster(backupIndex);</span>
<span class="source-line-no">110</span><span id="line-110"> LOG.debug("\n\nStopping a backup master: " + master.getServerName() + "\n");</span>
<span class="source-line-no">111</span><span id="line-111"> cluster.stopMaster(backupIndex, false);</span>
<span class="source-line-no">112</span><span id="line-112"> cluster.waitOnMaster(backupIndex);</span>
<span class="source-line-no">113</span><span id="line-113"></span>
<span class="source-line-no">114</span><span id="line-114"> // Verify still one active master and it's the same</span>
<span class="source-line-no">115</span><span id="line-115"> for (int i = 0; i &lt; masterThreads.size(); i++) {</span>
<span class="source-line-no">116</span><span id="line-116"> if (masterThreads.get(i).getMaster().isActiveMaster()) {</span>
<span class="source-line-no">117</span><span id="line-117"> assertEquals(activeName, masterThreads.get(i).getMaster().getServerName());</span>
<span class="source-line-no">118</span><span id="line-118"> activeIndex = i;</span>
<span class="source-line-no">119</span><span id="line-119"> active = masterThreads.get(activeIndex).getMaster();</span>
<span class="source-line-no">120</span><span id="line-120"> }</span>
<span class="source-line-no">121</span><span id="line-121"> }</span>
<span class="source-line-no">122</span><span id="line-122"> assertEquals(1, numActive);</span>
<span class="source-line-no">123</span><span id="line-123"> assertEquals(2, masterThreads.size());</span>
<span class="source-line-no">124</span><span id="line-124"> int rsCount = masterThreads.get(activeIndex).getMaster().getClusterMetrics()</span>
<span class="source-line-no">125</span><span id="line-125"> .getLiveServerMetrics().size();</span>
<span class="source-line-no">126</span><span id="line-126"> LOG.info(</span>
<span class="source-line-no">127</span><span id="line-127"> "Active master " + active.getServerName() + " managing " + rsCount + " regions servers");</span>
<span class="source-line-no">128</span><span id="line-128"> assertEquals(3, rsCount);</span>
<span class="source-line-no">129</span><span id="line-129"></span>
<span class="source-line-no">130</span><span id="line-130"> // wait for the active master to acknowledge loss of the backup from ZK</span>
<span class="source-line-no">131</span><span id="line-131"> final HMaster activeFinal = active;</span>
<span class="source-line-no">132</span><span id="line-132"> TEST_UTIL.waitFor(TimeUnit.MINUTES.toMillis(5),</span>
<span class="source-line-no">133</span><span id="line-133"> () -&gt; activeFinal.getBackupMasters().size() == 1);</span>
<span class="source-line-no">134</span><span id="line-134"></span>
<span class="source-line-no">135</span><span id="line-135"> // Check that ClusterStatus reports the correct active and backup masters</span>
<span class="source-line-no">136</span><span id="line-136"> assertNotNull(active);</span>
<span class="source-line-no">137</span><span id="line-137"> status = active.getClusterMetrics();</span>
<span class="source-line-no">138</span><span id="line-138"> assertEquals(activeName, status.getMasterName());</span>
<span class="source-line-no">139</span><span id="line-139"> assertEquals(1, status.getBackupMasterNames().size());</span>
<span class="source-line-no">140</span><span id="line-140"></span>
<span class="source-line-no">141</span><span id="line-141"> // kill the active master</span>
<span class="source-line-no">142</span><span id="line-142"> LOG.debug("\n\nStopping the active master " + active.getServerName() + "\n");</span>
<span class="source-line-no">143</span><span id="line-143"> cluster.stopMaster(activeIndex, false);</span>
<span class="source-line-no">144</span><span id="line-144"> cluster.waitOnMaster(activeIndex);</span>
<span class="source-line-no">145</span><span id="line-145"></span>
<span class="source-line-no">146</span><span id="line-146"> // wait for an active master to show up and be ready</span>
<span class="source-line-no">147</span><span id="line-147"> assertTrue(cluster.waitForActiveAndReadyMaster());</span>
<span class="source-line-no">148</span><span id="line-148"></span>
<span class="source-line-no">149</span><span id="line-149"> LOG.debug("\n\nVerifying backup master is now active\n");</span>
<span class="source-line-no">150</span><span id="line-150"> // should only have one master now</span>
<span class="source-line-no">151</span><span id="line-151"> assertEquals(1, masterThreads.size());</span>
<span class="source-line-no">152</span><span id="line-152"></span>
<span class="source-line-no">153</span><span id="line-153"> // and he should be active</span>
<span class="source-line-no">154</span><span id="line-154"> active = masterThreads.get(0).getMaster();</span>
<span class="source-line-no">155</span><span id="line-155"> assertNotNull(active);</span>
<span class="source-line-no">156</span><span id="line-156"> status = active.getClusterMetrics();</span>
<span class="source-line-no">157</span><span id="line-157"> ServerName masterName = status.getMasterName();</span>
<span class="source-line-no">158</span><span id="line-158"> assertNotNull(masterName);</span>
<span class="source-line-no">159</span><span id="line-159"> assertEquals(active.getServerName(), masterName);</span>
<span class="source-line-no">160</span><span id="line-160"> assertTrue(active.isActiveMaster());</span>
<span class="source-line-no">161</span><span id="line-161"> assertEquals(0, status.getBackupMasterNames().size());</span>
<span class="source-line-no">162</span><span id="line-162"> int rss = status.getLiveServerMetrics().size();</span>
<span class="source-line-no">163</span><span id="line-163"> LOG.info("Active master {} managing {} region servers", masterName.getServerName(), rss);</span>
<span class="source-line-no">164</span><span id="line-164"> assertEquals(3, rss);</span>
<span class="source-line-no">165</span><span id="line-165"> } finally {</span>
<span class="source-line-no">166</span><span id="line-166"> // Stop the cluster</span>
<span class="source-line-no">167</span><span id="line-167"> TEST_UTIL.shutdownMiniCluster();</span>
<span class="source-line-no">168</span><span id="line-168"> }</span>
<span class="source-line-no">169</span><span id="line-169"> }</span>
<span class="source-line-no">170</span><span id="line-170"></span>
<span class="source-line-no">171</span><span id="line-171"> /**</span>
<span class="source-line-no">172</span><span id="line-172"> * Test meta in transition when master failover. This test used to manipulate region state up in</span>
<span class="source-line-no">173</span><span id="line-173"> * zk. That is not allowed any more in hbase2 so I removed that messing. That makes this test</span>
<span class="source-line-no">174</span><span id="line-174"> * anemic.</span>
<span class="source-line-no">175</span><span id="line-175"> */</span>
<span class="source-line-no">176</span><span id="line-176"> @Test</span>
<span class="source-line-no">177</span><span id="line-177"> public void testMetaInTransitionWhenMasterFailover() throws Exception {</span>
<span class="source-line-no">178</span><span id="line-178"> // Start the cluster</span>
<span class="source-line-no">179</span><span id="line-179"> HBaseTestingUtil TEST_UTIL = new HBaseTestingUtil();</span>
<span class="source-line-no">180</span><span id="line-180"> TEST_UTIL.startMiniCluster();</span>
<span class="source-line-no">181</span><span id="line-181"> try {</span>
<span class="source-line-no">182</span><span id="line-182"> SingleProcessHBaseCluster cluster = TEST_UTIL.getHBaseCluster();</span>
<span class="source-line-no">183</span><span id="line-183"> LOG.info("Cluster started");</span>
<span class="source-line-no">184</span><span id="line-184"></span>
<span class="source-line-no">185</span><span id="line-185"> HMaster activeMaster = cluster.getMaster();</span>
<span class="source-line-no">186</span><span id="line-186"> ServerName metaServerName = cluster.getServerHoldingMeta();</span>
<span class="source-line-no">187</span><span id="line-187"> HRegionServer hrs = cluster.getRegionServer(metaServerName);</span>
<span class="source-line-no">188</span><span id="line-188"></span>
<span class="source-line-no">189</span><span id="line-189"> // Now kill master, meta should remain on rs, where we placed it before.</span>
<span class="source-line-no">190</span><span id="line-190"> LOG.info("Aborting master");</span>
<span class="source-line-no">191</span><span id="line-191"> activeMaster.abort("test-kill");</span>
<span class="source-line-no">192</span><span id="line-192"> cluster.waitForMasterToStop(activeMaster.getServerName(), 30000);</span>
<span class="source-line-no">193</span><span id="line-193"> LOG.info("Master has aborted");</span>
<span class="source-line-no">194</span><span id="line-194"></span>
<span class="source-line-no">195</span><span id="line-195"> // meta should remain where it was</span>
<span class="source-line-no">196</span><span id="line-196"> RegionState metaState = MetaTableLocator.getMetaRegionState(hrs.getZooKeeper());</span>
<span class="source-line-no">197</span><span id="line-197"> assertEquals("hbase:meta should be online on RS", metaState.getServerName(), metaServerName);</span>
<span class="source-line-no">198</span><span id="line-198"> assertEquals("hbase:meta should be online on RS", State.OPEN, metaState.getState());</span>
<span class="source-line-no">199</span><span id="line-199"></span>
<span class="source-line-no">200</span><span id="line-200"> // Start up a new master</span>
<span class="source-line-no">201</span><span id="line-201"> LOG.info("Starting up a new master");</span>
<span class="source-line-no">202</span><span id="line-202"> activeMaster = cluster.startMaster().getMaster();</span>
<span class="source-line-no">203</span><span id="line-203"> LOG.info("Waiting for master to be ready");</span>
<span class="source-line-no">204</span><span id="line-204"> cluster.waitForActiveAndReadyMaster();</span>
<span class="source-line-no">205</span><span id="line-205"> LOG.info("Master is ready");</span>
<span class="source-line-no">206</span><span id="line-206"></span>
<span class="source-line-no">207</span><span id="line-207"> // ensure meta is still deployed on RS</span>
<span class="source-line-no">208</span><span id="line-208"> metaState = MetaTableLocator.getMetaRegionState(activeMaster.getZooKeeper());</span>
<span class="source-line-no">209</span><span id="line-209"> assertEquals("hbase:meta should be online on RS", metaState.getServerName(), metaServerName);</span>
<span class="source-line-no">210</span><span id="line-210"> assertEquals("hbase:meta should be online on RS", State.OPEN, metaState.getState());</span>
<span class="source-line-no">211</span><span id="line-211"></span>
<span class="source-line-no">212</span><span id="line-212"> // Done, shutdown the cluster</span>
<span class="source-line-no">213</span><span id="line-213"> } finally {</span>
<span class="source-line-no">214</span><span id="line-214"> TEST_UTIL.shutdownMiniCluster();</span>
<span class="source-line-no">215</span><span id="line-215"> }</span>
<span class="source-line-no">216</span><span id="line-216"> }</span>
<span class="source-line-no">217</span><span id="line-217">}</span>
</pre>
</div>
</main>
</body>
</html>