blob: e5f9df546c616819992fcc9f6b7a717feec4e082 [file] [log] [blame]
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
<html lang="en">
<head>
<title>Source code</title>
<link rel="stylesheet" type="text/css" href="../../../../../../stylesheet.css" title="Style">
</head>
<body>
<div class="sourceContainer">
<pre><span class="sourceLineNo">001</span>/**<a name="line.1"></a>
<span class="sourceLineNo">002</span> * Licensed to the Apache Software Foundation (ASF) under one<a name="line.2"></a>
<span class="sourceLineNo">003</span> * or more contributor license agreements. See the NOTICE file<a name="line.3"></a>
<span class="sourceLineNo">004</span> * distributed with this work for additional information<a name="line.4"></a>
<span class="sourceLineNo">005</span> * regarding copyright ownership. The ASF licenses this file<a name="line.5"></a>
<span class="sourceLineNo">006</span> * to you under the Apache License, Version 2.0 (the<a name="line.6"></a>
<span class="sourceLineNo">007</span> * "License"); you may not use this file except in compliance<a name="line.7"></a>
<span class="sourceLineNo">008</span> * with the License. You may obtain a copy of the License at<a name="line.8"></a>
<span class="sourceLineNo">009</span> *<a name="line.9"></a>
<span class="sourceLineNo">010</span> * http://www.apache.org/licenses/LICENSE-2.0<a name="line.10"></a>
<span class="sourceLineNo">011</span> *<a name="line.11"></a>
<span class="sourceLineNo">012</span> * Unless required by applicable law or agreed to in writing, software<a name="line.12"></a>
<span class="sourceLineNo">013</span> * distributed under the License is distributed on an "AS IS" BASIS,<a name="line.13"></a>
<span class="sourceLineNo">014</span> * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.<a name="line.14"></a>
<span class="sourceLineNo">015</span> * See the License for the specific language governing permissions and<a name="line.15"></a>
<span class="sourceLineNo">016</span> * limitations under the License.<a name="line.16"></a>
<span class="sourceLineNo">017</span> */<a name="line.17"></a>
<span class="sourceLineNo">018</span>package org.apache.hadoop.hbase.procedure;<a name="line.18"></a>
<span class="sourceLineNo">019</span><a name="line.19"></a>
<span class="sourceLineNo">020</span>import java.io.IOException;<a name="line.20"></a>
<span class="sourceLineNo">021</span>import java.util.ArrayList;<a name="line.21"></a>
<span class="sourceLineNo">022</span>import java.util.HashMap;<a name="line.22"></a>
<span class="sourceLineNo">023</span>import java.util.List;<a name="line.23"></a>
<span class="sourceLineNo">024</span>import java.util.concurrent.Callable;<a name="line.24"></a>
<span class="sourceLineNo">025</span>import java.util.concurrent.CountDownLatch;<a name="line.25"></a>
<span class="sourceLineNo">026</span>import java.util.concurrent.TimeUnit;<a name="line.26"></a>
<span class="sourceLineNo">027</span><a name="line.27"></a>
<span class="sourceLineNo">028</span>import org.apache.yetus.audience.InterfaceAudience;<a name="line.28"></a>
<span class="sourceLineNo">029</span>import org.slf4j.Logger;<a name="line.29"></a>
<span class="sourceLineNo">030</span>import org.slf4j.LoggerFactory;<a name="line.30"></a>
<span class="sourceLineNo">031</span>import org.apache.hadoop.hbase.errorhandling.ForeignException;<a name="line.31"></a>
<span class="sourceLineNo">032</span>import org.apache.hadoop.hbase.errorhandling.ForeignExceptionDispatcher;<a name="line.32"></a>
<span class="sourceLineNo">033</span>import org.apache.hadoop.hbase.errorhandling.ForeignExceptionListener;<a name="line.33"></a>
<span class="sourceLineNo">034</span>import org.apache.hadoop.hbase.errorhandling.ForeignExceptionSnare;<a name="line.34"></a>
<span class="sourceLineNo">035</span>import org.apache.hadoop.hbase.errorhandling.TimeoutExceptionInjector;<a name="line.35"></a>
<span class="sourceLineNo">036</span><a name="line.36"></a>
<span class="sourceLineNo">037</span>import org.apache.hbase.thirdparty.com.google.common.collect.Lists;<a name="line.37"></a>
<span class="sourceLineNo">038</span><a name="line.38"></a>
<span class="sourceLineNo">039</span>/**<a name="line.39"></a>
<span class="sourceLineNo">040</span> * A globally-barriered distributed procedure. This class encapsulates state and methods for<a name="line.40"></a>
<span class="sourceLineNo">041</span> * tracking and managing a distributed procedure, as well as aborting if any member encounters<a name="line.41"></a>
<span class="sourceLineNo">042</span> * a problem or if a cancellation is requested.<a name="line.42"></a>
<span class="sourceLineNo">043</span> * &lt;p&gt;<a name="line.43"></a>
<span class="sourceLineNo">044</span> * All procedures first attempt to reach a barrier point with the {@link #sendGlobalBarrierStart()}<a name="line.44"></a>
<span class="sourceLineNo">045</span> * method. The procedure contacts all members and waits for all subprocedures to execute<a name="line.45"></a>
<span class="sourceLineNo">046</span> * {@link Subprocedure#acquireBarrier} to acquire its local piece of the global barrier and then<a name="line.46"></a>
<span class="sourceLineNo">047</span> * send acquisition info back to the coordinator. If all acquisitions at subprocedures succeed,<a name="line.47"></a>
<span class="sourceLineNo">048</span> * the coordinator then will call {@link #sendGlobalBarrierReached()}. This notifies members to<a name="line.48"></a>
<span class="sourceLineNo">049</span> * execute the {@link Subprocedure#insideBarrier()} method. The procedure is blocked until all<a name="line.49"></a>
<span class="sourceLineNo">050</span> * {@link Subprocedure#insideBarrier} executions complete at the members. When<a name="line.50"></a>
<span class="sourceLineNo">051</span> * {@link Subprocedure#insideBarrier} completes at each member, the member sends notification to<a name="line.51"></a>
<span class="sourceLineNo">052</span> * the coordinator. Once all members complete, the coordinator calls<a name="line.52"></a>
<span class="sourceLineNo">053</span> * {@link #sendGlobalBarrierComplete()}.<a name="line.53"></a>
<span class="sourceLineNo">054</span> * &lt;p&gt;<a name="line.54"></a>
<span class="sourceLineNo">055</span> * If errors are encountered remotely, they are forwarded to the coordinator, and<a name="line.55"></a>
<span class="sourceLineNo">056</span> * {@link Subprocedure#cleanup(Exception)} is called.<a name="line.56"></a>
<span class="sourceLineNo">057</span> * &lt;p&gt;<a name="line.57"></a>
<span class="sourceLineNo">058</span> * Each Procedure and each Subprocedure enforces a time limit on the execution time. If the time<a name="line.58"></a>
<span class="sourceLineNo">059</span> * limit expires before the procedure completes the {@link TimeoutExceptionInjector} will trigger<a name="line.59"></a>
<span class="sourceLineNo">060</span> * an {@link ForeignException} to abort the procedure. This is particularly useful for situations<a name="line.60"></a>
<span class="sourceLineNo">061</span> * when running a distributed {@link Subprocedure} so participants can avoid blocking for extreme<a name="line.61"></a>
<span class="sourceLineNo">062</span> * amounts of time if one of the participants fails or takes a really long time (e.g. GC pause).<a name="line.62"></a>
<span class="sourceLineNo">063</span> * &lt;p&gt;<a name="line.63"></a>
<span class="sourceLineNo">064</span> * Users should generally not directly create or subclass instances of this. They are created<a name="line.64"></a>
<span class="sourceLineNo">065</span> * for them implicitly via {@link ProcedureCoordinator#startProcedure(ForeignExceptionDispatcher,<a name="line.65"></a>
<span class="sourceLineNo">066</span> * String, byte[], List)}}<a name="line.66"></a>
<span class="sourceLineNo">067</span> */<a name="line.67"></a>
<span class="sourceLineNo">068</span>@InterfaceAudience.Private<a name="line.68"></a>
<span class="sourceLineNo">069</span>public class Procedure implements Callable&lt;Void&gt;, ForeignExceptionListener {<a name="line.69"></a>
<span class="sourceLineNo">070</span> private static final Logger LOG = LoggerFactory.getLogger(Procedure.class);<a name="line.70"></a>
<span class="sourceLineNo">071</span><a name="line.71"></a>
<span class="sourceLineNo">072</span> //<a name="line.72"></a>
<span class="sourceLineNo">073</span> // Arguments and naming<a name="line.73"></a>
<span class="sourceLineNo">074</span> //<a name="line.74"></a>
<span class="sourceLineNo">075</span><a name="line.75"></a>
<span class="sourceLineNo">076</span> // Name of the procedure<a name="line.76"></a>
<span class="sourceLineNo">077</span> final private String procName;<a name="line.77"></a>
<span class="sourceLineNo">078</span> // Arguments for this procedure execution<a name="line.78"></a>
<span class="sourceLineNo">079</span> final private byte[] args;<a name="line.79"></a>
<span class="sourceLineNo">080</span><a name="line.80"></a>
<span class="sourceLineNo">081</span> //<a name="line.81"></a>
<span class="sourceLineNo">082</span> // Execution State<a name="line.82"></a>
<span class="sourceLineNo">083</span> //<a name="line.83"></a>
<span class="sourceLineNo">084</span> /** latch for waiting until all members have acquire in barrier state */<a name="line.84"></a>
<span class="sourceLineNo">085</span> final CountDownLatch acquiredBarrierLatch;<a name="line.85"></a>
<span class="sourceLineNo">086</span> /** latch for waiting until all members have executed and released their in barrier state */<a name="line.86"></a>
<span class="sourceLineNo">087</span> final CountDownLatch releasedBarrierLatch;<a name="line.87"></a>
<span class="sourceLineNo">088</span> /** latch for waiting until a procedure has completed */<a name="line.88"></a>
<span class="sourceLineNo">089</span> final CountDownLatch completedLatch;<a name="line.89"></a>
<span class="sourceLineNo">090</span> /** monitor to check for errors */<a name="line.90"></a>
<span class="sourceLineNo">091</span> private final ForeignExceptionDispatcher monitor;<a name="line.91"></a>
<span class="sourceLineNo">092</span><a name="line.92"></a>
<span class="sourceLineNo">093</span> //<a name="line.93"></a>
<span class="sourceLineNo">094</span> // Execution Timeout Handling.<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> /** frequency to check for errors (ms) */<a name="line.97"></a>
<span class="sourceLineNo">098</span> protected final long wakeFrequency;<a name="line.98"></a>
<span class="sourceLineNo">099</span> protected final TimeoutExceptionInjector timeoutInjector;<a name="line.99"></a>
<span class="sourceLineNo">100</span><a name="line.100"></a>
<span class="sourceLineNo">101</span> //<a name="line.101"></a>
<span class="sourceLineNo">102</span> // Members' and Coordinator's state<a name="line.102"></a>
<span class="sourceLineNo">103</span> //<a name="line.103"></a>
<span class="sourceLineNo">104</span><a name="line.104"></a>
<span class="sourceLineNo">105</span> /** lock to prevent nodes from acquiring and then releasing before we can track them */<a name="line.105"></a>
<span class="sourceLineNo">106</span> private final Object joinBarrierLock = new Object();<a name="line.106"></a>
<span class="sourceLineNo">107</span> private final List&lt;String&gt; acquiringMembers;<a name="line.107"></a>
<span class="sourceLineNo">108</span> private final List&lt;String&gt; inBarrierMembers;<a name="line.108"></a>
<span class="sourceLineNo">109</span> private final HashMap&lt;String, byte[]&gt; dataFromFinishedMembers;<a name="line.109"></a>
<span class="sourceLineNo">110</span> private ProcedureCoordinator coord;<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> * Creates a procedure. (FOR TESTING)<a name="line.113"></a>
<span class="sourceLineNo">114</span> *<a name="line.114"></a>
<span class="sourceLineNo">115</span> * {@link Procedure} state to be run by a {@link ProcedureCoordinator}.<a name="line.115"></a>
<span class="sourceLineNo">116</span> * @param coord coordinator to call back to for general errors (e.g.<a name="line.116"></a>
<span class="sourceLineNo">117</span> * {@link ProcedureCoordinator#rpcConnectionFailure(String, IOException)}).<a name="line.117"></a>
<span class="sourceLineNo">118</span> * @param monitor error monitor to check for external errors<a name="line.118"></a>
<span class="sourceLineNo">119</span> * @param wakeFreq frequency to check for errors while waiting<a name="line.119"></a>
<span class="sourceLineNo">120</span> * @param timeout amount of time to allow the procedure to run before cancelling<a name="line.120"></a>
<span class="sourceLineNo">121</span> * @param procName name of the procedure instance<a name="line.121"></a>
<span class="sourceLineNo">122</span> * @param args argument data associated with the procedure instance<a name="line.122"></a>
<span class="sourceLineNo">123</span> * @param expectedMembers names of the expected members<a name="line.123"></a>
<span class="sourceLineNo">124</span> */<a name="line.124"></a>
<span class="sourceLineNo">125</span> public Procedure(ProcedureCoordinator coord, ForeignExceptionDispatcher monitor, long wakeFreq,<a name="line.125"></a>
<span class="sourceLineNo">126</span> long timeout, String procName, byte[] args, List&lt;String&gt; expectedMembers) {<a name="line.126"></a>
<span class="sourceLineNo">127</span> this.coord = coord;<a name="line.127"></a>
<span class="sourceLineNo">128</span> this.acquiringMembers = new ArrayList&lt;&gt;(expectedMembers);<a name="line.128"></a>
<span class="sourceLineNo">129</span> this.inBarrierMembers = new ArrayList&lt;&gt;(acquiringMembers.size());<a name="line.129"></a>
<span class="sourceLineNo">130</span> this.dataFromFinishedMembers = new HashMap&lt;&gt;();<a name="line.130"></a>
<span class="sourceLineNo">131</span> this.procName = procName;<a name="line.131"></a>
<span class="sourceLineNo">132</span> this.args = args;<a name="line.132"></a>
<span class="sourceLineNo">133</span> this.monitor = monitor;<a name="line.133"></a>
<span class="sourceLineNo">134</span> this.wakeFrequency = wakeFreq;<a name="line.134"></a>
<span class="sourceLineNo">135</span><a name="line.135"></a>
<span class="sourceLineNo">136</span> int count = expectedMembers.size();<a name="line.136"></a>
<span class="sourceLineNo">137</span> this.acquiredBarrierLatch = new CountDownLatch(count);<a name="line.137"></a>
<span class="sourceLineNo">138</span> this.releasedBarrierLatch = new CountDownLatch(count);<a name="line.138"></a>
<span class="sourceLineNo">139</span> this.completedLatch = new CountDownLatch(1);<a name="line.139"></a>
<span class="sourceLineNo">140</span> this.timeoutInjector = new TimeoutExceptionInjector(monitor, timeout);<a name="line.140"></a>
<span class="sourceLineNo">141</span> }<a name="line.141"></a>
<span class="sourceLineNo">142</span><a name="line.142"></a>
<span class="sourceLineNo">143</span> /**<a name="line.143"></a>
<span class="sourceLineNo">144</span> * Create a procedure.<a name="line.144"></a>
<span class="sourceLineNo">145</span> *<a name="line.145"></a>
<span class="sourceLineNo">146</span> * Users should generally not directly create instances of this. They are created them<a name="line.146"></a>
<span class="sourceLineNo">147</span> * implicitly via {@link ProcedureCoordinator#createProcedure(ForeignExceptionDispatcher,<a name="line.147"></a>
<span class="sourceLineNo">148</span> * String, byte[], List)}}<a name="line.148"></a>
<span class="sourceLineNo">149</span> *<a name="line.149"></a>
<span class="sourceLineNo">150</span> * @param coord coordinator to call back to for general errors (e.g.<a name="line.150"></a>
<span class="sourceLineNo">151</span> * {@link ProcedureCoordinator#rpcConnectionFailure(String, IOException)}).<a name="line.151"></a>
<span class="sourceLineNo">152</span> * @param wakeFreq frequency to check for errors while waiting<a name="line.152"></a>
<span class="sourceLineNo">153</span> * @param timeout amount of time to allow the procedure to run before cancelling<a name="line.153"></a>
<span class="sourceLineNo">154</span> * @param procName name of the procedure instance<a name="line.154"></a>
<span class="sourceLineNo">155</span> * @param args argument data associated with the procedure instance<a name="line.155"></a>
<span class="sourceLineNo">156</span> * @param expectedMembers names of the expected members<a name="line.156"></a>
<span class="sourceLineNo">157</span> */<a name="line.157"></a>
<span class="sourceLineNo">158</span> public Procedure(ProcedureCoordinator coord, long wakeFreq, long timeout,<a name="line.158"></a>
<span class="sourceLineNo">159</span> String procName, byte[] args, List&lt;String&gt; expectedMembers) {<a name="line.159"></a>
<span class="sourceLineNo">160</span> this(coord, new ForeignExceptionDispatcher(), wakeFreq, timeout, procName, args,<a name="line.160"></a>
<span class="sourceLineNo">161</span> expectedMembers);<a name="line.161"></a>
<span class="sourceLineNo">162</span> }<a name="line.162"></a>
<span class="sourceLineNo">163</span><a name="line.163"></a>
<span class="sourceLineNo">164</span> public String getName() {<a name="line.164"></a>
<span class="sourceLineNo">165</span> return procName;<a name="line.165"></a>
<span class="sourceLineNo">166</span> }<a name="line.166"></a>
<span class="sourceLineNo">167</span><a name="line.167"></a>
<span class="sourceLineNo">168</span> /**<a name="line.168"></a>
<span class="sourceLineNo">169</span> * @return String of the procedure members both trying to enter the barrier and already in barrier<a name="line.169"></a>
<span class="sourceLineNo">170</span> */<a name="line.170"></a>
<span class="sourceLineNo">171</span> public String getStatus() {<a name="line.171"></a>
<span class="sourceLineNo">172</span> String waiting, done;<a name="line.172"></a>
<span class="sourceLineNo">173</span> synchronized (joinBarrierLock) {<a name="line.173"></a>
<span class="sourceLineNo">174</span> waiting = acquiringMembers.toString();<a name="line.174"></a>
<span class="sourceLineNo">175</span> done = inBarrierMembers.toString();<a name="line.175"></a>
<span class="sourceLineNo">176</span> }<a name="line.176"></a>
<span class="sourceLineNo">177</span> return "Procedure " + procName + " { waiting=" + waiting + " done="+ done + " }";<a name="line.177"></a>
<span class="sourceLineNo">178</span> }<a name="line.178"></a>
<span class="sourceLineNo">179</span><a name="line.179"></a>
<span class="sourceLineNo">180</span> /**<a name="line.180"></a>
<span class="sourceLineNo">181</span> * Get the ForeignExceptionDispatcher<a name="line.181"></a>
<span class="sourceLineNo">182</span> * @return the Procedure's monitor.<a name="line.182"></a>
<span class="sourceLineNo">183</span> */<a name="line.183"></a>
<span class="sourceLineNo">184</span> public ForeignExceptionDispatcher getErrorMonitor() {<a name="line.184"></a>
<span class="sourceLineNo">185</span> return monitor;<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> /**<a name="line.188"></a>
<span class="sourceLineNo">189</span> * This call is the main execution thread of the barriered procedure. It sends messages and<a name="line.189"></a>
<span class="sourceLineNo">190</span> * essentially blocks until all procedure members acquire or later complete but periodically<a name="line.190"></a>
<span class="sourceLineNo">191</span> * checks for foreign exceptions.<a name="line.191"></a>
<span class="sourceLineNo">192</span> */<a name="line.192"></a>
<span class="sourceLineNo">193</span> @Override<a name="line.193"></a>
<span class="sourceLineNo">194</span> @SuppressWarnings("finally")<a name="line.194"></a>
<span class="sourceLineNo">195</span> final public Void call() {<a name="line.195"></a>
<span class="sourceLineNo">196</span> LOG.info("Starting procedure '" + procName + "'");<a name="line.196"></a>
<span class="sourceLineNo">197</span> // start the timer<a name="line.197"></a>
<span class="sourceLineNo">198</span> timeoutInjector.start();<a name="line.198"></a>
<span class="sourceLineNo">199</span><a name="line.199"></a>
<span class="sourceLineNo">200</span> // run the procedure<a name="line.200"></a>
<span class="sourceLineNo">201</span> try {<a name="line.201"></a>
<span class="sourceLineNo">202</span> // start by checking for error first<a name="line.202"></a>
<span class="sourceLineNo">203</span> monitor.rethrowException();<a name="line.203"></a>
<span class="sourceLineNo">204</span> LOG.debug("Procedure '" + procName + "' starting 'acquire'");<a name="line.204"></a>
<span class="sourceLineNo">205</span> sendGlobalBarrierStart();<a name="line.205"></a>
<span class="sourceLineNo">206</span><a name="line.206"></a>
<span class="sourceLineNo">207</span> // wait for all the members to report acquisition<a name="line.207"></a>
<span class="sourceLineNo">208</span> LOG.debug("Waiting for all members to 'acquire'");<a name="line.208"></a>
<span class="sourceLineNo">209</span> waitForLatch(acquiredBarrierLatch, monitor, wakeFrequency, "acquired");<a name="line.209"></a>
<span class="sourceLineNo">210</span> monitor.rethrowException();<a name="line.210"></a>
<span class="sourceLineNo">211</span><a name="line.211"></a>
<span class="sourceLineNo">212</span> LOG.debug("Procedure '" + procName + "' starting 'in-barrier' execution.");<a name="line.212"></a>
<span class="sourceLineNo">213</span> sendGlobalBarrierReached();<a name="line.213"></a>
<span class="sourceLineNo">214</span><a name="line.214"></a>
<span class="sourceLineNo">215</span> // wait for all members to report barrier release<a name="line.215"></a>
<span class="sourceLineNo">216</span> LOG.debug("Waiting for all members to 'release'");<a name="line.216"></a>
<span class="sourceLineNo">217</span> waitForLatch(releasedBarrierLatch, monitor, wakeFrequency, "released");<a name="line.217"></a>
<span class="sourceLineNo">218</span><a name="line.218"></a>
<span class="sourceLineNo">219</span> // make sure we didn't get an error during in barrier execution and release<a name="line.219"></a>
<span class="sourceLineNo">220</span> monitor.rethrowException();<a name="line.220"></a>
<span class="sourceLineNo">221</span> LOG.info("Procedure '" + procName + "' execution completed");<a name="line.221"></a>
<span class="sourceLineNo">222</span> } catch (Exception e) {<a name="line.222"></a>
<span class="sourceLineNo">223</span> if (e instanceof InterruptedException) {<a name="line.223"></a>
<span class="sourceLineNo">224</span> Thread.currentThread().interrupt();<a name="line.224"></a>
<span class="sourceLineNo">225</span> }<a name="line.225"></a>
<span class="sourceLineNo">226</span> String msg = "Procedure '" + procName +"' execution failed!";<a name="line.226"></a>
<span class="sourceLineNo">227</span> LOG.error(msg, e);<a name="line.227"></a>
<span class="sourceLineNo">228</span> receive(new ForeignException(getName(), e));<a name="line.228"></a>
<span class="sourceLineNo">229</span> } finally {<a name="line.229"></a>
<span class="sourceLineNo">230</span> LOG.debug("Running finish phase.");<a name="line.230"></a>
<span class="sourceLineNo">231</span> sendGlobalBarrierComplete();<a name="line.231"></a>
<span class="sourceLineNo">232</span> completedLatch.countDown();<a name="line.232"></a>
<span class="sourceLineNo">233</span><a name="line.233"></a>
<span class="sourceLineNo">234</span> // tell the timer we are done, if we get here successfully<a name="line.234"></a>
<span class="sourceLineNo">235</span> timeoutInjector.complete();<a name="line.235"></a>
<span class="sourceLineNo">236</span> return null;<a name="line.236"></a>
<span class="sourceLineNo">237</span> }<a name="line.237"></a>
<span class="sourceLineNo">238</span> }<a name="line.238"></a>
<span class="sourceLineNo">239</span><a name="line.239"></a>
<span class="sourceLineNo">240</span> /**<a name="line.240"></a>
<span class="sourceLineNo">241</span> * Sends a message to Members to create a new {@link Subprocedure} for this Procedure and execute<a name="line.241"></a>
<span class="sourceLineNo">242</span> * the {@link Subprocedure#acquireBarrier} step.<a name="line.242"></a>
<span class="sourceLineNo">243</span> * @throws ForeignException<a name="line.243"></a>
<span class="sourceLineNo">244</span> */<a name="line.244"></a>
<span class="sourceLineNo">245</span> public void sendGlobalBarrierStart() throws ForeignException {<a name="line.245"></a>
<span class="sourceLineNo">246</span> // start the procedure<a name="line.246"></a>
<span class="sourceLineNo">247</span> LOG.debug("Starting procedure '" + procName + "', kicking off acquire phase on members.");<a name="line.247"></a>
<span class="sourceLineNo">248</span> try {<a name="line.248"></a>
<span class="sourceLineNo">249</span> // send procedure barrier start to specified list of members. cloning the list to avoid<a name="line.249"></a>
<span class="sourceLineNo">250</span> // concurrent modification from the controller setting the prepared nodes<a name="line.250"></a>
<span class="sourceLineNo">251</span> coord.getRpcs().sendGlobalBarrierAcquire(this, args, Lists.newArrayList(this.acquiringMembers));<a name="line.251"></a>
<span class="sourceLineNo">252</span> } catch (IOException e) {<a name="line.252"></a>
<span class="sourceLineNo">253</span> coord.rpcConnectionFailure("Can't reach controller.", e);<a name="line.253"></a>
<span class="sourceLineNo">254</span> } catch (IllegalArgumentException e) {<a name="line.254"></a>
<span class="sourceLineNo">255</span> throw new ForeignException(getName(), e);<a name="line.255"></a>
<span class="sourceLineNo">256</span> }<a name="line.256"></a>
<span class="sourceLineNo">257</span> }<a name="line.257"></a>
<span class="sourceLineNo">258</span><a name="line.258"></a>
<span class="sourceLineNo">259</span> /**<a name="line.259"></a>
<span class="sourceLineNo">260</span> * Sends a message to all members that the global barrier condition has been satisfied. This<a name="line.260"></a>
<span class="sourceLineNo">261</span> * should only be executed after all members have completed its<a name="line.261"></a>
<span class="sourceLineNo">262</span> * {@link Subprocedure#acquireBarrier()} call successfully. This triggers the member<a name="line.262"></a>
<span class="sourceLineNo">263</span> * {@link Subprocedure#insideBarrier} method.<a name="line.263"></a>
<span class="sourceLineNo">264</span> * @throws ForeignException<a name="line.264"></a>
<span class="sourceLineNo">265</span> */<a name="line.265"></a>
<span class="sourceLineNo">266</span> public void sendGlobalBarrierReached() throws ForeignException {<a name="line.266"></a>
<span class="sourceLineNo">267</span> try {<a name="line.267"></a>
<span class="sourceLineNo">268</span> // trigger to have member run {@link Subprocedure#insideBarrier}<a name="line.268"></a>
<span class="sourceLineNo">269</span> coord.getRpcs().sendGlobalBarrierReached(this, Lists.newArrayList(inBarrierMembers));<a name="line.269"></a>
<span class="sourceLineNo">270</span> } catch (IOException e) {<a name="line.270"></a>
<span class="sourceLineNo">271</span> coord.rpcConnectionFailure("Can't reach controller.", e);<a name="line.271"></a>
<span class="sourceLineNo">272</span> }<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> * Sends a message to members that all {@link Subprocedure#insideBarrier} calls have completed.<a name="line.276"></a>
<span class="sourceLineNo">277</span> * After this executes, the coordinator can assume that any state resources about this barrier<a name="line.277"></a>
<span class="sourceLineNo">278</span> * procedure state has been released.<a name="line.278"></a>
<span class="sourceLineNo">279</span> */<a name="line.279"></a>
<span class="sourceLineNo">280</span> public void sendGlobalBarrierComplete() {<a name="line.280"></a>
<span class="sourceLineNo">281</span> LOG.debug("Finished coordinator procedure - removing self from list of running procedures");<a name="line.281"></a>
<span class="sourceLineNo">282</span> try {<a name="line.282"></a>
<span class="sourceLineNo">283</span> coord.getRpcs().resetMembers(this);<a name="line.283"></a>
<span class="sourceLineNo">284</span> } catch (IOException e) {<a name="line.284"></a>
<span class="sourceLineNo">285</span> coord.rpcConnectionFailure("Failed to reset procedure:" + procName, e);<a name="line.285"></a>
<span class="sourceLineNo">286</span> }<a name="line.286"></a>
<span class="sourceLineNo">287</span> }<a name="line.287"></a>
<span class="sourceLineNo">288</span><a name="line.288"></a>
<span class="sourceLineNo">289</span> //<a name="line.289"></a>
<span class="sourceLineNo">290</span> // Call backs from other external processes.<a name="line.290"></a>
<span class="sourceLineNo">291</span> //<a name="line.291"></a>
<span class="sourceLineNo">292</span><a name="line.292"></a>
<span class="sourceLineNo">293</span> /**<a name="line.293"></a>
<span class="sourceLineNo">294</span> * Call back triggered by an individual member upon successful local barrier acquisition<a name="line.294"></a>
<span class="sourceLineNo">295</span> * @param member<a name="line.295"></a>
<span class="sourceLineNo">296</span> */<a name="line.296"></a>
<span class="sourceLineNo">297</span> public void barrierAcquiredByMember(String member) {<a name="line.297"></a>
<span class="sourceLineNo">298</span> LOG.debug("member: '" + member + "' joining acquired barrier for procedure '" + procName<a name="line.298"></a>
<span class="sourceLineNo">299</span> + "' on coordinator");<a name="line.299"></a>
<span class="sourceLineNo">300</span> if (this.acquiringMembers.contains(member)) {<a name="line.300"></a>
<span class="sourceLineNo">301</span> synchronized (joinBarrierLock) {<a name="line.301"></a>
<span class="sourceLineNo">302</span> if (this.acquiringMembers.remove(member)) {<a name="line.302"></a>
<span class="sourceLineNo">303</span> this.inBarrierMembers.add(member);<a name="line.303"></a>
<span class="sourceLineNo">304</span> acquiredBarrierLatch.countDown();<a name="line.304"></a>
<span class="sourceLineNo">305</span> }<a name="line.305"></a>
<span class="sourceLineNo">306</span> }<a name="line.306"></a>
<span class="sourceLineNo">307</span> LOG.debug("Waiting on: " + acquiredBarrierLatch + " remaining members to acquire global barrier");<a name="line.307"></a>
<span class="sourceLineNo">308</span> } else {<a name="line.308"></a>
<span class="sourceLineNo">309</span> LOG.warn("Member " + member + " joined barrier, but we weren't waiting on it to join." +<a name="line.309"></a>
<span class="sourceLineNo">310</span> " Continuing on.");<a name="line.310"></a>
<span class="sourceLineNo">311</span> }<a name="line.311"></a>
<span class="sourceLineNo">312</span> }<a name="line.312"></a>
<span class="sourceLineNo">313</span><a name="line.313"></a>
<span class="sourceLineNo">314</span> /**<a name="line.314"></a>
<span class="sourceLineNo">315</span> * Call back triggered by a individual member upon successful local in-barrier execution and<a name="line.315"></a>
<span class="sourceLineNo">316</span> * release<a name="line.316"></a>
<span class="sourceLineNo">317</span> * @param member<a name="line.317"></a>
<span class="sourceLineNo">318</span> * @param dataFromMember<a name="line.318"></a>
<span class="sourceLineNo">319</span> */<a name="line.319"></a>
<span class="sourceLineNo">320</span> public void barrierReleasedByMember(String member, byte[] dataFromMember) {<a name="line.320"></a>
<span class="sourceLineNo">321</span> boolean removed = false;<a name="line.321"></a>
<span class="sourceLineNo">322</span> synchronized (joinBarrierLock) {<a name="line.322"></a>
<span class="sourceLineNo">323</span> removed = this.inBarrierMembers.remove(member);<a name="line.323"></a>
<span class="sourceLineNo">324</span> if (removed) {<a name="line.324"></a>
<span class="sourceLineNo">325</span> releasedBarrierLatch.countDown();<a name="line.325"></a>
<span class="sourceLineNo">326</span> }<a name="line.326"></a>
<span class="sourceLineNo">327</span> }<a name="line.327"></a>
<span class="sourceLineNo">328</span> if (removed) {<a name="line.328"></a>
<span class="sourceLineNo">329</span> LOG.debug("Member: '" + member + "' released barrier for procedure'" + procName<a name="line.329"></a>
<span class="sourceLineNo">330</span> + "', counting down latch. Waiting for " + releasedBarrierLatch.getCount()<a name="line.330"></a>
<span class="sourceLineNo">331</span> + " more");<a name="line.331"></a>
<span class="sourceLineNo">332</span> } else {<a name="line.332"></a>
<span class="sourceLineNo">333</span> LOG.warn("Member: '" + member + "' released barrier for procedure'" + procName<a name="line.333"></a>
<span class="sourceLineNo">334</span> + "', but we weren't waiting on it to release!");<a name="line.334"></a>
<span class="sourceLineNo">335</span> }<a name="line.335"></a>
<span class="sourceLineNo">336</span> dataFromFinishedMembers.put(member, dataFromMember);<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> /**<a name="line.339"></a>
<span class="sourceLineNo">340</span> * Waits until the entire procedure has globally completed, or has been aborted. If an<a name="line.340"></a>
<span class="sourceLineNo">341</span> * exception is thrown the procedure may or not have run cleanup to trigger the completion latch<a name="line.341"></a>
<span class="sourceLineNo">342</span> * yet.<a name="line.342"></a>
<span class="sourceLineNo">343</span> * @throws ForeignException<a name="line.343"></a>
<span class="sourceLineNo">344</span> * @throws InterruptedException<a name="line.344"></a>
<span class="sourceLineNo">345</span> */<a name="line.345"></a>
<span class="sourceLineNo">346</span> public void waitForCompleted() throws ForeignException, InterruptedException {<a name="line.346"></a>
<span class="sourceLineNo">347</span> waitForLatch(completedLatch, monitor, wakeFrequency, procName + " completed");<a name="line.347"></a>
<span class="sourceLineNo">348</span> }<a name="line.348"></a>
<span class="sourceLineNo">349</span><a name="line.349"></a>
<span class="sourceLineNo">350</span> /**<a name="line.350"></a>
<span class="sourceLineNo">351</span> * Waits until the entire procedure has globally completed, or has been aborted. If an<a name="line.351"></a>
<span class="sourceLineNo">352</span> * exception is thrown the procedure may or not have run cleanup to trigger the completion latch<a name="line.352"></a>
<span class="sourceLineNo">353</span> * yet.<a name="line.353"></a>
<span class="sourceLineNo">354</span> * @return data returned from procedure members upon successfully completing subprocedure.<a name="line.354"></a>
<span class="sourceLineNo">355</span> * @throws ForeignException<a name="line.355"></a>
<span class="sourceLineNo">356</span> * @throws InterruptedException<a name="line.356"></a>
<span class="sourceLineNo">357</span> */<a name="line.357"></a>
<span class="sourceLineNo">358</span> public HashMap&lt;String, byte[]&gt; waitForCompletedWithRet() throws ForeignException, InterruptedException {<a name="line.358"></a>
<span class="sourceLineNo">359</span> waitForCompleted();<a name="line.359"></a>
<span class="sourceLineNo">360</span> return dataFromFinishedMembers;<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> /**<a name="line.363"></a>
<span class="sourceLineNo">364</span> * Check if the entire procedure has globally completed, or has been aborted.<a name="line.364"></a>
<span class="sourceLineNo">365</span> * @throws ForeignException<a name="line.365"></a>
<span class="sourceLineNo">366</span> */<a name="line.366"></a>
<span class="sourceLineNo">367</span> public boolean isCompleted() throws ForeignException {<a name="line.367"></a>
<span class="sourceLineNo">368</span> // Rethrow exception if any<a name="line.368"></a>
<span class="sourceLineNo">369</span> monitor.rethrowException();<a name="line.369"></a>
<span class="sourceLineNo">370</span> return (completedLatch.getCount() == 0);<a name="line.370"></a>
<span class="sourceLineNo">371</span> }<a name="line.371"></a>
<span class="sourceLineNo">372</span><a name="line.372"></a>
<span class="sourceLineNo">373</span> /**<a name="line.373"></a>
<span class="sourceLineNo">374</span> * A callback that handles incoming ForeignExceptions.<a name="line.374"></a>
<span class="sourceLineNo">375</span> */<a name="line.375"></a>
<span class="sourceLineNo">376</span> @Override<a name="line.376"></a>
<span class="sourceLineNo">377</span> public void receive(ForeignException e) {<a name="line.377"></a>
<span class="sourceLineNo">378</span> monitor.receive(e);<a name="line.378"></a>
<span class="sourceLineNo">379</span> }<a name="line.379"></a>
<span class="sourceLineNo">380</span><a name="line.380"></a>
<span class="sourceLineNo">381</span> /**<a name="line.381"></a>
<span class="sourceLineNo">382</span> * Wait for latch to count to zero, ignoring any spurious wake-ups, but waking periodically to<a name="line.382"></a>
<span class="sourceLineNo">383</span> * check for errors<a name="line.383"></a>
<span class="sourceLineNo">384</span> * @param latch latch to wait on<a name="line.384"></a>
<span class="sourceLineNo">385</span> * @param monitor monitor to check for errors while waiting<a name="line.385"></a>
<span class="sourceLineNo">386</span> * @param wakeFrequency frequency to wake up and check for errors (in<a name="line.386"></a>
<span class="sourceLineNo">387</span> * {@link TimeUnit#MILLISECONDS})<a name="line.387"></a>
<span class="sourceLineNo">388</span> * @param latchDescription description of the latch, for logging<a name="line.388"></a>
<span class="sourceLineNo">389</span> * @throws ForeignException type of error the monitor can throw, if the task fails<a name="line.389"></a>
<span class="sourceLineNo">390</span> * @throws InterruptedException if we are interrupted while waiting on latch<a name="line.390"></a>
<span class="sourceLineNo">391</span> */<a name="line.391"></a>
<span class="sourceLineNo">392</span> public static void waitForLatch(CountDownLatch latch, ForeignExceptionSnare monitor,<a name="line.392"></a>
<span class="sourceLineNo">393</span> long wakeFrequency, String latchDescription) throws ForeignException,<a name="line.393"></a>
<span class="sourceLineNo">394</span> InterruptedException {<a name="line.394"></a>
<span class="sourceLineNo">395</span> boolean released = false;<a name="line.395"></a>
<span class="sourceLineNo">396</span> while (!released) {<a name="line.396"></a>
<span class="sourceLineNo">397</span> if (monitor != null) {<a name="line.397"></a>
<span class="sourceLineNo">398</span> monitor.rethrowException();<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>
<span class="sourceLineNo">401</span> ForeignExceptionDispatcher.LOG.debug("Waiting for '" + latchDescription + "' latch. (sleep:"<a name="line.401"></a>
<span class="sourceLineNo">402</span> + wakeFrequency + " ms)"); */<a name="line.402"></a>
<span class="sourceLineNo">403</span> released = latch.await(wakeFrequency, TimeUnit.MILLISECONDS);<a name="line.403"></a>
<span class="sourceLineNo">404</span> }<a name="line.404"></a>
<span class="sourceLineNo">405</span> // check error again in case an error raised during last wait<a name="line.405"></a>
<span class="sourceLineNo">406</span> if (monitor != null) {<a name="line.406"></a>
<span class="sourceLineNo">407</span> monitor.rethrowException();<a name="line.407"></a>
<span class="sourceLineNo">408</span> }<a name="line.408"></a>
<span class="sourceLineNo">409</span> }<a name="line.409"></a>
<span class="sourceLineNo">410</span>}<a name="line.410"></a>
</pre>
</div>
</body>
</html>