| <!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: MasterWalManager"> |
| <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 java.io.IOException;</span> |
| <span class="source-line-no">021</span><span id="line-21">import java.util.ArrayList;</span> |
| <span class="source-line-no">022</span><span id="line-22">import java.util.Collections;</span> |
| <span class="source-line-no">023</span><span id="line-23">import java.util.HashSet;</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.Set;</span> |
| <span class="source-line-no">026</span><span id="line-26">import java.util.concurrent.locks.Lock;</span> |
| <span class="source-line-no">027</span><span id="line-27">import java.util.concurrent.locks.ReentrantLock;</span> |
| <span class="source-line-no">028</span><span id="line-28">import java.util.stream.Collectors;</span> |
| <span class="source-line-no">029</span><span id="line-29">import java.util.stream.Stream;</span> |
| <span class="source-line-no">030</span><span id="line-30">import org.apache.hadoop.conf.Configuration;</span> |
| <span class="source-line-no">031</span><span id="line-31">import org.apache.hadoop.fs.FileStatus;</span> |
| <span class="source-line-no">032</span><span id="line-32">import org.apache.hadoop.fs.FileSystem;</span> |
| <span class="source-line-no">033</span><span id="line-33">import org.apache.hadoop.fs.Path;</span> |
| <span class="source-line-no">034</span><span id="line-34">import org.apache.hadoop.fs.PathFilter;</span> |
| <span class="source-line-no">035</span><span id="line-35">import org.apache.hadoop.hbase.HConstants;</span> |
| <span class="source-line-no">036</span><span id="line-36">import org.apache.hadoop.hbase.ServerName;</span> |
| <span class="source-line-no">037</span><span id="line-37">import org.apache.hadoop.hbase.regionserver.wal.AbstractFSWAL;</span> |
| <span class="source-line-no">038</span><span id="line-38">import org.apache.hadoop.hbase.util.CommonFSUtils;</span> |
| <span class="source-line-no">039</span><span id="line-39">import org.apache.hadoop.hbase.util.EnvironmentEdgeManager;</span> |
| <span class="source-line-no">040</span><span id="line-40">import org.apache.hadoop.hbase.util.FSUtils;</span> |
| <span class="source-line-no">041</span><span id="line-41">import org.apache.hadoop.hbase.wal.AbstractFSWALProvider;</span> |
| <span class="source-line-no">042</span><span id="line-42">import org.apache.hadoop.hbase.wal.WALSplitter;</span> |
| <span class="source-line-no">043</span><span id="line-43">import org.apache.yetus.audience.InterfaceAudience;</span> |
| <span class="source-line-no">044</span><span id="line-44">import org.slf4j.Logger;</span> |
| <span class="source-line-no">045</span><span id="line-45">import org.slf4j.LoggerFactory;</span> |
| <span class="source-line-no">046</span><span id="line-46"></span> |
| <span class="source-line-no">047</span><span id="line-47">/**</span> |
| <span class="source-line-no">048</span><span id="line-48"> * This class abstracts a bunch of operations the HMaster needs when splitting log files e.g.</span> |
| <span class="source-line-no">049</span><span id="line-49"> * finding log files, dirs etc.</span> |
| <span class="source-line-no">050</span><span id="line-50"> */</span> |
| <span class="source-line-no">051</span><span id="line-51">@InterfaceAudience.Private</span> |
| <span class="source-line-no">052</span><span id="line-52">public class MasterWalManager {</span> |
| <span class="source-line-no">053</span><span id="line-53"> private static final Logger LOG = LoggerFactory.getLogger(MasterWalManager.class);</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"> * Filter *in* WAL files that are for the hbase:meta Region.</span> |
| <span class="source-line-no">057</span><span id="line-57"> */</span> |
| <span class="source-line-no">058</span><span id="line-58"> final static PathFilter META_FILTER = new PathFilter() {</span> |
| <span class="source-line-no">059</span><span id="line-59"> @Override</span> |
| <span class="source-line-no">060</span><span id="line-60"> public boolean accept(Path p) {</span> |
| <span class="source-line-no">061</span><span id="line-61"> return AbstractFSWALProvider.isMetaFile(p);</span> |
| <span class="source-line-no">062</span><span id="line-62"> }</span> |
| <span class="source-line-no">063</span><span id="line-63"> };</span> |
| <span class="source-line-no">064</span><span id="line-64"></span> |
| <span class="source-line-no">065</span><span id="line-65"> /**</span> |
| <span class="source-line-no">066</span><span id="line-66"> * Filter *out* WAL files that are for the hbase:meta Region; i.e. return user-space WALs only.</span> |
| <span class="source-line-no">067</span><span id="line-67"> */</span> |
| <span class="source-line-no">068</span><span id="line-68"> public final static PathFilter NON_META_FILTER = new PathFilter() {</span> |
| <span class="source-line-no">069</span><span id="line-69"> @Override</span> |
| <span class="source-line-no">070</span><span id="line-70"> public boolean accept(Path p) {</span> |
| <span class="source-line-no">071</span><span id="line-71"> return !AbstractFSWALProvider.isMetaFile(p);</span> |
| <span class="source-line-no">072</span><span id="line-72"> }</span> |
| <span class="source-line-no">073</span><span id="line-73"> };</span> |
| <span class="source-line-no">074</span><span id="line-74"></span> |
| <span class="source-line-no">075</span><span id="line-75"> // metrics for master</span> |
| <span class="source-line-no">076</span><span id="line-76"> // TODO: Rename it, since those metrics are split-manager related</span> |
| <span class="source-line-no">077</span><span id="line-77"> private final MetricsMasterFileSystem metricsMasterFilesystem = new MetricsMasterFileSystem();</span> |
| <span class="source-line-no">078</span><span id="line-78"></span> |
| <span class="source-line-no">079</span><span id="line-79"> // Keep around for convenience.</span> |
| <span class="source-line-no">080</span><span id="line-80"> private final MasterServices services;</span> |
| <span class="source-line-no">081</span><span id="line-81"> private final Configuration conf;</span> |
| <span class="source-line-no">082</span><span id="line-82"> private final FileSystem fs;</span> |
| <span class="source-line-no">083</span><span id="line-83"></span> |
| <span class="source-line-no">084</span><span id="line-84"> // The Path to the old logs dir</span> |
| <span class="source-line-no">085</span><span id="line-85"> private final Path oldLogDir;</span> |
| <span class="source-line-no">086</span><span id="line-86"></span> |
| <span class="source-line-no">087</span><span id="line-87"> private final Path rootDir;</span> |
| <span class="source-line-no">088</span><span id="line-88"></span> |
| <span class="source-line-no">089</span><span id="line-89"> // create the split log lock</span> |
| <span class="source-line-no">090</span><span id="line-90"> private final Lock splitLogLock = new ReentrantLock();</span> |
| <span class="source-line-no">091</span><span id="line-91"></span> |
| <span class="source-line-no">092</span><span id="line-92"> // old WALs directory size in bytes</span> |
| <span class="source-line-no">093</span><span id="line-93"> private long oldWALsDirSize;</span> |
| <span class="source-line-no">094</span><span id="line-94"></span> |
| <span class="source-line-no">095</span><span id="line-95"> /**</span> |
| <span class="source-line-no">096</span><span id="line-96"> * Superceded by {@link SplitWALManager}; i.e. procedure-based WAL splitting rather than 'classic'</span> |
| <span class="source-line-no">097</span><span id="line-97"> * zk-coordinated WAL splitting.</span> |
| <span class="source-line-no">098</span><span id="line-98"> * @deprecated since 2.3.0 and 3.0.0 to be removed in 4.0.0; replaced by {@link SplitWALManager}.</span> |
| <span class="source-line-no">099</span><span id="line-99"> * @see SplitWALManager</span> |
| <span class="source-line-no">100</span><span id="line-100"> */</span> |
| <span class="source-line-no">101</span><span id="line-101"> @Deprecated</span> |
| <span class="source-line-no">102</span><span id="line-102"> private final SplitLogManager splitLogManager;</span> |
| <span class="source-line-no">103</span><span id="line-103"></span> |
| <span class="source-line-no">104</span><span id="line-104"> // Is the fileystem ok?</span> |
| <span class="source-line-no">105</span><span id="line-105"> private volatile boolean fsOk = true;</span> |
| <span class="source-line-no">106</span><span id="line-106"></span> |
| <span class="source-line-no">107</span><span id="line-107"> public MasterWalManager(MasterServices services) throws IOException {</span> |
| <span class="source-line-no">108</span><span id="line-108"> this(services.getConfiguration(), services.getMasterFileSystem().getWALFileSystem(),</span> |
| <span class="source-line-no">109</span><span id="line-109"> services.getMasterFileSystem().getWALRootDir(), services);</span> |
| <span class="source-line-no">110</span><span id="line-110"> }</span> |
| <span class="source-line-no">111</span><span id="line-111"></span> |
| <span class="source-line-no">112</span><span id="line-112"> public MasterWalManager(Configuration conf, FileSystem fs, Path rootDir, MasterServices services)</span> |
| <span class="source-line-no">113</span><span id="line-113"> throws IOException {</span> |
| <span class="source-line-no">114</span><span id="line-114"> this.fs = fs;</span> |
| <span class="source-line-no">115</span><span id="line-115"> this.conf = conf;</span> |
| <span class="source-line-no">116</span><span id="line-116"> this.rootDir = rootDir;</span> |
| <span class="source-line-no">117</span><span id="line-117"> this.services = services;</span> |
| <span class="source-line-no">118</span><span id="line-118"> this.splitLogManager = new SplitLogManager(services, conf);</span> |
| <span class="source-line-no">119</span><span id="line-119"> this.oldLogDir = new Path(rootDir, HConstants.HREGION_OLDLOGDIR_NAME);</span> |
| <span class="source-line-no">120</span><span id="line-120"> this.oldWALsDirSize = 0;</span> |
| <span class="source-line-no">121</span><span id="line-121"> }</span> |
| <span class="source-line-no">122</span><span id="line-122"></span> |
| <span class="source-line-no">123</span><span id="line-123"> public void stop() {</span> |
| <span class="source-line-no">124</span><span id="line-124"> if (splitLogManager != null) {</span> |
| <span class="source-line-no">125</span><span id="line-125"> splitLogManager.stop();</span> |
| <span class="source-line-no">126</span><span id="line-126"> }</span> |
| <span class="source-line-no">127</span><span id="line-127"> }</span> |
| <span class="source-line-no">128</span><span id="line-128"></span> |
| <span class="source-line-no">129</span><span id="line-129"> SplitLogManager getSplitLogManager() {</span> |
| <span class="source-line-no">130</span><span id="line-130"> return this.splitLogManager;</span> |
| <span class="source-line-no">131</span><span id="line-131"> }</span> |
| <span class="source-line-no">132</span><span id="line-132"></span> |
| <span class="source-line-no">133</span><span id="line-133"> /**</span> |
| <span class="source-line-no">134</span><span id="line-134"> * Get the directory where old logs go</span> |
| <span class="source-line-no">135</span><span id="line-135"> * @return the dir</span> |
| <span class="source-line-no">136</span><span id="line-136"> */</span> |
| <span class="source-line-no">137</span><span id="line-137"> Path getOldLogDir() {</span> |
| <span class="source-line-no">138</span><span id="line-138"> return this.oldLogDir;</span> |
| <span class="source-line-no">139</span><span id="line-139"> }</span> |
| <span class="source-line-no">140</span><span id="line-140"></span> |
| <span class="source-line-no">141</span><span id="line-141"> public void updateOldWALsDirSize() throws IOException {</span> |
| <span class="source-line-no">142</span><span id="line-142"> this.oldWALsDirSize = fs.getContentSummary(this.oldLogDir).getLength();</span> |
| <span class="source-line-no">143</span><span id="line-143"> }</span> |
| <span class="source-line-no">144</span><span id="line-144"></span> |
| <span class="source-line-no">145</span><span id="line-145"> public long getOldWALsDirSize() {</span> |
| <span class="source-line-no">146</span><span id="line-146"> return this.oldWALsDirSize;</span> |
| <span class="source-line-no">147</span><span id="line-147"> }</span> |
| <span class="source-line-no">148</span><span id="line-148"></span> |
| <span class="source-line-no">149</span><span id="line-149"> public FileSystem getFileSystem() {</span> |
| <span class="source-line-no">150</span><span id="line-150"> return this.fs;</span> |
| <span class="source-line-no">151</span><span id="line-151"> }</span> |
| <span class="source-line-no">152</span><span id="line-152"></span> |
| <span class="source-line-no">153</span><span id="line-153"> /**</span> |
| <span class="source-line-no">154</span><span id="line-154"> * Checks to see if the file system is still accessible. If not, sets closed</span> |
| <span class="source-line-no">155</span><span id="line-155"> * @return false if file system is not available</span> |
| <span class="source-line-no">156</span><span id="line-156"> */</span> |
| <span class="source-line-no">157</span><span id="line-157"> private boolean checkFileSystem() {</span> |
| <span class="source-line-no">158</span><span id="line-158"> if (this.fsOk) {</span> |
| <span class="source-line-no">159</span><span id="line-159"> try {</span> |
| <span class="source-line-no">160</span><span id="line-160"> FSUtils.checkFileSystemAvailable(this.fs);</span> |
| <span class="source-line-no">161</span><span id="line-161"> FSUtils.checkDfsSafeMode(this.conf);</span> |
| <span class="source-line-no">162</span><span id="line-162"> } catch (IOException e) {</span> |
| <span class="source-line-no">163</span><span id="line-163"> services.abort("Shutting down HBase cluster: file system not available", e);</span> |
| <span class="source-line-no">164</span><span id="line-164"> this.fsOk = false;</span> |
| <span class="source-line-no">165</span><span id="line-165"> }</span> |
| <span class="source-line-no">166</span><span id="line-166"> }</span> |
| <span class="source-line-no">167</span><span id="line-167"> return this.fsOk;</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"> * Get Servernames which are currently splitting; paths have a '-splitting' suffix.</span> |
| <span class="source-line-no">172</span><span id="line-172"> */</span> |
| <span class="source-line-no">173</span><span id="line-173"> public Set<ServerName> getSplittingServersFromWALDir() throws IOException {</span> |
| <span class="source-line-no">174</span><span id="line-174"> return getServerNamesFromWALDirPath(</span> |
| <span class="source-line-no">175</span><span id="line-175"> p -> p.getName().endsWith(AbstractFSWALProvider.SPLITTING_EXT));</span> |
| <span class="source-line-no">176</span><span id="line-176"> }</span> |
| <span class="source-line-no">177</span><span id="line-177"></span> |
| <span class="source-line-no">178</span><span id="line-178"> /**</span> |
| <span class="source-line-no">179</span><span id="line-179"> * Get Servernames that COULD BE 'alive'; excludes those that have a '-splitting' suffix as these</span> |
| <span class="source-line-no">180</span><span id="line-180"> * are already being split -- they cannot be 'alive'.</span> |
| <span class="source-line-no">181</span><span id="line-181"> */</span> |
| <span class="source-line-no">182</span><span id="line-182"> public Set<ServerName> getLiveServersFromWALDir() throws IOException {</span> |
| <span class="source-line-no">183</span><span id="line-183"> return getServerNamesFromWALDirPath(</span> |
| <span class="source-line-no">184</span><span id="line-184"> p -> !p.getName().endsWith(AbstractFSWALProvider.SPLITTING_EXT));</span> |
| <span class="source-line-no">185</span><span id="line-185"> }</span> |
| <span class="source-line-no">186</span><span id="line-186"></span> |
| <span class="source-line-no">187</span><span id="line-187"> /** Returns listing of ServerNames found by parsing WAL directory paths in FS. */</span> |
| <span class="source-line-no">188</span><span id="line-188"> public Set<ServerName> getServerNamesFromWALDirPath(final PathFilter filter) throws IOException {</span> |
| <span class="source-line-no">189</span><span id="line-189"> FileStatus[] walDirForServerNames = getWALDirPaths(filter);</span> |
| <span class="source-line-no">190</span><span id="line-190"> return Stream.of(walDirForServerNames).map(s -> {</span> |
| <span class="source-line-no">191</span><span id="line-191"> ServerName serverName = AbstractFSWALProvider.getServerNameFromWALDirectoryName(s.getPath());</span> |
| <span class="source-line-no">192</span><span id="line-192"> if (serverName == null) {</span> |
| <span class="source-line-no">193</span><span id="line-193"> LOG.warn("Log folder {} doesn't look like its name includes a "</span> |
| <span class="source-line-no">194</span><span id="line-194"> + "region server name; leaving in place. If you see later errors about missing "</span> |
| <span class="source-line-no">195</span><span id="line-195"> + "write ahead logs they may be saved in this location.", s.getPath());</span> |
| <span class="source-line-no">196</span><span id="line-196"> return null;</span> |
| <span class="source-line-no">197</span><span id="line-197"> }</span> |
| <span class="source-line-no">198</span><span id="line-198"> return serverName;</span> |
| <span class="source-line-no">199</span><span id="line-199"> }).filter(s -> s != null).collect(Collectors.toSet());</span> |
| <span class="source-line-no">200</span><span id="line-200"> }</span> |
| <span class="source-line-no">201</span><span id="line-201"></span> |
| <span class="source-line-no">202</span><span id="line-202"> /**</span> |
| <span class="source-line-no">203</span><span id="line-203"> * Returns List of all RegionServer WAL dirs; i.e. this.rootDir/HConstants.HREGION_LOGDIR_NAME.</span> |
| <span class="source-line-no">204</span><span id="line-204"> */</span> |
| <span class="source-line-no">205</span><span id="line-205"> public FileStatus[] getWALDirPaths(final PathFilter filter) throws IOException {</span> |
| <span class="source-line-no">206</span><span id="line-206"> Path walDirPath = new Path(CommonFSUtils.getWALRootDir(conf), HConstants.HREGION_LOGDIR_NAME);</span> |
| <span class="source-line-no">207</span><span id="line-207"> FileStatus[] walDirForServerNames = CommonFSUtils.listStatus(fs, walDirPath, filter);</span> |
| <span class="source-line-no">208</span><span id="line-208"> return walDirForServerNames == null ? new FileStatus[0] : walDirForServerNames;</span> |
| <span class="source-line-no">209</span><span id="line-209"> }</span> |
| <span class="source-line-no">210</span><span id="line-210"></span> |
| <span class="source-line-no">211</span><span id="line-211"> /**</span> |
| <span class="source-line-no">212</span><span id="line-212"> * Inspect the log directory to find dead servers which need recovery work</span> |
| <span class="source-line-no">213</span><span id="line-213"> * @return A set of ServerNames which aren't running but still have WAL files left in file system</span> |
| <span class="source-line-no">214</span><span id="line-214"> * @deprecated With proc-v2, we can record the crash server with procedure store, so do not need</span> |
| <span class="source-line-no">215</span><span id="line-215"> * to scan the wal directory to find out the splitting wal directory any more. Leave</span> |
| <span class="source-line-no">216</span><span id="line-216"> * it here only because {@code RecoverMetaProcedure}(which is also deprecated) uses</span> |
| <span class="source-line-no">217</span><span id="line-217"> * it.</span> |
| <span class="source-line-no">218</span><span id="line-218"> */</span> |
| <span class="source-line-no">219</span><span id="line-219"> @Deprecated</span> |
| <span class="source-line-no">220</span><span id="line-220"> public Set<ServerName> getFailedServersFromLogFolders() throws IOException {</span> |
| <span class="source-line-no">221</span><span id="line-221"> boolean retrySplitting =</span> |
| <span class="source-line-no">222</span><span id="line-222"> !conf.getBoolean(WALSplitter.SPLIT_SKIP_ERRORS_KEY, WALSplitter.SPLIT_SKIP_ERRORS_DEFAULT);</span> |
| <span class="source-line-no">223</span><span id="line-223"></span> |
| <span class="source-line-no">224</span><span id="line-224"> Set<ServerName> serverNames = new HashSet<>();</span> |
| <span class="source-line-no">225</span><span id="line-225"> Path logsDirPath = new Path(CommonFSUtils.getWALRootDir(conf), HConstants.HREGION_LOGDIR_NAME);</span> |
| <span class="source-line-no">226</span><span id="line-226"></span> |
| <span class="source-line-no">227</span><span id="line-227"> do {</span> |
| <span class="source-line-no">228</span><span id="line-228"> if (services.isStopped()) {</span> |
| <span class="source-line-no">229</span><span id="line-229"> LOG.warn("Master stopped while trying to get failed servers.");</span> |
| <span class="source-line-no">230</span><span id="line-230"> break;</span> |
| <span class="source-line-no">231</span><span id="line-231"> }</span> |
| <span class="source-line-no">232</span><span id="line-232"> try {</span> |
| <span class="source-line-no">233</span><span id="line-233"> if (!this.fs.exists(logsDirPath)) return serverNames;</span> |
| <span class="source-line-no">234</span><span id="line-234"> FileStatus[] logFolders = CommonFSUtils.listStatus(this.fs, logsDirPath, null);</span> |
| <span class="source-line-no">235</span><span id="line-235"> // Get online servers after getting log folders to avoid log folder deletion of newly</span> |
| <span class="source-line-no">236</span><span id="line-236"> // checked in region servers . see HBASE-5916</span> |
| <span class="source-line-no">237</span><span id="line-237"> Set<ServerName> onlineServers = services.getServerManager().getOnlineServers().keySet();</span> |
| <span class="source-line-no">238</span><span id="line-238"></span> |
| <span class="source-line-no">239</span><span id="line-239"> if (logFolders == null || logFolders.length == 0) {</span> |
| <span class="source-line-no">240</span><span id="line-240"> LOG.debug("No log files to split, proceeding...");</span> |
| <span class="source-line-no">241</span><span id="line-241"> return serverNames;</span> |
| <span class="source-line-no">242</span><span id="line-242"> }</span> |
| <span class="source-line-no">243</span><span id="line-243"> for (FileStatus status : logFolders) {</span> |
| <span class="source-line-no">244</span><span id="line-244"> FileStatus[] curLogFiles = CommonFSUtils.listStatus(this.fs, status.getPath(), null);</span> |
| <span class="source-line-no">245</span><span id="line-245"> if (curLogFiles == null || curLogFiles.length == 0) {</span> |
| <span class="source-line-no">246</span><span id="line-246"> // Empty log folder. No recovery needed</span> |
| <span class="source-line-no">247</span><span id="line-247"> continue;</span> |
| <span class="source-line-no">248</span><span id="line-248"> }</span> |
| <span class="source-line-no">249</span><span id="line-249"> final ServerName serverName =</span> |
| <span class="source-line-no">250</span><span id="line-250"> AbstractFSWALProvider.getServerNameFromWALDirectoryName(status.getPath());</span> |
| <span class="source-line-no">251</span><span id="line-251"> if (null == serverName) {</span> |
| <span class="source-line-no">252</span><span id="line-252"> LOG.warn("Log folder " + status.getPath() + " doesn't look like its name includes a "</span> |
| <span class="source-line-no">253</span><span id="line-253"> + "region server name; leaving in place. If you see later errors about missing "</span> |
| <span class="source-line-no">254</span><span id="line-254"> + "write ahead logs they may be saved in this location.");</span> |
| <span class="source-line-no">255</span><span id="line-255"> } else if (!onlineServers.contains(serverName)) {</span> |
| <span class="source-line-no">256</span><span id="line-256"> LOG.info("Log folder " + status.getPath() + " doesn't belong "</span> |
| <span class="source-line-no">257</span><span id="line-257"> + "to a known region server, splitting");</span> |
| <span class="source-line-no">258</span><span id="line-258"> serverNames.add(serverName);</span> |
| <span class="source-line-no">259</span><span id="line-259"> } else {</span> |
| <span class="source-line-no">260</span><span id="line-260"> LOG.info("Log folder " + status.getPath() + " belongs to an existing region server");</span> |
| <span class="source-line-no">261</span><span id="line-261"> }</span> |
| <span class="source-line-no">262</span><span id="line-262"> }</span> |
| <span class="source-line-no">263</span><span id="line-263"> retrySplitting = false;</span> |
| <span class="source-line-no">264</span><span id="line-264"> } catch (IOException ioe) {</span> |
| <span class="source-line-no">265</span><span id="line-265"> LOG.warn("Failed getting failed servers to be recovered.", ioe);</span> |
| <span class="source-line-no">266</span><span id="line-266"> if (!checkFileSystem()) {</span> |
| <span class="source-line-no">267</span><span id="line-267"> LOG.warn("Bad Filesystem, exiting");</span> |
| <span class="source-line-no">268</span><span id="line-268"> Runtime.getRuntime().halt(1);</span> |
| <span class="source-line-no">269</span><span id="line-269"> }</span> |
| <span class="source-line-no">270</span><span id="line-270"> try {</span> |
| <span class="source-line-no">271</span><span id="line-271"> if (retrySplitting) {</span> |
| <span class="source-line-no">272</span><span id="line-272"> Thread.sleep(conf.getInt("hbase.hlog.split.failure.retry.interval", 30 * 1000));</span> |
| <span class="source-line-no">273</span><span id="line-273"> }</span> |
| <span class="source-line-no">274</span><span id="line-274"> } catch (InterruptedException e) {</span> |
| <span class="source-line-no">275</span><span id="line-275"> LOG.warn("Interrupted, aborting since cannot return w/o splitting");</span> |
| <span class="source-line-no">276</span><span id="line-276"> Thread.currentThread().interrupt();</span> |
| <span class="source-line-no">277</span><span id="line-277"> retrySplitting = false;</span> |
| <span class="source-line-no">278</span><span id="line-278"> Runtime.getRuntime().halt(1);</span> |
| <span class="source-line-no">279</span><span id="line-279"> }</span> |
| <span class="source-line-no">280</span><span id="line-280"> }</span> |
| <span class="source-line-no">281</span><span id="line-281"> } while (retrySplitting);</span> |
| <span class="source-line-no">282</span><span id="line-282"></span> |
| <span class="source-line-no">283</span><span id="line-283"> return serverNames;</span> |
| <span class="source-line-no">284</span><span id="line-284"> }</span> |
| <span class="source-line-no">285</span><span id="line-285"></span> |
| <span class="source-line-no">286</span><span id="line-286"> public void splitLog(final ServerName serverName) throws IOException {</span> |
| <span class="source-line-no">287</span><span id="line-287"> splitLog(Collections.<ServerName> singleton(serverName));</span> |
| <span class="source-line-no">288</span><span id="line-288"> }</span> |
| <span class="source-line-no">289</span><span id="line-289"></span> |
| <span class="source-line-no">290</span><span id="line-290"> /**</span> |
| <span class="source-line-no">291</span><span id="line-291"> * Specialized method to handle the splitting for meta WAL</span> |
| <span class="source-line-no">292</span><span id="line-292"> * @param serverName logs belonging to this server will be split</span> |
| <span class="source-line-no">293</span><span id="line-293"> */</span> |
| <span class="source-line-no">294</span><span id="line-294"> public void splitMetaLog(final ServerName serverName) throws IOException {</span> |
| <span class="source-line-no">295</span><span id="line-295"> splitMetaLog(Collections.<ServerName> singleton(serverName));</span> |
| <span class="source-line-no">296</span><span id="line-296"> }</span> |
| <span class="source-line-no">297</span><span id="line-297"></span> |
| <span class="source-line-no">298</span><span id="line-298"> /**</span> |
| <span class="source-line-no">299</span><span id="line-299"> * Specialized method to handle the splitting for meta WAL</span> |
| <span class="source-line-no">300</span><span id="line-300"> * @param serverNames logs belonging to these servers will be split</span> |
| <span class="source-line-no">301</span><span id="line-301"> */</span> |
| <span class="source-line-no">302</span><span id="line-302"> public void splitMetaLog(final Set<ServerName> serverNames) throws IOException {</span> |
| <span class="source-line-no">303</span><span id="line-303"> splitLog(serverNames, META_FILTER);</span> |
| <span class="source-line-no">304</span><span id="line-304"> }</span> |
| <span class="source-line-no">305</span><span id="line-305"></span> |
| <span class="source-line-no">306</span><span id="line-306"> @edu.umd.cs.findbugs.annotations.SuppressWarnings(value = "UL_UNRELEASED_LOCK",</span> |
| <span class="source-line-no">307</span><span id="line-307"> justification = "We only release this lock when we set it. Updates to code "</span> |
| <span class="source-line-no">308</span><span id="line-308"> + "that uses it should verify use of the guard boolean.")</span> |
| <span class="source-line-no">309</span><span id="line-309"> List<Path> getLogDirs(final Set<ServerName> serverNames) throws IOException {</span> |
| <span class="source-line-no">310</span><span id="line-310"> List<Path> logDirs = new ArrayList<>();</span> |
| <span class="source-line-no">311</span><span id="line-311"> boolean needReleaseLock = false;</span> |
| <span class="source-line-no">312</span><span id="line-312"> if (!this.services.isInitialized()) {</span> |
| <span class="source-line-no">313</span><span id="line-313"> // during master initialization, we could have multiple places splitting a same wal</span> |
| <span class="source-line-no">314</span><span id="line-314"> // XXX: Does this still exist after we move to proc-v2?</span> |
| <span class="source-line-no">315</span><span id="line-315"> this.splitLogLock.lock();</span> |
| <span class="source-line-no">316</span><span id="line-316"> needReleaseLock = true;</span> |
| <span class="source-line-no">317</span><span id="line-317"> }</span> |
| <span class="source-line-no">318</span><span id="line-318"> try {</span> |
| <span class="source-line-no">319</span><span id="line-319"> for (ServerName serverName : serverNames) {</span> |
| <span class="source-line-no">320</span><span id="line-320"> Path logDir =</span> |
| <span class="source-line-no">321</span><span id="line-321"> new Path(this.rootDir, AbstractFSWALProvider.getWALDirectoryName(serverName.toString()));</span> |
| <span class="source-line-no">322</span><span id="line-322"> Path splitDir = logDir.suffix(AbstractFSWALProvider.SPLITTING_EXT);</span> |
| <span class="source-line-no">323</span><span id="line-323"> // Rename the directory so a rogue RS doesn't create more WALs</span> |
| <span class="source-line-no">324</span><span id="line-324"> if (fs.exists(logDir)) {</span> |
| <span class="source-line-no">325</span><span id="line-325"> if (!this.fs.rename(logDir, splitDir)) {</span> |
| <span class="source-line-no">326</span><span id="line-326"> throw new IOException("Failed fs.rename for log split: " + logDir);</span> |
| <span class="source-line-no">327</span><span id="line-327"> }</span> |
| <span class="source-line-no">328</span><span id="line-328"> logDir = splitDir;</span> |
| <span class="source-line-no">329</span><span id="line-329"> LOG.debug("Renamed region directory: " + splitDir);</span> |
| <span class="source-line-no">330</span><span id="line-330"> } else if (!fs.exists(splitDir)) {</span> |
| <span class="source-line-no">331</span><span id="line-331"> LOG.info("Log dir for server " + serverName + " does not exist");</span> |
| <span class="source-line-no">332</span><span id="line-332"> continue;</span> |
| <span class="source-line-no">333</span><span id="line-333"> }</span> |
| <span class="source-line-no">334</span><span id="line-334"> logDirs.add(splitDir);</span> |
| <span class="source-line-no">335</span><span id="line-335"> }</span> |
| <span class="source-line-no">336</span><span id="line-336"> } catch (IOException ioe) {</span> |
| <span class="source-line-no">337</span><span id="line-337"> if (!checkFileSystem()) {</span> |
| <span class="source-line-no">338</span><span id="line-338"> this.services.abort("Aborting due to filesystem unavailable", ioe);</span> |
| <span class="source-line-no">339</span><span id="line-339"> throw ioe;</span> |
| <span class="source-line-no">340</span><span id="line-340"> }</span> |
| <span class="source-line-no">341</span><span id="line-341"> } finally {</span> |
| <span class="source-line-no">342</span><span id="line-342"> if (needReleaseLock) {</span> |
| <span class="source-line-no">343</span><span id="line-343"> this.splitLogLock.unlock();</span> |
| <span class="source-line-no">344</span><span id="line-344"> }</span> |
| <span class="source-line-no">345</span><span id="line-345"> }</span> |
| <span class="source-line-no">346</span><span id="line-346"> return logDirs;</span> |
| <span class="source-line-no">347</span><span id="line-347"> }</span> |
| <span class="source-line-no">348</span><span id="line-348"></span> |
| <span class="source-line-no">349</span><span id="line-349"> public void splitLog(final Set<ServerName> serverNames) throws IOException {</span> |
| <span class="source-line-no">350</span><span id="line-350"> splitLog(serverNames, NON_META_FILTER);</span> |
| <span class="source-line-no">351</span><span id="line-351"> }</span> |
| <span class="source-line-no">352</span><span id="line-352"></span> |
| <span class="source-line-no">353</span><span id="line-353"> /**</span> |
| <span class="source-line-no">354</span><span id="line-354"> * This method is the base split method that splits WAL files matching a filter. Callers should</span> |
| <span class="source-line-no">355</span><span id="line-355"> * pass the appropriate filter for meta and non-meta WALs.</span> |
| <span class="source-line-no">356</span><span id="line-356"> * @param serverNames logs belonging to these servers will be split; this will rename the log</span> |
| <span class="source-line-no">357</span><span id="line-357"> * directory out from under a soft-failed server</span> |
| <span class="source-line-no">358</span><span id="line-358"> */</span> |
| <span class="source-line-no">359</span><span id="line-359"> public void splitLog(final Set<ServerName> serverNames, PathFilter filter) throws IOException {</span> |
| <span class="source-line-no">360</span><span id="line-360"> long splitTime = 0, splitLogSize = 0;</span> |
| <span class="source-line-no">361</span><span id="line-361"> List<Path> logDirs = getLogDirs(serverNames);</span> |
| <span class="source-line-no">362</span><span id="line-362"></span> |
| <span class="source-line-no">363</span><span id="line-363"> splitLogManager.handleDeadWorkers(serverNames);</span> |
| <span class="source-line-no">364</span><span id="line-364"> splitTime = EnvironmentEdgeManager.currentTime();</span> |
| <span class="source-line-no">365</span><span id="line-365"> splitLogSize = splitLogManager.splitLogDistributed(serverNames, logDirs, filter);</span> |
| <span class="source-line-no">366</span><span id="line-366"> splitTime = EnvironmentEdgeManager.currentTime() - splitTime;</span> |
| <span class="source-line-no">367</span><span id="line-367"></span> |
| <span class="source-line-no">368</span><span id="line-368"> if (this.metricsMasterFilesystem != null) {</span> |
| <span class="source-line-no">369</span><span id="line-369"> if (filter == META_FILTER) {</span> |
| <span class="source-line-no">370</span><span id="line-370"> this.metricsMasterFilesystem.addMetaWALSplit(splitTime, splitLogSize);</span> |
| <span class="source-line-no">371</span><span id="line-371"> } else {</span> |
| <span class="source-line-no">372</span><span id="line-372"> this.metricsMasterFilesystem.addSplit(splitTime, splitLogSize);</span> |
| <span class="source-line-no">373</span><span id="line-373"> }</span> |
| <span class="source-line-no">374</span><span id="line-374"> }</span> |
| <span class="source-line-no">375</span><span id="line-375"> }</span> |
| <span class="source-line-no">376</span><span id="line-376"></span> |
| <span class="source-line-no">377</span><span id="line-377"> /**</span> |
| <span class="source-line-no">378</span><span id="line-378"> * The hbase:meta region may OPEN and CLOSE without issue on a server and then move elsewhere. On</span> |
| <span class="source-line-no">379</span><span id="line-379"> * CLOSE, the WAL for the hbase:meta table may not be archived yet (The WAL is only needed if</span> |
| <span class="source-line-no">380</span><span id="line-380"> * hbase:meta did not close cleanaly). Since meta region is no long on this server, the</span> |
| <span class="source-line-no">381</span><span id="line-381"> * ServerCrashProcedure won't split these leftover hbase:meta WALs, just leaving them in the WAL</span> |
| <span class="source-line-no">382</span><span id="line-382"> * splitting dir. If we try to delete the WAL splitting for the server, it fail since the dir is</span> |
| <span class="source-line-no">383</span><span id="line-383"> * not totally empty. We can safely archive these hbase:meta log; then the WAL dir can be deleted.</span> |
| <span class="source-line-no">384</span><span id="line-384"> * @param serverName the server to archive meta log</span> |
| <span class="source-line-no">385</span><span id="line-385"> */</span> |
| <span class="source-line-no">386</span><span id="line-386"> public void archiveMetaLog(final ServerName serverName) {</span> |
| <span class="source-line-no">387</span><span id="line-387"> try {</span> |
| <span class="source-line-no">388</span><span id="line-388"> Path logDir =</span> |
| <span class="source-line-no">389</span><span id="line-389"> new Path(this.rootDir, AbstractFSWALProvider.getWALDirectoryName(serverName.toString()));</span> |
| <span class="source-line-no">390</span><span id="line-390"> Path splitDir = logDir.suffix(AbstractFSWALProvider.SPLITTING_EXT);</span> |
| <span class="source-line-no">391</span><span id="line-391"> if (fs.exists(splitDir)) {</span> |
| <span class="source-line-no">392</span><span id="line-392"> FileStatus[] logfiles = CommonFSUtils.listStatus(fs, splitDir, META_FILTER);</span> |
| <span class="source-line-no">393</span><span id="line-393"> if (logfiles != null) {</span> |
| <span class="source-line-no">394</span><span id="line-394"> for (FileStatus status : logfiles) {</span> |
| <span class="source-line-no">395</span><span id="line-395"> if (!status.isDir()) {</span> |
| <span class="source-line-no">396</span><span id="line-396"> Path newPath = AbstractFSWAL.getWALArchivePath(this.oldLogDir, status.getPath());</span> |
| <span class="source-line-no">397</span><span id="line-397"> if (!CommonFSUtils.renameAndSetModifyTime(fs, status.getPath(), newPath)) {</span> |
| <span class="source-line-no">398</span><span id="line-398"> LOG.warn("Unable to move " + status.getPath() + " to " + newPath);</span> |
| <span class="source-line-no">399</span><span id="line-399"> } else {</span> |
| <span class="source-line-no">400</span><span id="line-400"> LOG.debug("Archived meta log " + status.getPath() + " to " + newPath);</span> |
| <span class="source-line-no">401</span><span id="line-401"> }</span> |
| <span class="source-line-no">402</span><span id="line-402"> }</span> |
| <span class="source-line-no">403</span><span id="line-403"> }</span> |
| <span class="source-line-no">404</span><span id="line-404"> }</span> |
| <span class="source-line-no">405</span><span id="line-405"> if (!fs.delete(splitDir, false)) {</span> |
| <span class="source-line-no">406</span><span id="line-406"> LOG.warn("Unable to delete log dir. Ignoring. " + splitDir);</span> |
| <span class="source-line-no">407</span><span id="line-407"> }</span> |
| <span class="source-line-no">408</span><span id="line-408"> }</span> |
| <span class="source-line-no">409</span><span id="line-409"> } catch (IOException ie) {</span> |
| <span class="source-line-no">410</span><span id="line-410"> LOG.warn("Failed archiving meta log for server " + serverName, ie);</span> |
| <span class="source-line-no">411</span><span id="line-411"> }</span> |
| <span class="source-line-no">412</span><span id="line-412"> }</span> |
| <span class="source-line-no">413</span><span id="line-413">}</span> |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| </pre> |
| </div> |
| </main> |
| </body> |
| </html> |