| <!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.regionserver, class: TestRSKilledWhenInitializing"> |
| <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.regionserver;</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.assertTrue;</span> |
| <span class="source-line-no">022</span><span id="line-22"></span> |
| <span class="source-line-no">023</span><span id="line-23">import java.io.IOException;</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.Map;</span> |
| <span class="source-line-no">026</span><span id="line-26">import java.util.concurrent.atomic.AtomicBoolean;</span> |
| <span class="source-line-no">027</span><span id="line-27">import java.util.concurrent.atomic.AtomicReference;</span> |
| <span class="source-line-no">028</span><span id="line-28">import org.apache.hadoop.conf.Configuration;</span> |
| <span class="source-line-no">029</span><span id="line-29">import org.apache.hadoop.hbase.HBaseClassTestRule;</span> |
| <span class="source-line-no">030</span><span id="line-30">import org.apache.hadoop.hbase.HBaseConfiguration;</span> |
| <span class="source-line-no">031</span><span id="line-31">import org.apache.hadoop.hbase.HBaseTestingUtil;</span> |
| <span class="source-line-no">032</span><span id="line-32">import org.apache.hadoop.hbase.LocalHBaseCluster;</span> |
| <span class="source-line-no">033</span><span id="line-33">import org.apache.hadoop.hbase.ServerName;</span> |
| <span class="source-line-no">034</span><span id="line-34">import org.apache.hadoop.hbase.SingleProcessHBaseCluster;</span> |
| <span class="source-line-no">035</span><span id="line-35">import org.apache.hadoop.hbase.client.RegionInfo;</span> |
| <span class="source-line-no">036</span><span id="line-36">import org.apache.hadoop.hbase.master.HMaster;</span> |
| <span class="source-line-no">037</span><span id="line-37">import org.apache.hadoop.hbase.master.ServerListener;</span> |
| <span class="source-line-no">038</span><span id="line-38">import org.apache.hadoop.hbase.master.ServerManager;</span> |
| <span class="source-line-no">039</span><span id="line-39">import org.apache.hadoop.hbase.testclassification.MediumTests;</span> |
| <span class="source-line-no">040</span><span id="line-40">import org.apache.hadoop.hbase.testclassification.RegionServerTests;</span> |
| <span class="source-line-no">041</span><span id="line-41">import org.apache.hadoop.hbase.util.Bytes;</span> |
| <span class="source-line-no">042</span><span id="line-42">import org.apache.hadoop.hbase.util.JVMClusterUtil.MasterThread;</span> |
| <span class="source-line-no">043</span><span id="line-43">import org.apache.hadoop.hbase.util.Threads;</span> |
| <span class="source-line-no">044</span><span id="line-44">import org.junit.ClassRule;</span> |
| <span class="source-line-no">045</span><span id="line-45">import org.junit.Ignore;</span> |
| <span class="source-line-no">046</span><span id="line-46">import org.junit.Rule;</span> |
| <span class="source-line-no">047</span><span id="line-47">import org.junit.Test;</span> |
| <span class="source-line-no">048</span><span id="line-48">import org.junit.experimental.categories.Category;</span> |
| <span class="source-line-no">049</span><span id="line-49">import org.junit.rules.TestName;</span> |
| <span class="source-line-no">050</span><span id="line-50">import org.slf4j.Logger;</span> |
| <span class="source-line-no">051</span><span id="line-51">import org.slf4j.LoggerFactory;</span> |
| <span class="source-line-no">052</span><span id="line-52"></span> |
| <span class="source-line-no">053</span><span id="line-53">import org.apache.hadoop.hbase.shaded.protobuf.generated.RegionServerStatusProtos.RegionServerStartupResponse;</span> |
| <span class="source-line-no">054</span><span id="line-54"></span> |
| <span class="source-line-no">055</span><span id="line-55">/**</span> |
| <span class="source-line-no">056</span><span id="line-56"> * Tests that a regionserver that dies after reporting for duty gets removed from list of online</span> |
| <span class="source-line-no">057</span><span id="line-57"> * regions. See HBASE-9593.</span> |
| <span class="source-line-no">058</span><span id="line-58"> */</span> |
| <span class="source-line-no">059</span><span id="line-59">@Category({ RegionServerTests.class, MediumTests.class })</span> |
| <span class="source-line-no">060</span><span id="line-60">@Ignore("See HBASE-19515")</span> |
| <span class="source-line-no">061</span><span id="line-61">public class TestRSKilledWhenInitializing {</span> |
| <span class="source-line-no">062</span><span id="line-62"></span> |
| <span class="source-line-no">063</span><span id="line-63"> @ClassRule</span> |
| <span class="source-line-no">064</span><span id="line-64"> public static final HBaseClassTestRule CLASS_RULE =</span> |
| <span class="source-line-no">065</span><span id="line-65"> HBaseClassTestRule.forClass(TestRSKilledWhenInitializing.class);</span> |
| <span class="source-line-no">066</span><span id="line-66"></span> |
| <span class="source-line-no">067</span><span id="line-67"> private static final Logger LOG = LoggerFactory.getLogger(TestRSKilledWhenInitializing.class);</span> |
| <span class="source-line-no">068</span><span id="line-68"></span> |
| <span class="source-line-no">069</span><span id="line-69"> @Rule</span> |
| <span class="source-line-no">070</span><span id="line-70"> public TestName testName = new TestName();</span> |
| <span class="source-line-no">071</span><span id="line-71"></span> |
| <span class="source-line-no">072</span><span id="line-72"> // This boolean needs to be globally available. It is used below in our</span> |
| <span class="source-line-no">073</span><span id="line-73"> // mocked up regionserver so it knows when to die.</span> |
| <span class="source-line-no">074</span><span id="line-74"> private static AtomicBoolean masterActive = new AtomicBoolean(false);</span> |
| <span class="source-line-no">075</span><span id="line-75"> // Ditto for this variable. It also is used in the mocked regionserver class.</span> |
| <span class="source-line-no">076</span><span id="line-76"> private static final AtomicReference<ServerName> killedRS = new AtomicReference<ServerName>();</span> |
| <span class="source-line-no">077</span><span id="line-77"></span> |
| <span class="source-line-no">078</span><span id="line-78"> private static final int NUM_MASTERS = 1;</span> |
| <span class="source-line-no">079</span><span id="line-79"> private static final int NUM_RS = 2;</span> |
| <span class="source-line-no">080</span><span id="line-80"></span> |
| <span class="source-line-no">081</span><span id="line-81"> /**</span> |
| <span class="source-line-no">082</span><span id="line-82"> * Test verifies whether a region server is removed from online servers list in master if it went</span> |
| <span class="source-line-no">083</span><span id="line-83"> * down after registering with master. Test will TIMEOUT if an error!!!!</span> |
| <span class="source-line-no">084</span><span id="line-84"> */</span> |
| <span class="source-line-no">085</span><span id="line-85"> @Test</span> |
| <span class="source-line-no">086</span><span id="line-86"> public void testRSTerminationAfterRegisteringToMasterBeforeCreatingEphemeralNode()</span> |
| <span class="source-line-no">087</span><span id="line-87"> throws Exception {</span> |
| <span class="source-line-no">088</span><span id="line-88"> // Create config to use for this cluster</span> |
| <span class="source-line-no">089</span><span id="line-89"> Configuration conf = HBaseConfiguration.create();</span> |
| <span class="source-line-no">090</span><span id="line-90"> conf.setInt(ServerManager.WAIT_ON_REGIONSERVERS_MINTOSTART, 1);</span> |
| <span class="source-line-no">091</span><span id="line-91"> // Start the cluster</span> |
| <span class="source-line-no">092</span><span id="line-92"> final HBaseTestingUtil TEST_UTIL = new HBaseTestingUtil(conf);</span> |
| <span class="source-line-no">093</span><span id="line-93"> TEST_UTIL.startMiniDFSCluster(3);</span> |
| <span class="source-line-no">094</span><span id="line-94"> TEST_UTIL.startMiniZKCluster();</span> |
| <span class="source-line-no">095</span><span id="line-95"> TEST_UTIL.createRootDir();</span> |
| <span class="source-line-no">096</span><span id="line-96"> final LocalHBaseCluster cluster = new LocalHBaseCluster(conf, NUM_MASTERS, NUM_RS,</span> |
| <span class="source-line-no">097</span><span id="line-97"> HMaster.class, RegisterAndDieRegionServer.class);</span> |
| <span class="source-line-no">098</span><span id="line-98"> final MasterThread master = startMaster(cluster.getMasters().get(0));</span> |
| <span class="source-line-no">099</span><span id="line-99"> try {</span> |
| <span class="source-line-no">100</span><span id="line-100"> // Master is up waiting on RegionServers to check in. Now start RegionServers.</span> |
| <span class="source-line-no">101</span><span id="line-101"> for (int i = 0; i < NUM_RS; i++) {</span> |
| <span class="source-line-no">102</span><span id="line-102"> cluster.getRegionServers().get(i).start();</span> |
| <span class="source-line-no">103</span><span id="line-103"> }</span> |
| <span class="source-line-no">104</span><span id="line-104"> // Expected total regionservers depends on whether Master can host regions or not.</span> |
| <span class="source-line-no">105</span><span id="line-105"> int expectedTotalRegionServers = NUM_RS;</span> |
| <span class="source-line-no">106</span><span id="line-106"> List<ServerName> onlineServersList = null;</span> |
| <span class="source-line-no">107</span><span id="line-107"> do {</span> |
| <span class="source-line-no">108</span><span id="line-108"> onlineServersList = master.getMaster().getServerManager().getOnlineServersList();</span> |
| <span class="source-line-no">109</span><span id="line-109"> } while (onlineServersList.size() < expectedTotalRegionServers);</span> |
| <span class="source-line-no">110</span><span id="line-110"> // Wait until killedRS is set. Means RegionServer is starting to go down.</span> |
| <span class="source-line-no">111</span><span id="line-111"> while (killedRS.get() == null) {</span> |
| <span class="source-line-no">112</span><span id="line-112"> Threads.sleep(1);</span> |
| <span class="source-line-no">113</span><span id="line-113"> }</span> |
| <span class="source-line-no">114</span><span id="line-114"> // Wait on the RegionServer to fully die.</span> |
| <span class="source-line-no">115</span><span id="line-115"> while (cluster.getLiveRegionServers().size() >= expectedTotalRegionServers) {</span> |
| <span class="source-line-no">116</span><span id="line-116"> Threads.sleep(1);</span> |
| <span class="source-line-no">117</span><span id="line-117"> }</span> |
| <span class="source-line-no">118</span><span id="line-118"> // Make sure Master is fully up before progressing. Could take a while if regions</span> |
| <span class="source-line-no">119</span><span id="line-119"> // being reassigned.</span> |
| <span class="source-line-no">120</span><span id="line-120"> while (!master.getMaster().isInitialized()) {</span> |
| <span class="source-line-no">121</span><span id="line-121"> Threads.sleep(1);</span> |
| <span class="source-line-no">122</span><span id="line-122"> }</span> |
| <span class="source-line-no">123</span><span id="line-123"></span> |
| <span class="source-line-no">124</span><span id="line-124"> // Now in steady state. How many regions open? Master should have too many regionservers</span> |
| <span class="source-line-no">125</span><span id="line-125"> // showing still. The downed RegionServer should still be showing as registered.</span> |
| <span class="source-line-no">126</span><span id="line-126"> assertTrue(master.getMaster().getServerManager().isServerOnline(killedRS.get()));</span> |
| <span class="source-line-no">127</span><span id="line-127"> // Find non-meta region (namespace?) and assign to the killed server. That'll trigger cleanup.</span> |
| <span class="source-line-no">128</span><span id="line-128"> Map<RegionInfo, ServerName> assignments = null;</span> |
| <span class="source-line-no">129</span><span id="line-129"> do {</span> |
| <span class="source-line-no">130</span><span id="line-130"> assignments =</span> |
| <span class="source-line-no">131</span><span id="line-131"> master.getMaster().getAssignmentManager().getRegionStates().getRegionAssignments();</span> |
| <span class="source-line-no">132</span><span id="line-132"> } while (assignments == null || assignments.size() < 2);</span> |
| <span class="source-line-no">133</span><span id="line-133"> RegionInfo hri = null;</span> |
| <span class="source-line-no">134</span><span id="line-134"> for (Map.Entry<RegionInfo, ServerName> e : assignments.entrySet()) {</span> |
| <span class="source-line-no">135</span><span id="line-135"> if (e.getKey().isMetaRegion()) continue;</span> |
| <span class="source-line-no">136</span><span id="line-136"> hri = e.getKey();</span> |
| <span class="source-line-no">137</span><span id="line-137"> break;</span> |
| <span class="source-line-no">138</span><span id="line-138"> }</span> |
| <span class="source-line-no">139</span><span id="line-139"> // Try moving region to the killed server. It will fail. As by-product, we will</span> |
| <span class="source-line-no">140</span><span id="line-140"> // remove the RS from Master online list because no corresponding znode.</span> |
| <span class="source-line-no">141</span><span id="line-141"> assertEquals(expectedTotalRegionServers,</span> |
| <span class="source-line-no">142</span><span id="line-142"> master.getMaster().getServerManager().getOnlineServersList().size());</span> |
| <span class="source-line-no">143</span><span id="line-143"> LOG.info("Move " + hri.getEncodedName() + " to " + killedRS.get());</span> |
| <span class="source-line-no">144</span><span id="line-144"> master.getMaster().move(hri.getEncodedNameAsBytes(),</span> |
| <span class="source-line-no">145</span><span id="line-145"> Bytes.toBytes(killedRS.get().toString()));</span> |
| <span class="source-line-no">146</span><span id="line-146"></span> |
| <span class="source-line-no">147</span><span id="line-147"> // TODO: This test could do more to verify fix. It could create a table</span> |
| <span class="source-line-no">148</span><span id="line-148"> // and do round-robin assign. It should fail if zombie RS. HBASE-19515.</span> |
| <span class="source-line-no">149</span><span id="line-149"></span> |
| <span class="source-line-no">150</span><span id="line-150"> // Wait until the RS no longer shows as registered in Master.</span> |
| <span class="source-line-no">151</span><span id="line-151"> while (onlineServersList.size() > (NUM_RS + 1)) {</span> |
| <span class="source-line-no">152</span><span id="line-152"> Thread.sleep(100);</span> |
| <span class="source-line-no">153</span><span id="line-153"> onlineServersList = master.getMaster().getServerManager().getOnlineServersList();</span> |
| <span class="source-line-no">154</span><span id="line-154"> }</span> |
| <span class="source-line-no">155</span><span id="line-155"> } finally {</span> |
| <span class="source-line-no">156</span><span id="line-156"> // Shutdown is messy with complaints about fs being closed. Why? TODO.</span> |
| <span class="source-line-no">157</span><span id="line-157"> cluster.shutdown();</span> |
| <span class="source-line-no">158</span><span id="line-158"> cluster.join();</span> |
| <span class="source-line-no">159</span><span id="line-159"> TEST_UTIL.shutdownMiniDFSCluster();</span> |
| <span class="source-line-no">160</span><span id="line-160"> TEST_UTIL.shutdownMiniZKCluster();</span> |
| <span class="source-line-no">161</span><span id="line-161"> TEST_UTIL.cleanupTestDir();</span> |
| <span class="source-line-no">162</span><span id="line-162"> }</span> |
| <span class="source-line-no">163</span><span id="line-163"> }</span> |
| <span class="source-line-no">164</span><span id="line-164"></span> |
| <span class="source-line-no">165</span><span id="line-165"> /**</span> |
| <span class="source-line-no">166</span><span id="line-166"> * Start Master. Get as far as the state where Master is waiting on RegionServers to check in,</span> |
| <span class="source-line-no">167</span><span id="line-167"> * then return.</span> |
| <span class="source-line-no">168</span><span id="line-168"> */</span> |
| <span class="source-line-no">169</span><span id="line-169"> private MasterThread startMaster(MasterThread master) {</span> |
| <span class="source-line-no">170</span><span id="line-170"> master.start();</span> |
| <span class="source-line-no">171</span><span id="line-171"> // It takes a while until ServerManager creation to happen inside Master startup.</span> |
| <span class="source-line-no">172</span><span id="line-172"> while (master.getMaster().getServerManager() == null) {</span> |
| <span class="source-line-no">173</span><span id="line-173"> continue;</span> |
| <span class="source-line-no">174</span><span id="line-174"> }</span> |
| <span class="source-line-no">175</span><span id="line-175"> // Set a listener for the waiting-on-RegionServers state. We want to wait</span> |
| <span class="source-line-no">176</span><span id="line-176"> // until this condition before we leave this method and start regionservers.</span> |
| <span class="source-line-no">177</span><span id="line-177"> final AtomicBoolean waiting = new AtomicBoolean(false);</span> |
| <span class="source-line-no">178</span><span id="line-178"> if (master.getMaster().getServerManager() == null) throw new NullPointerException("SM");</span> |
| <span class="source-line-no">179</span><span id="line-179"> master.getMaster().getServerManager().registerListener(new ServerListener() {</span> |
| <span class="source-line-no">180</span><span id="line-180"> @Override</span> |
| <span class="source-line-no">181</span><span id="line-181"> public void waiting() {</span> |
| <span class="source-line-no">182</span><span id="line-182"> waiting.set(true);</span> |
| <span class="source-line-no">183</span><span id="line-183"> }</span> |
| <span class="source-line-no">184</span><span id="line-184"> });</span> |
| <span class="source-line-no">185</span><span id="line-185"> // Wait until the Master gets to place where it is waiting on RegionServers to check in.</span> |
| <span class="source-line-no">186</span><span id="line-186"> while (!waiting.get()) {</span> |
| <span class="source-line-no">187</span><span id="line-187"> continue;</span> |
| <span class="source-line-no">188</span><span id="line-188"> }</span> |
| <span class="source-line-no">189</span><span id="line-189"> // Set the global master-is-active; gets picked up by regionservers later.</span> |
| <span class="source-line-no">190</span><span id="line-190"> masterActive.set(true);</span> |
| <span class="source-line-no">191</span><span id="line-191"> return master;</span> |
| <span class="source-line-no">192</span><span id="line-192"> }</span> |
| <span class="source-line-no">193</span><span id="line-193"></span> |
| <span class="source-line-no">194</span><span id="line-194"> /**</span> |
| <span class="source-line-no">195</span><span id="line-195"> * A RegionServer that reports for duty and then immediately dies if it is the first to receive</span> |
| <span class="source-line-no">196</span><span id="line-196"> * the response to a reportForDuty. When it dies, it clears its ephemeral znode which the master</span> |
| <span class="source-line-no">197</span><span id="line-197"> * notices and so removes the region from its set of online regionservers.</span> |
| <span class="source-line-no">198</span><span id="line-198"> */</span> |
| <span class="source-line-no">199</span><span id="line-199"> static class RegisterAndDieRegionServer</span> |
| <span class="source-line-no">200</span><span id="line-200"> extends SingleProcessHBaseCluster.MiniHBaseClusterRegionServer {</span> |
| <span class="source-line-no">201</span><span id="line-201"> public RegisterAndDieRegionServer(Configuration conf) throws IOException, InterruptedException {</span> |
| <span class="source-line-no">202</span><span id="line-202"> super(conf);</span> |
| <span class="source-line-no">203</span><span id="line-203"> }</span> |
| <span class="source-line-no">204</span><span id="line-204"></span> |
| <span class="source-line-no">205</span><span id="line-205"> @Override</span> |
| <span class="source-line-no">206</span><span id="line-206"> protected void handleReportForDutyResponse(RegionServerStartupResponse c) throws IOException {</span> |
| <span class="source-line-no">207</span><span id="line-207"> if (killedRS.compareAndSet(null, getServerName())) {</span> |
| <span class="source-line-no">208</span><span id="line-208"> // Make sure Master is up so it will see the removal of the ephemeral znode for this RS.</span> |
| <span class="source-line-no">209</span><span id="line-209"> while (!masterActive.get()) {</span> |
| <span class="source-line-no">210</span><span id="line-210"> Threads.sleep(100);</span> |
| <span class="source-line-no">211</span><span id="line-211"> }</span> |
| <span class="source-line-no">212</span><span id="line-212"> super.kill();</span> |
| <span class="source-line-no">213</span><span id="line-213"> } else {</span> |
| <span class="source-line-no">214</span><span id="line-214"> super.handleReportForDutyResponse(c);</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> |
| <span class="source-line-no">218</span><span id="line-218">}</span> |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| </pre> |
| </div> |
| </main> |
| </body> |
| </html> |