| <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd"> |
| <html lang="en"> |
| <head> |
| <title>Source code</title> |
| <link rel="stylesheet" type="text/css" href="../../../../../../stylesheet.css" title="Style"> |
| </head> |
| <body> |
| <div class="sourceContainer"> |
| <pre><span class="sourceLineNo">001</span>/**<a name="line.1"></a> |
| <span class="sourceLineNo">002</span> *<a name="line.2"></a> |
| <span class="sourceLineNo">003</span> * Licensed to the Apache Software Foundation (ASF) under one<a name="line.3"></a> |
| <span class="sourceLineNo">004</span> * or more contributor license agreements. See the NOTICE file<a name="line.4"></a> |
| <span class="sourceLineNo">005</span> * distributed with this work for additional information<a name="line.5"></a> |
| <span class="sourceLineNo">006</span> * regarding copyright ownership. The ASF licenses this file<a name="line.6"></a> |
| <span class="sourceLineNo">007</span> * to you under the Apache License, Version 2.0 (the<a name="line.7"></a> |
| <span class="sourceLineNo">008</span> * "License"); you may not use this file except in compliance<a name="line.8"></a> |
| <span class="sourceLineNo">009</span> * with the License. You may obtain a copy of the License at<a name="line.9"></a> |
| <span class="sourceLineNo">010</span> *<a name="line.10"></a> |
| <span class="sourceLineNo">011</span> * http://www.apache.org/licenses/LICENSE-2.0<a name="line.11"></a> |
| <span class="sourceLineNo">012</span> *<a name="line.12"></a> |
| <span class="sourceLineNo">013</span> * Unless required by applicable law or agreed to in writing, software<a name="line.13"></a> |
| <span class="sourceLineNo">014</span> * distributed under the License is distributed on an "AS IS" BASIS,<a name="line.14"></a> |
| <span class="sourceLineNo">015</span> * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.<a name="line.15"></a> |
| <span class="sourceLineNo">016</span> * See the License for the specific language governing permissions and<a name="line.16"></a> |
| <span class="sourceLineNo">017</span> * limitations under the License.<a name="line.17"></a> |
| <span class="sourceLineNo">018</span> */<a name="line.18"></a> |
| <span class="sourceLineNo">019</span>package org.apache.hadoop.hbase.master;<a name="line.19"></a> |
| <span class="sourceLineNo">020</span><a name="line.20"></a> |
| <span class="sourceLineNo">021</span>import java.io.IOException;<a name="line.21"></a> |
| <span class="sourceLineNo">022</span>import java.util.ArrayList;<a name="line.22"></a> |
| <span class="sourceLineNo">023</span>import java.util.Collections;<a name="line.23"></a> |
| <span class="sourceLineNo">024</span>import java.util.HashSet;<a name="line.24"></a> |
| <span class="sourceLineNo">025</span>import java.util.List;<a name="line.25"></a> |
| <span class="sourceLineNo">026</span>import java.util.Set;<a name="line.26"></a> |
| <span class="sourceLineNo">027</span>import java.util.concurrent.locks.Lock;<a name="line.27"></a> |
| <span class="sourceLineNo">028</span>import java.util.concurrent.locks.ReentrantLock;<a name="line.28"></a> |
| <span class="sourceLineNo">029</span>import java.util.stream.Collectors;<a name="line.29"></a> |
| <span class="sourceLineNo">030</span>import java.util.stream.Stream;<a name="line.30"></a> |
| <span class="sourceLineNo">031</span>import org.apache.hadoop.conf.Configuration;<a name="line.31"></a> |
| <span class="sourceLineNo">032</span>import org.apache.hadoop.fs.FileStatus;<a name="line.32"></a> |
| <span class="sourceLineNo">033</span>import org.apache.hadoop.fs.FileSystem;<a name="line.33"></a> |
| <span class="sourceLineNo">034</span>import org.apache.hadoop.fs.Path;<a name="line.34"></a> |
| <span class="sourceLineNo">035</span>import org.apache.hadoop.fs.PathFilter;<a name="line.35"></a> |
| <span class="sourceLineNo">036</span>import org.apache.hadoop.hbase.HConstants;<a name="line.36"></a> |
| <span class="sourceLineNo">037</span>import org.apache.hadoop.hbase.ServerName;<a name="line.37"></a> |
| <span class="sourceLineNo">038</span>import org.apache.hadoop.hbase.regionserver.wal.AbstractFSWAL;<a name="line.38"></a> |
| <span class="sourceLineNo">039</span>import org.apache.hadoop.hbase.util.CommonFSUtils;<a name="line.39"></a> |
| <span class="sourceLineNo">040</span>import org.apache.hadoop.hbase.util.EnvironmentEdgeManager;<a name="line.40"></a> |
| <span class="sourceLineNo">041</span>import org.apache.hadoop.hbase.util.FSUtils;<a name="line.41"></a> |
| <span class="sourceLineNo">042</span>import org.apache.hadoop.hbase.wal.AbstractFSWALProvider;<a name="line.42"></a> |
| <span class="sourceLineNo">043</span>import org.apache.hadoop.hbase.wal.WALSplitter;<a name="line.43"></a> |
| <span class="sourceLineNo">044</span>import org.apache.yetus.audience.InterfaceAudience;<a name="line.44"></a> |
| <span class="sourceLineNo">045</span>import org.slf4j.Logger;<a name="line.45"></a> |
| <span class="sourceLineNo">046</span>import org.slf4j.LoggerFactory;<a name="line.46"></a> |
| <span class="sourceLineNo">047</span>import org.apache.hbase.thirdparty.com.google.common.annotations.VisibleForTesting;<a name="line.47"></a> |
| <span class="sourceLineNo">048</span><a name="line.48"></a> |
| <span class="sourceLineNo">049</span>/**<a name="line.49"></a> |
| <span class="sourceLineNo">050</span> * This class abstracts a bunch of operations the HMaster needs<a name="line.50"></a> |
| <span class="sourceLineNo">051</span> * when splitting log files e.g. finding log files, dirs etc.<a name="line.51"></a> |
| <span class="sourceLineNo">052</span> */<a name="line.52"></a> |
| <span class="sourceLineNo">053</span>@InterfaceAudience.Private<a name="line.53"></a> |
| <span class="sourceLineNo">054</span>public class MasterWalManager {<a name="line.54"></a> |
| <span class="sourceLineNo">055</span> private static final Logger LOG = LoggerFactory.getLogger(MasterWalManager.class);<a name="line.55"></a> |
| <span class="sourceLineNo">056</span><a name="line.56"></a> |
| <span class="sourceLineNo">057</span> final static PathFilter META_FILTER = new PathFilter() {<a name="line.57"></a> |
| <span class="sourceLineNo">058</span> @Override<a name="line.58"></a> |
| <span class="sourceLineNo">059</span> public boolean accept(Path p) {<a name="line.59"></a> |
| <span class="sourceLineNo">060</span> return AbstractFSWALProvider.isMetaFile(p);<a name="line.60"></a> |
| <span class="sourceLineNo">061</span> }<a name="line.61"></a> |
| <span class="sourceLineNo">062</span> };<a name="line.62"></a> |
| <span class="sourceLineNo">063</span><a name="line.63"></a> |
| <span class="sourceLineNo">064</span> @VisibleForTesting<a name="line.64"></a> |
| <span class="sourceLineNo">065</span> public final static PathFilter NON_META_FILTER = new PathFilter() {<a name="line.65"></a> |
| <span class="sourceLineNo">066</span> @Override<a name="line.66"></a> |
| <span class="sourceLineNo">067</span> public boolean accept(Path p) {<a name="line.67"></a> |
| <span class="sourceLineNo">068</span> return !AbstractFSWALProvider.isMetaFile(p);<a name="line.68"></a> |
| <span class="sourceLineNo">069</span> }<a name="line.69"></a> |
| <span class="sourceLineNo">070</span> };<a name="line.70"></a> |
| <span class="sourceLineNo">071</span><a name="line.71"></a> |
| <span class="sourceLineNo">072</span> // metrics for master<a name="line.72"></a> |
| <span class="sourceLineNo">073</span> // TODO: Rename it, since those metrics are split-manager related<a name="line.73"></a> |
| <span class="sourceLineNo">074</span> private final MetricsMasterFileSystem metricsMasterFilesystem = new MetricsMasterFileSystem();<a name="line.74"></a> |
| <span class="sourceLineNo">075</span><a name="line.75"></a> |
| <span class="sourceLineNo">076</span> // Keep around for convenience.<a name="line.76"></a> |
| <span class="sourceLineNo">077</span> private final MasterServices services;<a name="line.77"></a> |
| <span class="sourceLineNo">078</span> private final Configuration conf;<a name="line.78"></a> |
| <span class="sourceLineNo">079</span> private final FileSystem fs;<a name="line.79"></a> |
| <span class="sourceLineNo">080</span><a name="line.80"></a> |
| <span class="sourceLineNo">081</span> // The Path to the old logs dir<a name="line.81"></a> |
| <span class="sourceLineNo">082</span> private final Path oldLogDir;<a name="line.82"></a> |
| <span class="sourceLineNo">083</span> private final Path rootDir;<a name="line.83"></a> |
| <span class="sourceLineNo">084</span><a name="line.84"></a> |
| <span class="sourceLineNo">085</span> // create the split log lock<a name="line.85"></a> |
| <span class="sourceLineNo">086</span> private final Lock splitLogLock = new ReentrantLock();<a name="line.86"></a> |
| <span class="sourceLineNo">087</span> private final SplitLogManager splitLogManager;<a name="line.87"></a> |
| <span class="sourceLineNo">088</span><a name="line.88"></a> |
| <span class="sourceLineNo">089</span> // Is the fileystem ok?<a name="line.89"></a> |
| <span class="sourceLineNo">090</span> private volatile boolean fsOk = true;<a name="line.90"></a> |
| <span class="sourceLineNo">091</span><a name="line.91"></a> |
| <span class="sourceLineNo">092</span> public MasterWalManager(MasterServices services) throws IOException {<a name="line.92"></a> |
| <span class="sourceLineNo">093</span> this(services.getConfiguration(), services.getMasterFileSystem().getWALFileSystem(),<a name="line.93"></a> |
| <span class="sourceLineNo">094</span> services.getMasterFileSystem().getWALRootDir(), services);<a name="line.94"></a> |
| <span class="sourceLineNo">095</span> }<a name="line.95"></a> |
| <span class="sourceLineNo">096</span><a name="line.96"></a> |
| <span class="sourceLineNo">097</span> public MasterWalManager(Configuration conf, FileSystem fs, Path rootDir, MasterServices services)<a name="line.97"></a> |
| <span class="sourceLineNo">098</span> throws IOException {<a name="line.98"></a> |
| <span class="sourceLineNo">099</span> this.fs = fs;<a name="line.99"></a> |
| <span class="sourceLineNo">100</span> this.conf = conf;<a name="line.100"></a> |
| <span class="sourceLineNo">101</span> this.rootDir = rootDir;<a name="line.101"></a> |
| <span class="sourceLineNo">102</span> this.services = services;<a name="line.102"></a> |
| <span class="sourceLineNo">103</span> this.splitLogManager = new SplitLogManager(services, conf);<a name="line.103"></a> |
| <span class="sourceLineNo">104</span><a name="line.104"></a> |
| <span class="sourceLineNo">105</span> this.oldLogDir = new Path(rootDir, HConstants.HREGION_OLDLOGDIR_NAME);<a name="line.105"></a> |
| <span class="sourceLineNo">106</span> }<a name="line.106"></a> |
| <span class="sourceLineNo">107</span><a name="line.107"></a> |
| <span class="sourceLineNo">108</span> public void stop() {<a name="line.108"></a> |
| <span class="sourceLineNo">109</span> if (splitLogManager != null) {<a name="line.109"></a> |
| <span class="sourceLineNo">110</span> splitLogManager.stop();<a name="line.110"></a> |
| <span class="sourceLineNo">111</span> }<a name="line.111"></a> |
| <span class="sourceLineNo">112</span> }<a name="line.112"></a> |
| <span class="sourceLineNo">113</span><a name="line.113"></a> |
| <span class="sourceLineNo">114</span> @VisibleForTesting<a name="line.114"></a> |
| <span class="sourceLineNo">115</span> SplitLogManager getSplitLogManager() {<a name="line.115"></a> |
| <span class="sourceLineNo">116</span> return this.splitLogManager;<a name="line.116"></a> |
| <span class="sourceLineNo">117</span> }<a name="line.117"></a> |
| <span class="sourceLineNo">118</span><a name="line.118"></a> |
| <span class="sourceLineNo">119</span> /**<a name="line.119"></a> |
| <span class="sourceLineNo">120</span> * Get the directory where old logs go<a name="line.120"></a> |
| <span class="sourceLineNo">121</span> * @return the dir<a name="line.121"></a> |
| <span class="sourceLineNo">122</span> */<a name="line.122"></a> |
| <span class="sourceLineNo">123</span> Path getOldLogDir() {<a name="line.123"></a> |
| <span class="sourceLineNo">124</span> return this.oldLogDir;<a name="line.124"></a> |
| <span class="sourceLineNo">125</span> }<a name="line.125"></a> |
| <span class="sourceLineNo">126</span><a name="line.126"></a> |
| <span class="sourceLineNo">127</span> public FileSystem getFileSystem() {<a name="line.127"></a> |
| <span class="sourceLineNo">128</span> return this.fs;<a name="line.128"></a> |
| <span class="sourceLineNo">129</span> }<a name="line.129"></a> |
| <span class="sourceLineNo">130</span><a name="line.130"></a> |
| <span class="sourceLineNo">131</span> /**<a name="line.131"></a> |
| <span class="sourceLineNo">132</span> * Checks to see if the file system is still accessible.<a name="line.132"></a> |
| <span class="sourceLineNo">133</span> * If not, sets closed<a name="line.133"></a> |
| <span class="sourceLineNo">134</span> * @return false if file system is not available<a name="line.134"></a> |
| <span class="sourceLineNo">135</span> */<a name="line.135"></a> |
| <span class="sourceLineNo">136</span> private boolean checkFileSystem() {<a name="line.136"></a> |
| <span class="sourceLineNo">137</span> if (this.fsOk) {<a name="line.137"></a> |
| <span class="sourceLineNo">138</span> try {<a name="line.138"></a> |
| <span class="sourceLineNo">139</span> FSUtils.checkFileSystemAvailable(this.fs);<a name="line.139"></a> |
| <span class="sourceLineNo">140</span> FSUtils.checkDfsSafeMode(this.conf);<a name="line.140"></a> |
| <span class="sourceLineNo">141</span> } catch (IOException e) {<a name="line.141"></a> |
| <span class="sourceLineNo">142</span> services.abort("Shutting down HBase cluster: file system not available", e);<a name="line.142"></a> |
| <span class="sourceLineNo">143</span> this.fsOk = false;<a name="line.143"></a> |
| <span class="sourceLineNo">144</span> }<a name="line.144"></a> |
| <span class="sourceLineNo">145</span> }<a name="line.145"></a> |
| <span class="sourceLineNo">146</span> return this.fsOk;<a name="line.146"></a> |
| <span class="sourceLineNo">147</span> }<a name="line.147"></a> |
| <span class="sourceLineNo">148</span><a name="line.148"></a> |
| <span class="sourceLineNo">149</span> /**<a name="line.149"></a> |
| <span class="sourceLineNo">150</span> * Get Servernames which are currently splitting; paths have a '-splitting' suffix.<a name="line.150"></a> |
| <span class="sourceLineNo">151</span> * @return ServerName<a name="line.151"></a> |
| <span class="sourceLineNo">152</span> * @throws IOException IOException<a name="line.152"></a> |
| <span class="sourceLineNo">153</span> */<a name="line.153"></a> |
| <span class="sourceLineNo">154</span> public Set<ServerName> getSplittingServersFromWALDir() throws IOException {<a name="line.154"></a> |
| <span class="sourceLineNo">155</span> return getServerNamesFromWALDirPath(<a name="line.155"></a> |
| <span class="sourceLineNo">156</span> p -> p.getName().endsWith(AbstractFSWALProvider.SPLITTING_EXT));<a name="line.156"></a> |
| <span class="sourceLineNo">157</span> }<a name="line.157"></a> |
| <span class="sourceLineNo">158</span><a name="line.158"></a> |
| <span class="sourceLineNo">159</span> /**<a name="line.159"></a> |
| <span class="sourceLineNo">160</span> * Get Servernames that COULD BE 'alive'; excludes those that have a '-splitting' suffix as these<a name="line.160"></a> |
| <span class="sourceLineNo">161</span> * are already being split -- they cannot be 'alive'.<a name="line.161"></a> |
| <span class="sourceLineNo">162</span> * @return ServerName<a name="line.162"></a> |
| <span class="sourceLineNo">163</span> * @throws IOException IOException<a name="line.163"></a> |
| <span class="sourceLineNo">164</span> */<a name="line.164"></a> |
| <span class="sourceLineNo">165</span> public Set<ServerName> getLiveServersFromWALDir() throws IOException {<a name="line.165"></a> |
| <span class="sourceLineNo">166</span> return getServerNamesFromWALDirPath(<a name="line.166"></a> |
| <span class="sourceLineNo">167</span> p -> !p.getName().endsWith(AbstractFSWALProvider.SPLITTING_EXT));<a name="line.167"></a> |
| <span class="sourceLineNo">168</span> }<a name="line.168"></a> |
| <span class="sourceLineNo">169</span><a name="line.169"></a> |
| <span class="sourceLineNo">170</span> /**<a name="line.170"></a> |
| <span class="sourceLineNo">171</span> * @return listing of ServerNames found by parsing WAL directory paths in FS.<a name="line.171"></a> |
| <span class="sourceLineNo">172</span> */<a name="line.172"></a> |
| <span class="sourceLineNo">173</span> public Set<ServerName> getServerNamesFromWALDirPath(final PathFilter filter) throws IOException {<a name="line.173"></a> |
| <span class="sourceLineNo">174</span> FileStatus[] walDirForServerNames = getWALDirPaths(filter);<a name="line.174"></a> |
| <span class="sourceLineNo">175</span> return Stream.of(walDirForServerNames).map(s -> {<a name="line.175"></a> |
| <span class="sourceLineNo">176</span> ServerName serverName = AbstractFSWALProvider.getServerNameFromWALDirectoryName(s.getPath());<a name="line.176"></a> |
| <span class="sourceLineNo">177</span> if (serverName == null) {<a name="line.177"></a> |
| <span class="sourceLineNo">178</span> LOG.warn("Log folder {} doesn't look like its name includes a " +<a name="line.178"></a> |
| <span class="sourceLineNo">179</span> "region server name; leaving in place. If you see later errors about missing " +<a name="line.179"></a> |
| <span class="sourceLineNo">180</span> "write ahead logs they may be saved in this location.", s.getPath());<a name="line.180"></a> |
| <span class="sourceLineNo">181</span> return null;<a name="line.181"></a> |
| <span class="sourceLineNo">182</span> }<a name="line.182"></a> |
| <span class="sourceLineNo">183</span> return serverName;<a name="line.183"></a> |
| <span class="sourceLineNo">184</span> }).filter(s -> s != null).collect(Collectors.toSet());<a name="line.184"></a> |
| <span class="sourceLineNo">185</span> }<a name="line.185"></a> |
| <span class="sourceLineNo">186</span><a name="line.186"></a> |
| <span class="sourceLineNo">187</span> /**<a name="line.187"></a> |
| <span class="sourceLineNo">188</span> * @return List of all RegionServer WAL dirs; i.e. this.rootDir/HConstants.HREGION_LOGDIR_NAME.<a name="line.188"></a> |
| <span class="sourceLineNo">189</span> */<a name="line.189"></a> |
| <span class="sourceLineNo">190</span> public FileStatus[] getWALDirPaths(final PathFilter filter) throws IOException {<a name="line.190"></a> |
| <span class="sourceLineNo">191</span> Path walDirPath = new Path(CommonFSUtils.getWALRootDir(conf), HConstants.HREGION_LOGDIR_NAME);<a name="line.191"></a> |
| <span class="sourceLineNo">192</span> FileStatus[] walDirForServerNames = FSUtils.listStatus(fs, walDirPath, filter);<a name="line.192"></a> |
| <span class="sourceLineNo">193</span> return walDirForServerNames == null? new FileStatus[0]: walDirForServerNames;<a name="line.193"></a> |
| <span class="sourceLineNo">194</span> }<a name="line.194"></a> |
| <span class="sourceLineNo">195</span><a name="line.195"></a> |
| <span class="sourceLineNo">196</span> /**<a name="line.196"></a> |
| <span class="sourceLineNo">197</span> * Inspect the log directory to find dead servers which need recovery work<a name="line.197"></a> |
| <span class="sourceLineNo">198</span> * @return A set of ServerNames which aren't running but still have WAL files left in file system<a name="line.198"></a> |
| <span class="sourceLineNo">199</span> * @deprecated With proc-v2, we can record the crash server with procedure store, so do not need<a name="line.199"></a> |
| <span class="sourceLineNo">200</span> * to scan the wal directory to find out the splitting wal directory any more. Leave<a name="line.200"></a> |
| <span class="sourceLineNo">201</span> * it here only because {@code RecoverMetaProcedure}(which is also deprecated) uses<a name="line.201"></a> |
| <span class="sourceLineNo">202</span> * it.<a name="line.202"></a> |
| <span class="sourceLineNo">203</span> */<a name="line.203"></a> |
| <span class="sourceLineNo">204</span> @Deprecated<a name="line.204"></a> |
| <span class="sourceLineNo">205</span> public Set<ServerName> getFailedServersFromLogFolders() throws IOException {<a name="line.205"></a> |
| <span class="sourceLineNo">206</span> boolean retrySplitting = !conf.getBoolean("hbase.hlog.split.skip.errors",<a name="line.206"></a> |
| <span class="sourceLineNo">207</span> WALSplitter.SPLIT_SKIP_ERRORS_DEFAULT);<a name="line.207"></a> |
| <span class="sourceLineNo">208</span><a name="line.208"></a> |
| <span class="sourceLineNo">209</span> Set<ServerName> serverNames = new HashSet<>();<a name="line.209"></a> |
| <span class="sourceLineNo">210</span> Path logsDirPath = new Path(CommonFSUtils.getWALRootDir(conf), HConstants.HREGION_LOGDIR_NAME);<a name="line.210"></a> |
| <span class="sourceLineNo">211</span><a name="line.211"></a> |
| <span class="sourceLineNo">212</span> do {<a name="line.212"></a> |
| <span class="sourceLineNo">213</span> if (services.isStopped()) {<a name="line.213"></a> |
| <span class="sourceLineNo">214</span> LOG.warn("Master stopped while trying to get failed servers.");<a name="line.214"></a> |
| <span class="sourceLineNo">215</span> break;<a name="line.215"></a> |
| <span class="sourceLineNo">216</span> }<a name="line.216"></a> |
| <span class="sourceLineNo">217</span> try {<a name="line.217"></a> |
| <span class="sourceLineNo">218</span> if (!this.fs.exists(logsDirPath)) return serverNames;<a name="line.218"></a> |
| <span class="sourceLineNo">219</span> FileStatus[] logFolders = FSUtils.listStatus(this.fs, logsDirPath, null);<a name="line.219"></a> |
| <span class="sourceLineNo">220</span> // Get online servers after getting log folders to avoid log folder deletion of newly<a name="line.220"></a> |
| <span class="sourceLineNo">221</span> // checked in region servers . see HBASE-5916<a name="line.221"></a> |
| <span class="sourceLineNo">222</span> Set<ServerName> onlineServers = services.getServerManager().getOnlineServers().keySet();<a name="line.222"></a> |
| <span class="sourceLineNo">223</span><a name="line.223"></a> |
| <span class="sourceLineNo">224</span> if (logFolders == null || logFolders.length == 0) {<a name="line.224"></a> |
| <span class="sourceLineNo">225</span> LOG.debug("No log files to split, proceeding...");<a name="line.225"></a> |
| <span class="sourceLineNo">226</span> return serverNames;<a name="line.226"></a> |
| <span class="sourceLineNo">227</span> }<a name="line.227"></a> |
| <span class="sourceLineNo">228</span> for (FileStatus status : logFolders) {<a name="line.228"></a> |
| <span class="sourceLineNo">229</span> FileStatus[] curLogFiles = FSUtils.listStatus(this.fs, status.getPath(), null);<a name="line.229"></a> |
| <span class="sourceLineNo">230</span> if (curLogFiles == null || curLogFiles.length == 0) {<a name="line.230"></a> |
| <span class="sourceLineNo">231</span> // Empty log folder. No recovery needed<a name="line.231"></a> |
| <span class="sourceLineNo">232</span> continue;<a name="line.232"></a> |
| <span class="sourceLineNo">233</span> }<a name="line.233"></a> |
| <span class="sourceLineNo">234</span> final ServerName serverName = AbstractFSWALProvider.getServerNameFromWALDirectoryName(<a name="line.234"></a> |
| <span class="sourceLineNo">235</span> status.getPath());<a name="line.235"></a> |
| <span class="sourceLineNo">236</span> if (null == serverName) {<a name="line.236"></a> |
| <span class="sourceLineNo">237</span> LOG.warn("Log folder " + status.getPath() + " doesn't look like its name includes a " +<a name="line.237"></a> |
| <span class="sourceLineNo">238</span> "region server name; leaving in place. If you see later errors about missing " +<a name="line.238"></a> |
| <span class="sourceLineNo">239</span> "write ahead logs they may be saved in this location.");<a name="line.239"></a> |
| <span class="sourceLineNo">240</span> } else if (!onlineServers.contains(serverName)) {<a name="line.240"></a> |
| <span class="sourceLineNo">241</span> LOG.info("Log folder " + status.getPath() + " doesn't belong "<a name="line.241"></a> |
| <span class="sourceLineNo">242</span> + "to a known region server, splitting");<a name="line.242"></a> |
| <span class="sourceLineNo">243</span> serverNames.add(serverName);<a name="line.243"></a> |
| <span class="sourceLineNo">244</span> } else {<a name="line.244"></a> |
| <span class="sourceLineNo">245</span> LOG.info("Log folder " + status.getPath() + " belongs to an existing region server");<a name="line.245"></a> |
| <span class="sourceLineNo">246</span> }<a name="line.246"></a> |
| <span class="sourceLineNo">247</span> }<a name="line.247"></a> |
| <span class="sourceLineNo">248</span> retrySplitting = false;<a name="line.248"></a> |
| <span class="sourceLineNo">249</span> } catch (IOException ioe) {<a name="line.249"></a> |
| <span class="sourceLineNo">250</span> LOG.warn("Failed getting failed servers to be recovered.", ioe);<a name="line.250"></a> |
| <span class="sourceLineNo">251</span> if (!checkFileSystem()) {<a name="line.251"></a> |
| <span class="sourceLineNo">252</span> LOG.warn("Bad Filesystem, exiting");<a name="line.252"></a> |
| <span class="sourceLineNo">253</span> Runtime.getRuntime().halt(1);<a name="line.253"></a> |
| <span class="sourceLineNo">254</span> }<a name="line.254"></a> |
| <span class="sourceLineNo">255</span> try {<a name="line.255"></a> |
| <span class="sourceLineNo">256</span> if (retrySplitting) {<a name="line.256"></a> |
| <span class="sourceLineNo">257</span> Thread.sleep(conf.getInt("hbase.hlog.split.failure.retry.interval", 30 * 1000));<a name="line.257"></a> |
| <span class="sourceLineNo">258</span> }<a name="line.258"></a> |
| <span class="sourceLineNo">259</span> } catch (InterruptedException e) {<a name="line.259"></a> |
| <span class="sourceLineNo">260</span> LOG.warn("Interrupted, aborting since cannot return w/o splitting");<a name="line.260"></a> |
| <span class="sourceLineNo">261</span> Thread.currentThread().interrupt();<a name="line.261"></a> |
| <span class="sourceLineNo">262</span> retrySplitting = false;<a name="line.262"></a> |
| <span class="sourceLineNo">263</span> Runtime.getRuntime().halt(1);<a name="line.263"></a> |
| <span class="sourceLineNo">264</span> }<a name="line.264"></a> |
| <span class="sourceLineNo">265</span> }<a name="line.265"></a> |
| <span class="sourceLineNo">266</span> } while (retrySplitting);<a name="line.266"></a> |
| <span class="sourceLineNo">267</span><a name="line.267"></a> |
| <span class="sourceLineNo">268</span> return serverNames;<a name="line.268"></a> |
| <span class="sourceLineNo">269</span> }<a name="line.269"></a> |
| <span class="sourceLineNo">270</span><a name="line.270"></a> |
| <span class="sourceLineNo">271</span> public void splitLog(final ServerName serverName) throws IOException {<a name="line.271"></a> |
| <span class="sourceLineNo">272</span> splitLog(Collections.<ServerName>singleton(serverName));<a name="line.272"></a> |
| <span class="sourceLineNo">273</span> }<a name="line.273"></a> |
| <span class="sourceLineNo">274</span><a name="line.274"></a> |
| <span class="sourceLineNo">275</span> /**<a name="line.275"></a> |
| <span class="sourceLineNo">276</span> * Specialized method to handle the splitting for meta WAL<a name="line.276"></a> |
| <span class="sourceLineNo">277</span> * @param serverName logs belonging to this server will be split<a name="line.277"></a> |
| <span class="sourceLineNo">278</span> */<a name="line.278"></a> |
| <span class="sourceLineNo">279</span> public void splitMetaLog(final ServerName serverName) throws IOException {<a name="line.279"></a> |
| <span class="sourceLineNo">280</span> splitMetaLog(Collections.<ServerName>singleton(serverName));<a name="line.280"></a> |
| <span class="sourceLineNo">281</span> }<a name="line.281"></a> |
| <span class="sourceLineNo">282</span><a name="line.282"></a> |
| <span class="sourceLineNo">283</span> /**<a name="line.283"></a> |
| <span class="sourceLineNo">284</span> * Specialized method to handle the splitting for meta WAL<a name="line.284"></a> |
| <span class="sourceLineNo">285</span> * @param serverNames logs belonging to these servers will be split<a name="line.285"></a> |
| <span class="sourceLineNo">286</span> */<a name="line.286"></a> |
| <span class="sourceLineNo">287</span> public void splitMetaLog(final Set<ServerName> serverNames) throws IOException {<a name="line.287"></a> |
| <span class="sourceLineNo">288</span> splitLog(serverNames, META_FILTER);<a name="line.288"></a> |
| <span class="sourceLineNo">289</span> }<a name="line.289"></a> |
| <span class="sourceLineNo">290</span><a name="line.290"></a> |
| <span class="sourceLineNo">291</span> @edu.umd.cs.findbugs.annotations.SuppressWarnings(value="UL_UNRELEASED_LOCK", justification=<a name="line.291"></a> |
| <span class="sourceLineNo">292</span> "We only release this lock when we set it. Updates to code that uses it should verify use " +<a name="line.292"></a> |
| <span class="sourceLineNo">293</span> "of the guard boolean.")<a name="line.293"></a> |
| <span class="sourceLineNo">294</span> List<Path> getLogDirs(final Set<ServerName> serverNames) throws IOException {<a name="line.294"></a> |
| <span class="sourceLineNo">295</span> List<Path> logDirs = new ArrayList<>();<a name="line.295"></a> |
| <span class="sourceLineNo">296</span> boolean needReleaseLock = false;<a name="line.296"></a> |
| <span class="sourceLineNo">297</span> if (!this.services.isInitialized()) {<a name="line.297"></a> |
| <span class="sourceLineNo">298</span> // during master initialization, we could have multiple places splitting a same wal<a name="line.298"></a> |
| <span class="sourceLineNo">299</span> // XXX: Does this still exist after we move to proc-v2?<a name="line.299"></a> |
| <span class="sourceLineNo">300</span> this.splitLogLock.lock();<a name="line.300"></a> |
| <span class="sourceLineNo">301</span> needReleaseLock = true;<a name="line.301"></a> |
| <span class="sourceLineNo">302</span> }<a name="line.302"></a> |
| <span class="sourceLineNo">303</span> try {<a name="line.303"></a> |
| <span class="sourceLineNo">304</span> for (ServerName serverName : serverNames) {<a name="line.304"></a> |
| <span class="sourceLineNo">305</span> Path logDir = new Path(this.rootDir,<a name="line.305"></a> |
| <span class="sourceLineNo">306</span> AbstractFSWALProvider.getWALDirectoryName(serverName.toString()));<a name="line.306"></a> |
| <span class="sourceLineNo">307</span> Path splitDir = logDir.suffix(AbstractFSWALProvider.SPLITTING_EXT);<a name="line.307"></a> |
| <span class="sourceLineNo">308</span> // Rename the directory so a rogue RS doesn't create more WALs<a name="line.308"></a> |
| <span class="sourceLineNo">309</span> if (fs.exists(logDir)) {<a name="line.309"></a> |
| <span class="sourceLineNo">310</span> if (!this.fs.rename(logDir, splitDir)) {<a name="line.310"></a> |
| <span class="sourceLineNo">311</span> throw new IOException("Failed fs.rename for log split: " + logDir);<a name="line.311"></a> |
| <span class="sourceLineNo">312</span> }<a name="line.312"></a> |
| <span class="sourceLineNo">313</span> logDir = splitDir;<a name="line.313"></a> |
| <span class="sourceLineNo">314</span> LOG.debug("Renamed region directory: " + splitDir);<a name="line.314"></a> |
| <span class="sourceLineNo">315</span> } else if (!fs.exists(splitDir)) {<a name="line.315"></a> |
| <span class="sourceLineNo">316</span> LOG.info("Log dir for server " + serverName + " does not exist");<a name="line.316"></a> |
| <span class="sourceLineNo">317</span> continue;<a name="line.317"></a> |
| <span class="sourceLineNo">318</span> }<a name="line.318"></a> |
| <span class="sourceLineNo">319</span> logDirs.add(splitDir);<a name="line.319"></a> |
| <span class="sourceLineNo">320</span> }<a name="line.320"></a> |
| <span class="sourceLineNo">321</span> } catch (IOException ioe) {<a name="line.321"></a> |
| <span class="sourceLineNo">322</span> if (!checkFileSystem()) {<a name="line.322"></a> |
| <span class="sourceLineNo">323</span> this.services.abort("Aborting due to filesystem unavailable", ioe);<a name="line.323"></a> |
| <span class="sourceLineNo">324</span> throw ioe;<a name="line.324"></a> |
| <span class="sourceLineNo">325</span> }<a name="line.325"></a> |
| <span class="sourceLineNo">326</span> } finally {<a name="line.326"></a> |
| <span class="sourceLineNo">327</span> if (needReleaseLock) {<a name="line.327"></a> |
| <span class="sourceLineNo">328</span> this.splitLogLock.unlock();<a name="line.328"></a> |
| <span class="sourceLineNo">329</span> }<a name="line.329"></a> |
| <span class="sourceLineNo">330</span> }<a name="line.330"></a> |
| <span class="sourceLineNo">331</span> return logDirs;<a name="line.331"></a> |
| <span class="sourceLineNo">332</span> }<a name="line.332"></a> |
| <span class="sourceLineNo">333</span><a name="line.333"></a> |
| <span class="sourceLineNo">334</span> public void splitLog(final Set<ServerName> serverNames) throws IOException {<a name="line.334"></a> |
| <span class="sourceLineNo">335</span> splitLog(serverNames, NON_META_FILTER);<a name="line.335"></a> |
| <span class="sourceLineNo">336</span> }<a name="line.336"></a> |
| <span class="sourceLineNo">337</span><a name="line.337"></a> |
| <span class="sourceLineNo">338</span> /**<a name="line.338"></a> |
| <span class="sourceLineNo">339</span> * This method is the base split method that splits WAL files matching a filter. Callers should<a name="line.339"></a> |
| <span class="sourceLineNo">340</span> * pass the appropriate filter for meta and non-meta WALs.<a name="line.340"></a> |
| <span class="sourceLineNo">341</span> * @param serverNames logs belonging to these servers will be split; this will rename the log<a name="line.341"></a> |
| <span class="sourceLineNo">342</span> * directory out from under a soft-failed server<a name="line.342"></a> |
| <span class="sourceLineNo">343</span> */<a name="line.343"></a> |
| <span class="sourceLineNo">344</span> public void splitLog(final Set<ServerName> serverNames, PathFilter filter) throws IOException {<a name="line.344"></a> |
| <span class="sourceLineNo">345</span> long splitTime = 0, splitLogSize = 0;<a name="line.345"></a> |
| <span class="sourceLineNo">346</span> List<Path> logDirs = getLogDirs(serverNames);<a name="line.346"></a> |
| <span class="sourceLineNo">347</span><a name="line.347"></a> |
| <span class="sourceLineNo">348</span> splitLogManager.handleDeadWorkers(serverNames);<a name="line.348"></a> |
| <span class="sourceLineNo">349</span> splitTime = EnvironmentEdgeManager.currentTime();<a name="line.349"></a> |
| <span class="sourceLineNo">350</span> splitLogSize = splitLogManager.splitLogDistributed(serverNames, logDirs, filter);<a name="line.350"></a> |
| <span class="sourceLineNo">351</span> splitTime = EnvironmentEdgeManager.currentTime() - splitTime;<a name="line.351"></a> |
| <span class="sourceLineNo">352</span><a name="line.352"></a> |
| <span class="sourceLineNo">353</span> if (this.metricsMasterFilesystem != null) {<a name="line.353"></a> |
| <span class="sourceLineNo">354</span> if (filter == META_FILTER) {<a name="line.354"></a> |
| <span class="sourceLineNo">355</span> this.metricsMasterFilesystem.addMetaWALSplit(splitTime, splitLogSize);<a name="line.355"></a> |
| <span class="sourceLineNo">356</span> } else {<a name="line.356"></a> |
| <span class="sourceLineNo">357</span> this.metricsMasterFilesystem.addSplit(splitTime, splitLogSize);<a name="line.357"></a> |
| <span class="sourceLineNo">358</span> }<a name="line.358"></a> |
| <span class="sourceLineNo">359</span> }<a name="line.359"></a> |
| <span class="sourceLineNo">360</span> }<a name="line.360"></a> |
| <span class="sourceLineNo">361</span><a name="line.361"></a> |
| <span class="sourceLineNo">362</span> /**<a name="line.362"></a> |
| <span class="sourceLineNo">363</span> * For meta region open and closed normally on a server, it may leave some meta<a name="line.363"></a> |
| <span class="sourceLineNo">364</span> * WAL in the server's wal dir. Since meta region is no long on this server,<a name="line.364"></a> |
| <span class="sourceLineNo">365</span> * The SCP won't split those meta wals, just leaving them there. So deleting<a name="line.365"></a> |
| <span class="sourceLineNo">366</span> * the wal dir will fail since the dir is not empty. Actually We can safely achive those<a name="line.366"></a> |
| <span class="sourceLineNo">367</span> * meta log and Archiving the meta log and delete the dir.<a name="line.367"></a> |
| <span class="sourceLineNo">368</span> * @param serverName the server to archive meta log<a name="line.368"></a> |
| <span class="sourceLineNo">369</span> */<a name="line.369"></a> |
| <span class="sourceLineNo">370</span> public void archiveMetaLog(final ServerName serverName) {<a name="line.370"></a> |
| <span class="sourceLineNo">371</span> try {<a name="line.371"></a> |
| <span class="sourceLineNo">372</span> Path logDir = new Path(this.rootDir,<a name="line.372"></a> |
| <span class="sourceLineNo">373</span> AbstractFSWALProvider.getWALDirectoryName(serverName.toString()));<a name="line.373"></a> |
| <span class="sourceLineNo">374</span> Path splitDir = logDir.suffix(AbstractFSWALProvider.SPLITTING_EXT);<a name="line.374"></a> |
| <span class="sourceLineNo">375</span> if (fs.exists(splitDir)) {<a name="line.375"></a> |
| <span class="sourceLineNo">376</span> FileStatus[] logfiles = FSUtils.listStatus(fs, splitDir, META_FILTER);<a name="line.376"></a> |
| <span class="sourceLineNo">377</span> if (logfiles != null) {<a name="line.377"></a> |
| <span class="sourceLineNo">378</span> for (FileStatus status : logfiles) {<a name="line.378"></a> |
| <span class="sourceLineNo">379</span> if (!status.isDir()) {<a name="line.379"></a> |
| <span class="sourceLineNo">380</span> Path newPath = AbstractFSWAL.getWALArchivePath(this.oldLogDir,<a name="line.380"></a> |
| <span class="sourceLineNo">381</span> status.getPath());<a name="line.381"></a> |
| <span class="sourceLineNo">382</span> if (!FSUtils.renameAndSetModifyTime(fs, status.getPath(), newPath)) {<a name="line.382"></a> |
| <span class="sourceLineNo">383</span> LOG.warn("Unable to move " + status.getPath() + " to " + newPath);<a name="line.383"></a> |
| <span class="sourceLineNo">384</span> } else {<a name="line.384"></a> |
| <span class="sourceLineNo">385</span> LOG.debug("Archived meta log " + status.getPath() + " to " + newPath);<a name="line.385"></a> |
| <span class="sourceLineNo">386</span> }<a name="line.386"></a> |
| <span class="sourceLineNo">387</span> }<a name="line.387"></a> |
| <span class="sourceLineNo">388</span> }<a name="line.388"></a> |
| <span class="sourceLineNo">389</span> }<a name="line.389"></a> |
| <span class="sourceLineNo">390</span> if (!fs.delete(splitDir, false)) {<a name="line.390"></a> |
| <span class="sourceLineNo">391</span> LOG.warn("Unable to delete log dir. Ignoring. " + splitDir);<a name="line.391"></a> |
| <span class="sourceLineNo">392</span> }<a name="line.392"></a> |
| <span class="sourceLineNo">393</span> }<a name="line.393"></a> |
| <span class="sourceLineNo">394</span> } catch (IOException ie) {<a name="line.394"></a> |
| <span class="sourceLineNo">395</span> LOG.warn("Failed archiving meta log for server " + serverName, ie);<a name="line.395"></a> |
| <span class="sourceLineNo">396</span> }<a name="line.396"></a> |
| <span class="sourceLineNo">397</span> }<a name="line.397"></a> |
| <span class="sourceLineNo">398</span><a name="line.398"></a> |
| <span class="sourceLineNo">399</span><a name="line.399"></a> |
| <span class="sourceLineNo">400</span>}<a name="line.400"></a> |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| </pre> |
| </div> |
| </body> |
| </html> |