| <!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.chaos;<a name="line.18"></a> |
| <span class="sourceLineNo">019</span><a name="line.19"></a> |
| <span class="sourceLineNo">020</span>import java.io.Closeable;<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.nio.charset.StandardCharsets;<a name="line.22"></a> |
| <span class="sourceLineNo">023</span>import java.util.ArrayList;<a name="line.23"></a> |
| <span class="sourceLineNo">024</span>import java.util.List;<a name="line.24"></a> |
| <span class="sourceLineNo">025</span>import java.util.concurrent.atomic.AtomicBoolean;<a name="line.25"></a> |
| <span class="sourceLineNo">026</span>import org.apache.hadoop.conf.Configuration;<a name="line.26"></a> |
| <span class="sourceLineNo">027</span>import org.apache.hadoop.hbase.util.Pair;<a name="line.27"></a> |
| <span class="sourceLineNo">028</span>import org.apache.hadoop.hbase.util.RetryCounter;<a name="line.28"></a> |
| <span class="sourceLineNo">029</span>import org.apache.hadoop.hbase.util.RetryCounterFactory;<a name="line.29"></a> |
| <span class="sourceLineNo">030</span>import org.apache.hadoop.util.Shell;<a name="line.30"></a> |
| <span class="sourceLineNo">031</span>import org.apache.yetus.audience.InterfaceAudience;<a name="line.31"></a> |
| <span class="sourceLineNo">032</span>import org.apache.zookeeper.AsyncCallback;<a name="line.32"></a> |
| <span class="sourceLineNo">033</span>import org.apache.zookeeper.CreateMode;<a name="line.33"></a> |
| <span class="sourceLineNo">034</span>import org.apache.zookeeper.KeeperException;<a name="line.34"></a> |
| <span class="sourceLineNo">035</span>import org.apache.zookeeper.WatchedEvent;<a name="line.35"></a> |
| <span class="sourceLineNo">036</span>import org.apache.zookeeper.Watcher;<a name="line.36"></a> |
| <span class="sourceLineNo">037</span>import org.apache.zookeeper.ZooDefs;<a name="line.37"></a> |
| <span class="sourceLineNo">038</span>import org.apache.zookeeper.ZooKeeper;<a name="line.38"></a> |
| <span class="sourceLineNo">039</span>import org.apache.zookeeper.data.Stat;<a name="line.39"></a> |
| <span class="sourceLineNo">040</span>import org.slf4j.Logger;<a name="line.40"></a> |
| <span class="sourceLineNo">041</span>import org.slf4j.LoggerFactory;<a name="line.41"></a> |
| <span class="sourceLineNo">042</span><a name="line.42"></a> |
| <span class="sourceLineNo">043</span>/***<a name="line.43"></a> |
| <span class="sourceLineNo">044</span> * An agent for executing destructive actions for ChaosMonkey. Uses ZooKeeper Watchers and<a name="line.44"></a> |
| <span class="sourceLineNo">045</span> * LocalShell, to do the killing and getting status of service on targeted host without SSH. uses<a name="line.45"></a> |
| <span class="sourceLineNo">046</span> * given ZNode Structure: /perfChaosTest (root) | | /chaosAgents (Used for registration has hostname<a name="line.46"></a> |
| <span class="sourceLineNo">047</span> * ephemeral nodes as children) | | /chaosAgentTaskStatus (Used for task Execution, has hostname<a name="line.47"></a> |
| <span class="sourceLineNo">048</span> * persistent nodes as child with tasks as their children) | | /hostname | | /task0000001 (command<a name="line.48"></a> |
| <span class="sourceLineNo">049</span> * as data) (has two types of command : 1: starts with "exec" for executing a destructive action. 2:<a name="line.49"></a> |
| <span class="sourceLineNo">050</span> * starts with "bool" for getting only status of service.<a name="line.50"></a> |
| <span class="sourceLineNo">051</span> */<a name="line.51"></a> |
| <span class="sourceLineNo">052</span>@InterfaceAudience.Private<a name="line.52"></a> |
| <span class="sourceLineNo">053</span>public class ChaosAgent implements Watcher, Closeable, Runnable {<a name="line.53"></a> |
| <span class="sourceLineNo">054</span><a name="line.54"></a> |
| <span class="sourceLineNo">055</span> private static final Logger LOG = LoggerFactory.getLogger(ChaosAgent.class);<a name="line.55"></a> |
| <span class="sourceLineNo">056</span> static AtomicBoolean stopChaosAgent = new AtomicBoolean();<a name="line.56"></a> |
| <span class="sourceLineNo">057</span> private ZooKeeper zk;<a name="line.57"></a> |
| <span class="sourceLineNo">058</span> private String quorum;<a name="line.58"></a> |
| <span class="sourceLineNo">059</span> private String agentName;<a name="line.59"></a> |
| <span class="sourceLineNo">060</span> private Configuration conf;<a name="line.60"></a> |
| <span class="sourceLineNo">061</span> private RetryCounterFactory retryCounterFactory;<a name="line.61"></a> |
| <span class="sourceLineNo">062</span> private volatile boolean connected = false;<a name="line.62"></a> |
| <span class="sourceLineNo">063</span><a name="line.63"></a> |
| <span class="sourceLineNo">064</span> public ChaosAgent(Configuration conf, String quorum, String agentName) {<a name="line.64"></a> |
| <span class="sourceLineNo">065</span> initChaosAgent(conf, quorum, agentName);<a name="line.65"></a> |
| <span class="sourceLineNo">066</span> }<a name="line.66"></a> |
| <span class="sourceLineNo">067</span><a name="line.67"></a> |
| <span class="sourceLineNo">068</span> /***<a name="line.68"></a> |
| <span class="sourceLineNo">069</span> * sets global params and initiates connection with ZooKeeper then does registration.<a name="line.69"></a> |
| <span class="sourceLineNo">070</span> * @param conf initial configuration to use<a name="line.70"></a> |
| <span class="sourceLineNo">071</span> * @param quorum ZK Quorum<a name="line.71"></a> |
| <span class="sourceLineNo">072</span> * @param agentName AgentName to use<a name="line.72"></a> |
| <span class="sourceLineNo">073</span> */<a name="line.73"></a> |
| <span class="sourceLineNo">074</span> private void initChaosAgent(Configuration conf, String quorum, String agentName) {<a name="line.74"></a> |
| <span class="sourceLineNo">075</span> this.conf = conf;<a name="line.75"></a> |
| <span class="sourceLineNo">076</span> this.quorum = quorum;<a name="line.76"></a> |
| <span class="sourceLineNo">077</span> this.agentName = agentName;<a name="line.77"></a> |
| <span class="sourceLineNo">078</span> this.retryCounterFactory = new RetryCounterFactory(new RetryCounter.RetryConfig()<a name="line.78"></a> |
| <span class="sourceLineNo">079</span> .setMaxAttempts(<a name="line.79"></a> |
| <span class="sourceLineNo">080</span> conf.getInt(ChaosConstants.RETRY_ATTEMPTS_KEY, ChaosConstants.DEFAULT_RETRY_ATTEMPTS))<a name="line.80"></a> |
| <span class="sourceLineNo">081</span> .setSleepInterval(conf.getLong(ChaosConstants.RETRY_SLEEP_INTERVAL_KEY,<a name="line.81"></a> |
| <span class="sourceLineNo">082</span> ChaosConstants.DEFAULT_RETRY_SLEEP_INTERVAL)));<a name="line.82"></a> |
| <span class="sourceLineNo">083</span> try {<a name="line.83"></a> |
| <span class="sourceLineNo">084</span> this.createZKConnection(null);<a name="line.84"></a> |
| <span class="sourceLineNo">085</span> this.register();<a name="line.85"></a> |
| <span class="sourceLineNo">086</span> } catch (IOException e) {<a name="line.86"></a> |
| <span class="sourceLineNo">087</span> LOG.error("Error Creating Connection: " + e);<a name="line.87"></a> |
| <span class="sourceLineNo">088</span> }<a name="line.88"></a> |
| <span class="sourceLineNo">089</span> }<a name="line.89"></a> |
| <span class="sourceLineNo">090</span><a name="line.90"></a> |
| <span class="sourceLineNo">091</span> /***<a name="line.91"></a> |
| <span class="sourceLineNo">092</span> * Creates Connection with ZooKeeper.<a name="line.92"></a> |
| <span class="sourceLineNo">093</span> * @throws IOException if something goes wrong<a name="line.93"></a> |
| <span class="sourceLineNo">094</span> */<a name="line.94"></a> |
| <span class="sourceLineNo">095</span> private void createZKConnection(Watcher watcher) throws IOException {<a name="line.95"></a> |
| <span class="sourceLineNo">096</span> if (watcher == null) {<a name="line.96"></a> |
| <span class="sourceLineNo">097</span> zk = new ZooKeeper(quorum, ChaosConstants.SESSION_TIMEOUT_ZK, this);<a name="line.97"></a> |
| <span class="sourceLineNo">098</span> } else {<a name="line.98"></a> |
| <span class="sourceLineNo">099</span> zk = new ZooKeeper(quorum, ChaosConstants.SESSION_TIMEOUT_ZK, watcher);<a name="line.99"></a> |
| <span class="sourceLineNo">100</span> }<a name="line.100"></a> |
| <span class="sourceLineNo">101</span> LOG.info("ZooKeeper Connection created for ChaosAgent: " + agentName);<a name="line.101"></a> |
| <span class="sourceLineNo">102</span> }<a name="line.102"></a> |
| <span class="sourceLineNo">103</span><a name="line.103"></a> |
| <span class="sourceLineNo">104</span> // WATCHERS: Below are the Watches used by ChaosAgent<a name="line.104"></a> |
| <span class="sourceLineNo">105</span><a name="line.105"></a> |
| <span class="sourceLineNo">106</span> /***<a name="line.106"></a> |
| <span class="sourceLineNo">107</span> * Watcher for notifying if any task is assigned to agent or not, by seeking if any Node is being<a name="line.107"></a> |
| <span class="sourceLineNo">108</span> * added to agent as Child.<a name="line.108"></a> |
| <span class="sourceLineNo">109</span> */<a name="line.109"></a> |
| <span class="sourceLineNo">110</span> Watcher newTaskCreatedWatcher = new Watcher() {<a name="line.110"></a> |
| <span class="sourceLineNo">111</span> @Override<a name="line.111"></a> |
| <span class="sourceLineNo">112</span> public void process(WatchedEvent watchedEvent) {<a name="line.112"></a> |
| <span class="sourceLineNo">113</span> if (watchedEvent.getType() == Event.EventType.NodeChildrenChanged) {<a name="line.113"></a> |
| <span class="sourceLineNo">114</span> if (<a name="line.114"></a> |
| <span class="sourceLineNo">115</span> !(ChaosConstants.CHAOS_AGENT_STATUS_PERSISTENT_ZNODE + ChaosConstants.ZNODE_PATH_SEPARATOR<a name="line.115"></a> |
| <span class="sourceLineNo">116</span> + agentName).equals(watchedEvent.getPath())<a name="line.116"></a> |
| <span class="sourceLineNo">117</span> ) {<a name="line.117"></a> |
| <span class="sourceLineNo">118</span> throw new RuntimeException(<a name="line.118"></a> |
| <span class="sourceLineNo">119</span> KeeperException.create(KeeperException.Code.DATAINCONSISTENCY));<a name="line.119"></a> |
| <span class="sourceLineNo">120</span> }<a name="line.120"></a> |
| <span class="sourceLineNo">121</span><a name="line.121"></a> |
| <span class="sourceLineNo">122</span> LOG.info("Change in Tasks Node, checking for Tasks again.");<a name="line.122"></a> |
| <span class="sourceLineNo">123</span> getTasks();<a name="line.123"></a> |
| <span class="sourceLineNo">124</span> }<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> };<a name="line.127"></a> |
| <span class="sourceLineNo">128</span><a name="line.128"></a> |
| <span class="sourceLineNo">129</span> // CALLBACKS: Below are the Callbacks used by Chaos Agent<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> * Callback used while setting status of a given task, Logs given status.<a name="line.132"></a> |
| <span class="sourceLineNo">133</span> */<a name="line.133"></a> |
| <span class="sourceLineNo">134</span> AsyncCallback.StatCallback setStatusOfTaskZNodeCallback = (rc, path, ctx, stat) -> {<a name="line.134"></a> |
| <span class="sourceLineNo">135</span> switch (KeeperException.Code.get(rc)) {<a name="line.135"></a> |
| <span class="sourceLineNo">136</span> case CONNECTIONLOSS:<a name="line.136"></a> |
| <span class="sourceLineNo">137</span> // Connection to the server was lost while setting status setting again.<a name="line.137"></a> |
| <span class="sourceLineNo">138</span> try {<a name="line.138"></a> |
| <span class="sourceLineNo">139</span> recreateZKConnection();<a name="line.139"></a> |
| <span class="sourceLineNo">140</span> } catch (Exception e) {<a name="line.140"></a> |
| <span class="sourceLineNo">141</span> break;<a name="line.141"></a> |
| <span class="sourceLineNo">142</span> }<a name="line.142"></a> |
| <span class="sourceLineNo">143</span> setStatusOfTaskZNode(path, (String) ctx);<a name="line.143"></a> |
| <span class="sourceLineNo">144</span> break;<a name="line.144"></a> |
| <span class="sourceLineNo">145</span><a name="line.145"></a> |
| <span class="sourceLineNo">146</span> case OK:<a name="line.146"></a> |
| <span class="sourceLineNo">147</span> LOG.info("Status of Task has been set");<a name="line.147"></a> |
| <span class="sourceLineNo">148</span> break;<a name="line.148"></a> |
| <span class="sourceLineNo">149</span><a name="line.149"></a> |
| <span class="sourceLineNo">150</span> case NONODE:<a name="line.150"></a> |
| <span class="sourceLineNo">151</span> LOG.error("Chaos Agent status node does not exists: "<a name="line.151"></a> |
| <span class="sourceLineNo">152</span> + "check for ZNode directory structure again.");<a name="line.152"></a> |
| <span class="sourceLineNo">153</span> break;<a name="line.153"></a> |
| <span class="sourceLineNo">154</span><a name="line.154"></a> |
| <span class="sourceLineNo">155</span> default:<a name="line.155"></a> |
| <span class="sourceLineNo">156</span> LOG.error("Error while setting status of task ZNode: " + path,<a name="line.156"></a> |
| <span class="sourceLineNo">157</span> KeeperException.create(KeeperException.Code.get(rc), path));<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><a name="line.160"></a> |
| <span class="sourceLineNo">161</span> /**<a name="line.161"></a> |
| <span class="sourceLineNo">162</span> * Callback used while creating a Persistent ZNode tries to create ZNode again if Connection was<a name="line.162"></a> |
| <span class="sourceLineNo">163</span> * lost in previous try.<a name="line.163"></a> |
| <span class="sourceLineNo">164</span> */<a name="line.164"></a> |
| <span class="sourceLineNo">165</span> AsyncCallback.StringCallback createZNodeCallback = (rc, path, ctx, name) -> {<a name="line.165"></a> |
| <span class="sourceLineNo">166</span> switch (KeeperException.Code.get(rc)) {<a name="line.166"></a> |
| <span class="sourceLineNo">167</span> case CONNECTIONLOSS:<a name="line.167"></a> |
| <span class="sourceLineNo">168</span> try {<a name="line.168"></a> |
| <span class="sourceLineNo">169</span> recreateZKConnection();<a name="line.169"></a> |
| <span class="sourceLineNo">170</span> } catch (Exception e) {<a name="line.170"></a> |
| <span class="sourceLineNo">171</span> break;<a name="line.171"></a> |
| <span class="sourceLineNo">172</span> }<a name="line.172"></a> |
| <span class="sourceLineNo">173</span> createZNode(path, (byte[]) ctx);<a name="line.173"></a> |
| <span class="sourceLineNo">174</span> break;<a name="line.174"></a> |
| <span class="sourceLineNo">175</span> case OK:<a name="line.175"></a> |
| <span class="sourceLineNo">176</span> LOG.info("ZNode created : " + path);<a name="line.176"></a> |
| <span class="sourceLineNo">177</span> break;<a name="line.177"></a> |
| <span class="sourceLineNo">178</span> case NODEEXISTS:<a name="line.178"></a> |
| <span class="sourceLineNo">179</span> LOG.warn("ZNode already registered: " + path);<a name="line.179"></a> |
| <span class="sourceLineNo">180</span> break;<a name="line.180"></a> |
| <span class="sourceLineNo">181</span> default:<a name="line.181"></a> |
| <span class="sourceLineNo">182</span> LOG.error("Error occurred while creating Persistent ZNode: " + path,<a name="line.182"></a> |
| <span class="sourceLineNo">183</span> KeeperException.create(KeeperException.Code.get(rc), path));<a name="line.183"></a> |
| <span class="sourceLineNo">184</span> }<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> * Callback used while creating a Ephemeral ZNode tries to create ZNode again if Connection was<a name="line.188"></a> |
| <span class="sourceLineNo">189</span> * lost in previous try.<a name="line.189"></a> |
| <span class="sourceLineNo">190</span> */<a name="line.190"></a> |
| <span class="sourceLineNo">191</span> AsyncCallback.StringCallback createEphemeralZNodeCallback = (rc, path, ctx, name) -> {<a name="line.191"></a> |
| <span class="sourceLineNo">192</span> switch (KeeperException.Code.get(rc)) {<a name="line.192"></a> |
| <span class="sourceLineNo">193</span> case CONNECTIONLOSS:<a name="line.193"></a> |
| <span class="sourceLineNo">194</span> try {<a name="line.194"></a> |
| <span class="sourceLineNo">195</span> recreateZKConnection();<a name="line.195"></a> |
| <span class="sourceLineNo">196</span> } catch (Exception e) {<a name="line.196"></a> |
| <span class="sourceLineNo">197</span> break;<a name="line.197"></a> |
| <span class="sourceLineNo">198</span> }<a name="line.198"></a> |
| <span class="sourceLineNo">199</span> createEphemeralZNode(path, (byte[]) ctx);<a name="line.199"></a> |
| <span class="sourceLineNo">200</span> break;<a name="line.200"></a> |
| <span class="sourceLineNo">201</span> case OK:<a name="line.201"></a> |
| <span class="sourceLineNo">202</span> LOG.info("ZNode created : " + path);<a name="line.202"></a> |
| <span class="sourceLineNo">203</span> break;<a name="line.203"></a> |
| <span class="sourceLineNo">204</span> case NODEEXISTS:<a name="line.204"></a> |
| <span class="sourceLineNo">205</span> LOG.warn("ZNode already registered: " + path);<a name="line.205"></a> |
| <span class="sourceLineNo">206</span> break;<a name="line.206"></a> |
| <span class="sourceLineNo">207</span> default:<a name="line.207"></a> |
| <span class="sourceLineNo">208</span> LOG.error("Error occurred while creating Ephemeral ZNode: ",<a name="line.208"></a> |
| <span class="sourceLineNo">209</span> KeeperException.create(KeeperException.Code.get(rc), path));<a name="line.209"></a> |
| <span class="sourceLineNo">210</span> }<a name="line.210"></a> |
| <span class="sourceLineNo">211</span> };<a name="line.211"></a> |
| <span class="sourceLineNo">212</span><a name="line.212"></a> |
| <span class="sourceLineNo">213</span> /**<a name="line.213"></a> |
| <span class="sourceLineNo">214</span> * Callback used by getTasksForAgentCallback while getting command, after getting command<a name="line.214"></a> |
| <span class="sourceLineNo">215</span> * successfully, it executes command and set its status with respect to the command type.<a name="line.215"></a> |
| <span class="sourceLineNo">216</span> */<a name="line.216"></a> |
| <span class="sourceLineNo">217</span> AsyncCallback.DataCallback getTaskForExecutionCallback = new AsyncCallback.DataCallback() {<a name="line.217"></a> |
| <span class="sourceLineNo">218</span> @Override<a name="line.218"></a> |
| <span class="sourceLineNo">219</span> public void processResult(int rc, String path, Object ctx, byte[] data, Stat stat) {<a name="line.219"></a> |
| <span class="sourceLineNo">220</span> switch (KeeperException.Code.get(rc)) {<a name="line.220"></a> |
| <span class="sourceLineNo">221</span> case CONNECTIONLOSS:<a name="line.221"></a> |
| <span class="sourceLineNo">222</span> // Connection to the server has been lost while getting task, getting data again.<a name="line.222"></a> |
| <span class="sourceLineNo">223</span> try {<a name="line.223"></a> |
| <span class="sourceLineNo">224</span> recreateZKConnection();<a name="line.224"></a> |
| <span class="sourceLineNo">225</span> } catch (Exception e) {<a name="line.225"></a> |
| <span class="sourceLineNo">226</span> break;<a name="line.226"></a> |
| <span class="sourceLineNo">227</span> }<a name="line.227"></a> |
| <span class="sourceLineNo">228</span> zk.getData(path, false, getTaskForExecutionCallback,<a name="line.228"></a> |
| <span class="sourceLineNo">229</span> new String(data, StandardCharsets.UTF_8));<a name="line.229"></a> |
| <span class="sourceLineNo">230</span> break;<a name="line.230"></a> |
| <span class="sourceLineNo">231</span> case OK:<a name="line.231"></a> |
| <span class="sourceLineNo">232</span> String cmd = new String(data, StandardCharsets.UTF_8);<a name="line.232"></a> |
| <span class="sourceLineNo">233</span> LOG.info("Executing command : " + cmd);<a name="line.233"></a> |
| <span class="sourceLineNo">234</span> String status = ChaosConstants.TASK_COMPLETION_STRING;<a name="line.234"></a> |
| <span class="sourceLineNo">235</span> try {<a name="line.235"></a> |
| <span class="sourceLineNo">236</span> String user =<a name="line.236"></a> |
| <span class="sourceLineNo">237</span> conf.get(ChaosConstants.CHAOSAGENT_SHELL_USER, ChaosConstants.DEFAULT_SHELL_USER);<a name="line.237"></a> |
| <span class="sourceLineNo">238</span> switch (cmd.substring(0, 4)) {<a name="line.238"></a> |
| <span class="sourceLineNo">239</span> case "bool":<a name="line.239"></a> |
| <span class="sourceLineNo">240</span> String ret = execWithRetries(user, cmd.substring(4)).getSecond();<a name="line.240"></a> |
| <span class="sourceLineNo">241</span> status = Boolean.toString(ret.length() > 0);<a name="line.241"></a> |
| <span class="sourceLineNo">242</span> break;<a name="line.242"></a> |
| <span class="sourceLineNo">243</span><a name="line.243"></a> |
| <span class="sourceLineNo">244</span> case "exec":<a name="line.244"></a> |
| <span class="sourceLineNo">245</span> execWithRetries(user, cmd.substring(4));<a name="line.245"></a> |
| <span class="sourceLineNo">246</span> break;<a name="line.246"></a> |
| <span class="sourceLineNo">247</span><a name="line.247"></a> |
| <span class="sourceLineNo">248</span> default:<a name="line.248"></a> |
| <span class="sourceLineNo">249</span> LOG.error("Unknown Command Type");<a name="line.249"></a> |
| <span class="sourceLineNo">250</span> status = ChaosConstants.TASK_ERROR_STRING;<a name="line.250"></a> |
| <span class="sourceLineNo">251</span> }<a name="line.251"></a> |
| <span class="sourceLineNo">252</span> } catch (IOException e) {<a name="line.252"></a> |
| <span class="sourceLineNo">253</span> LOG.error("Got error while executing command : " + cmd + " On agent : " + agentName<a name="line.253"></a> |
| <span class="sourceLineNo">254</span> + " Error : " + e);<a name="line.254"></a> |
| <span class="sourceLineNo">255</span> status = ChaosConstants.TASK_ERROR_STRING;<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> try {<a name="line.258"></a> |
| <span class="sourceLineNo">259</span> setStatusOfTaskZNode(path, status);<a name="line.259"></a> |
| <span class="sourceLineNo">260</span> Thread.sleep(ChaosConstants.SET_STATUS_SLEEP_TIME);<a name="line.260"></a> |
| <span class="sourceLineNo">261</span> } catch (InterruptedException e) {<a name="line.261"></a> |
| <span class="sourceLineNo">262</span> LOG.error("Error occured after setting status: " + e);<a name="line.262"></a> |
| <span class="sourceLineNo">263</span> }<a name="line.263"></a> |
| <span class="sourceLineNo">264</span><a name="line.264"></a> |
| <span class="sourceLineNo">265</span> default:<a name="line.265"></a> |
| <span class="sourceLineNo">266</span> LOG.error("Error occurred while getting data",<a name="line.266"></a> |
| <span class="sourceLineNo">267</span> KeeperException.create(KeeperException.Code.get(rc), path));<a name="line.267"></a> |
| <span class="sourceLineNo">268</span> }<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><a name="line.271"></a> |
| <span class="sourceLineNo">272</span> /***<a name="line.272"></a> |
| <span class="sourceLineNo">273</span> * Callback used while getting Tasks for agent if call executed without Exception, It creates a<a name="line.273"></a> |
| <span class="sourceLineNo">274</span> * separate thread for each children to execute given Tasks parallely.<a name="line.274"></a> |
| <span class="sourceLineNo">275</span> */<a name="line.275"></a> |
| <span class="sourceLineNo">276</span> AsyncCallback.ChildrenCallback getTasksForAgentCallback = new AsyncCallback.ChildrenCallback() {<a name="line.276"></a> |
| <span class="sourceLineNo">277</span> @Override<a name="line.277"></a> |
| <span class="sourceLineNo">278</span> public void processResult(int rc, String path, Object ctx, List<String> children) {<a name="line.278"></a> |
| <span class="sourceLineNo">279</span> switch (KeeperException.Code.get(rc)) {<a name="line.279"></a> |
| <span class="sourceLineNo">280</span> case CONNECTIONLOSS: {<a name="line.280"></a> |
| <span class="sourceLineNo">281</span> // Connection to the server has been lost, getting tasks again.<a name="line.281"></a> |
| <span class="sourceLineNo">282</span> try {<a name="line.282"></a> |
| <span class="sourceLineNo">283</span> recreateZKConnection();<a name="line.283"></a> |
| <span class="sourceLineNo">284</span> } catch (Exception e) {<a name="line.284"></a> |
| <span class="sourceLineNo">285</span> break;<a name="line.285"></a> |
| <span class="sourceLineNo">286</span> }<a name="line.286"></a> |
| <span class="sourceLineNo">287</span> getTasks();<a name="line.287"></a> |
| <span class="sourceLineNo">288</span> break;<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> case OK: {<a name="line.291"></a> |
| <span class="sourceLineNo">292</span> if (children != null) {<a name="line.292"></a> |
| <span class="sourceLineNo">293</span> try {<a name="line.293"></a> |
| <span class="sourceLineNo">294</span><a name="line.294"></a> |
| <span class="sourceLineNo">295</span> LOG.info("Executing each task as a separate thread");<a name="line.295"></a> |
| <span class="sourceLineNo">296</span> List<Thread> tasksList = new ArrayList<>();<a name="line.296"></a> |
| <span class="sourceLineNo">297</span> for (String task : children) {<a name="line.297"></a> |
| <span class="sourceLineNo">298</span> String threadName = agentName + "_" + task;<a name="line.298"></a> |
| <span class="sourceLineNo">299</span> Thread t = new Thread(() -> {<a name="line.299"></a> |
| <span class="sourceLineNo">300</span><a name="line.300"></a> |
| <span class="sourceLineNo">301</span> LOG.info("Executing task : " + task + " of agent : " + agentName);<a name="line.301"></a> |
| <span class="sourceLineNo">302</span> zk.getData(<a name="line.302"></a> |
| <span class="sourceLineNo">303</span> ChaosConstants.CHAOS_AGENT_STATUS_PERSISTENT_ZNODE<a name="line.303"></a> |
| <span class="sourceLineNo">304</span> + ChaosConstants.ZNODE_PATH_SEPARATOR + agentName<a name="line.304"></a> |
| <span class="sourceLineNo">305</span> + ChaosConstants.ZNODE_PATH_SEPARATOR + task,<a name="line.305"></a> |
| <span class="sourceLineNo">306</span> false, getTaskForExecutionCallback, task);<a name="line.306"></a> |
| <span class="sourceLineNo">307</span><a name="line.307"></a> |
| <span class="sourceLineNo">308</span> });<a name="line.308"></a> |
| <span class="sourceLineNo">309</span> t.setName(threadName);<a name="line.309"></a> |
| <span class="sourceLineNo">310</span> t.start();<a name="line.310"></a> |
| <span class="sourceLineNo">311</span> tasksList.add(t);<a name="line.311"></a> |
| <span class="sourceLineNo">312</span><a name="line.312"></a> |
| <span class="sourceLineNo">313</span> for (Thread thread : tasksList) {<a name="line.313"></a> |
| <span class="sourceLineNo">314</span> thread.join();<a name="line.314"></a> |
| <span class="sourceLineNo">315</span> }<a name="line.315"></a> |
| <span class="sourceLineNo">316</span> }<a name="line.316"></a> |
| <span class="sourceLineNo">317</span> } catch (InterruptedException e) {<a name="line.317"></a> |
| <span class="sourceLineNo">318</span> LOG.error(<a name="line.318"></a> |
| <span class="sourceLineNo">319</span> "Error scheduling next task : " + " for agent : " + agentName + " Error : " + e);<a name="line.319"></a> |
| <span class="sourceLineNo">320</span> }<a name="line.320"></a> |
| <span class="sourceLineNo">321</span> }<a name="line.321"></a> |
| <span class="sourceLineNo">322</span> break;<a name="line.322"></a> |
| <span class="sourceLineNo">323</span> }<a name="line.323"></a> |
| <span class="sourceLineNo">324</span><a name="line.324"></a> |
| <span class="sourceLineNo">325</span> default:<a name="line.325"></a> |
| <span class="sourceLineNo">326</span> LOG.error("Error occurred while getting task",<a name="line.326"></a> |
| <span class="sourceLineNo">327</span> KeeperException.create(KeeperException.Code.get(rc), path));<a name="line.327"></a> |
| <span class="sourceLineNo">328</span> }<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><a name="line.331"></a> |
| <span class="sourceLineNo">332</span> /***<a name="line.332"></a> |
| <span class="sourceLineNo">333</span> * Function to create PERSISTENT ZNODE with given path and data given as params<a name="line.333"></a> |
| <span class="sourceLineNo">334</span> * @param path Path at which ZNode to create<a name="line.334"></a> |
| <span class="sourceLineNo">335</span> * @param data Data to put under ZNode<a name="line.335"></a> |
| <span class="sourceLineNo">336</span> */<a name="line.336"></a> |
| <span class="sourceLineNo">337</span> public void createZNode(String path, byte[] data) {<a name="line.337"></a> |
| <span class="sourceLineNo">338</span> zk.create(path, data, ZooDefs.Ids.OPEN_ACL_UNSAFE, CreateMode.PERSISTENT, createZNodeCallback,<a name="line.338"></a> |
| <span class="sourceLineNo">339</span> data);<a name="line.339"></a> |
| <span class="sourceLineNo">340</span> }<a name="line.340"></a> |
| <span class="sourceLineNo">341</span><a name="line.341"></a> |
| <span class="sourceLineNo">342</span> /***<a name="line.342"></a> |
| <span class="sourceLineNo">343</span> * Function to create EPHEMERAL ZNODE with given path and data as params.<a name="line.343"></a> |
| <span class="sourceLineNo">344</span> * @param path Path at which Ephemeral ZNode to create<a name="line.344"></a> |
| <span class="sourceLineNo">345</span> * @param data Data to put under ZNode<a name="line.345"></a> |
| <span class="sourceLineNo">346</span> */<a name="line.346"></a> |
| <span class="sourceLineNo">347</span> public void createEphemeralZNode(String path, byte[] data) {<a name="line.347"></a> |
| <span class="sourceLineNo">348</span> zk.create(path, data, ZooDefs.Ids.OPEN_ACL_UNSAFE, CreateMode.EPHEMERAL,<a name="line.348"></a> |
| <span class="sourceLineNo">349</span> createEphemeralZNodeCallback, data);<a name="line.349"></a> |
| <span class="sourceLineNo">350</span> }<a name="line.350"></a> |
| <span class="sourceLineNo">351</span><a name="line.351"></a> |
| <span class="sourceLineNo">352</span> /**<a name="line.352"></a> |
| <span class="sourceLineNo">353</span> * Checks if given ZNode exists, if not creates a PERSISTENT ZNODE for same.<a name="line.353"></a> |
| <span class="sourceLineNo">354</span> * @param path Path to check for ZNode<a name="line.354"></a> |
| <span class="sourceLineNo">355</span> */<a name="line.355"></a> |
| <span class="sourceLineNo">356</span> private void createIfZNodeNotExists(String path) {<a name="line.356"></a> |
| <span class="sourceLineNo">357</span> try {<a name="line.357"></a> |
| <span class="sourceLineNo">358</span> if (zk.exists(path, false) == null) {<a name="line.358"></a> |
| <span class="sourceLineNo">359</span> createZNode(path, new byte[0]);<a name="line.359"></a> |
| <span class="sourceLineNo">360</span> }<a name="line.360"></a> |
| <span class="sourceLineNo">361</span> } catch (KeeperException | InterruptedException e) {<a name="line.361"></a> |
| <span class="sourceLineNo">362</span> LOG.error("Error checking given node : " + path + " " + e);<a name="line.362"></a> |
| <span class="sourceLineNo">363</span> }<a name="line.363"></a> |
| <span class="sourceLineNo">364</span> }<a name="line.364"></a> |
| <span class="sourceLineNo">365</span><a name="line.365"></a> |
| <span class="sourceLineNo">366</span> /**<a name="line.366"></a> |
| <span class="sourceLineNo">367</span> * sets given Status for Task Znode<a name="line.367"></a> |
| <span class="sourceLineNo">368</span> * @param taskZNode ZNode to set status<a name="line.368"></a> |
| <span class="sourceLineNo">369</span> * @param status Status value<a name="line.369"></a> |
| <span class="sourceLineNo">370</span> */<a name="line.370"></a> |
| <span class="sourceLineNo">371</span> public void setStatusOfTaskZNode(String taskZNode, String status) {<a name="line.371"></a> |
| <span class="sourceLineNo">372</span> LOG.info("Setting status of Task ZNode: " + taskZNode + " status : " + status);<a name="line.372"></a> |
| <span class="sourceLineNo">373</span> zk.setData(taskZNode, status.getBytes(StandardCharsets.UTF_8), -1, setStatusOfTaskZNodeCallback,<a name="line.373"></a> |
| <span class="sourceLineNo">374</span> null);<a name="line.374"></a> |
| <span class="sourceLineNo">375</span> }<a name="line.375"></a> |
| <span class="sourceLineNo">376</span><a name="line.376"></a> |
| <span class="sourceLineNo">377</span> /**<a name="line.377"></a> |
| <span class="sourceLineNo">378</span> * registration of ChaosAgent by checking and creating necessary ZNodes.<a name="line.378"></a> |
| <span class="sourceLineNo">379</span> */<a name="line.379"></a> |
| <span class="sourceLineNo">380</span> private void register() {<a name="line.380"></a> |
| <span class="sourceLineNo">381</span> createIfZNodeNotExists(ChaosConstants.CHAOS_TEST_ROOT_ZNODE);<a name="line.381"></a> |
| <span class="sourceLineNo">382</span> createIfZNodeNotExists(ChaosConstants.CHAOS_AGENT_REGISTRATION_EPIMERAL_ZNODE);<a name="line.382"></a> |
| <span class="sourceLineNo">383</span> createIfZNodeNotExists(ChaosConstants.CHAOS_AGENT_STATUS_PERSISTENT_ZNODE);<a name="line.383"></a> |
| <span class="sourceLineNo">384</span> createIfZNodeNotExists(ChaosConstants.CHAOS_AGENT_STATUS_PERSISTENT_ZNODE<a name="line.384"></a> |
| <span class="sourceLineNo">385</span> + ChaosConstants.ZNODE_PATH_SEPARATOR + agentName);<a name="line.385"></a> |
| <span class="sourceLineNo">386</span><a name="line.386"></a> |
| <span class="sourceLineNo">387</span> createEphemeralZNode(ChaosConstants.CHAOS_AGENT_REGISTRATION_EPIMERAL_ZNODE<a name="line.387"></a> |
| <span class="sourceLineNo">388</span> + ChaosConstants.ZNODE_PATH_SEPARATOR + agentName, new byte[0]);<a name="line.388"></a> |
| <span class="sourceLineNo">389</span> }<a name="line.389"></a> |
| <span class="sourceLineNo">390</span><a name="line.390"></a> |
| <span class="sourceLineNo">391</span> /***<a name="line.391"></a> |
| <span class="sourceLineNo">392</span> * Gets tasks for execution, basically sets Watch on it's respective host's Znode and waits for<a name="line.392"></a> |
| <span class="sourceLineNo">393</span> * tasks to be assigned, also has a getTasksForAgentCallback which handles execution of task.<a name="line.393"></a> |
| <span class="sourceLineNo">394</span> */<a name="line.394"></a> |
| <span class="sourceLineNo">395</span> private void getTasks() {<a name="line.395"></a> |
| <span class="sourceLineNo">396</span> LOG.info("Getting Tasks for Agent: " + agentName + "and setting watch for new Tasks");<a name="line.396"></a> |
| <span class="sourceLineNo">397</span> zk.getChildren(ChaosConstants.CHAOS_AGENT_STATUS_PERSISTENT_ZNODE<a name="line.397"></a> |
| <span class="sourceLineNo">398</span> + ChaosConstants.ZNODE_PATH_SEPARATOR + agentName, newTaskCreatedWatcher,<a name="line.398"></a> |
| <span class="sourceLineNo">399</span> getTasksForAgentCallback, null);<a name="line.399"></a> |
| <span class="sourceLineNo">400</span> }<a name="line.400"></a> |
| <span class="sourceLineNo">401</span><a name="line.401"></a> |
| <span class="sourceLineNo">402</span> /**<a name="line.402"></a> |
| <span class="sourceLineNo">403</span> * Below function executes command with retries with given user. Uses LocalShell to execute a<a name="line.403"></a> |
| <span class="sourceLineNo">404</span> * command.<a name="line.404"></a> |
| <span class="sourceLineNo">405</span> * @param user user name, default none<a name="line.405"></a> |
| <span class="sourceLineNo">406</span> * @param cmd Command to execute<a name="line.406"></a> |
| <span class="sourceLineNo">407</span> * @return A pair of Exit Code and Shell output<a name="line.407"></a> |
| <span class="sourceLineNo">408</span> * @throws IOException Exception while executing shell command<a name="line.408"></a> |
| <span class="sourceLineNo">409</span> */<a name="line.409"></a> |
| <span class="sourceLineNo">410</span> private Pair<Integer, String> execWithRetries(String user, String cmd) throws IOException {<a name="line.410"></a> |
| <span class="sourceLineNo">411</span> RetryCounter retryCounter = retryCounterFactory.create();<a name="line.411"></a> |
| <span class="sourceLineNo">412</span> while (true) {<a name="line.412"></a> |
| <span class="sourceLineNo">413</span> try {<a name="line.413"></a> |
| <span class="sourceLineNo">414</span> return exec(user, cmd);<a name="line.414"></a> |
| <span class="sourceLineNo">415</span> } catch (IOException e) {<a name="line.415"></a> |
| <span class="sourceLineNo">416</span> retryOrThrow(retryCounter, e, user, cmd);<a name="line.416"></a> |
| <span class="sourceLineNo">417</span> }<a name="line.417"></a> |
| <span class="sourceLineNo">418</span> try {<a name="line.418"></a> |
| <span class="sourceLineNo">419</span> retryCounter.sleepUntilNextRetry();<a name="line.419"></a> |
| <span class="sourceLineNo">420</span> } catch (InterruptedException e) {<a name="line.420"></a> |
| <span class="sourceLineNo">421</span> LOG.warn("Sleep Interrupted: " + e);<a name="line.421"></a> |
| <span class="sourceLineNo">422</span> }<a name="line.422"></a> |
| <span class="sourceLineNo">423</span> }<a name="line.423"></a> |
| <span class="sourceLineNo">424</span> }<a name="line.424"></a> |
| <span class="sourceLineNo">425</span><a name="line.425"></a> |
| <span class="sourceLineNo">426</span> private Pair<Integer, String> exec(String user, String cmd) throws IOException {<a name="line.426"></a> |
| <span class="sourceLineNo">427</span> LOG.info("Executing Shell command: " + cmd + " , user: " + user);<a name="line.427"></a> |
| <span class="sourceLineNo">428</span><a name="line.428"></a> |
| <span class="sourceLineNo">429</span> LocalShell shell = new LocalShell(user, cmd);<a name="line.429"></a> |
| <span class="sourceLineNo">430</span> try {<a name="line.430"></a> |
| <span class="sourceLineNo">431</span> shell.execute();<a name="line.431"></a> |
| <span class="sourceLineNo">432</span> } catch (Shell.ExitCodeException e) {<a name="line.432"></a> |
| <span class="sourceLineNo">433</span> String output = shell.getOutput();<a name="line.433"></a> |
| <span class="sourceLineNo">434</span> throw new Shell.ExitCodeException(e.getExitCode(),<a name="line.434"></a> |
| <span class="sourceLineNo">435</span> "stderr: " + e.getMessage() + ", stdout: " + output);<a name="line.435"></a> |
| <span class="sourceLineNo">436</span> }<a name="line.436"></a> |
| <span class="sourceLineNo">437</span> LOG.info("Executed Shell command, exit code: {}, output n{}", shell.getExitCode(),<a name="line.437"></a> |
| <span class="sourceLineNo">438</span> shell.getOutput());<a name="line.438"></a> |
| <span class="sourceLineNo">439</span><a name="line.439"></a> |
| <span class="sourceLineNo">440</span> return new Pair<>(shell.getExitCode(), shell.getOutput());<a name="line.440"></a> |
| <span class="sourceLineNo">441</span> }<a name="line.441"></a> |
| <span class="sourceLineNo">442</span><a name="line.442"></a> |
| <span class="sourceLineNo">443</span> private <E extends Exception> void retryOrThrow(RetryCounter retryCounter, E ex, String user,<a name="line.443"></a> |
| <span class="sourceLineNo">444</span> String cmd) throws E {<a name="line.444"></a> |
| <span class="sourceLineNo">445</span> if (retryCounter.shouldRetry()) {<a name="line.445"></a> |
| <span class="sourceLineNo">446</span> LOG.warn(<a name="line.446"></a> |
| <span class="sourceLineNo">447</span> "Local command: {}, user: {}, failed at attempt {}. Retrying until maxAttempts: {}."<a name="line.447"></a> |
| <span class="sourceLineNo">448</span> + "Exception {}",<a name="line.448"></a> |
| <span class="sourceLineNo">449</span> cmd, user, retryCounter.getAttemptTimes(), retryCounter.getMaxAttempts(), ex.getMessage());<a name="line.449"></a> |
| <span class="sourceLineNo">450</span> return;<a name="line.450"></a> |
| <span class="sourceLineNo">451</span> }<a name="line.451"></a> |
| <span class="sourceLineNo">452</span> throw ex;<a name="line.452"></a> |
| <span class="sourceLineNo">453</span> }<a name="line.453"></a> |
| <span class="sourceLineNo">454</span><a name="line.454"></a> |
| <span class="sourceLineNo">455</span> private boolean isConnected() {<a name="line.455"></a> |
| <span class="sourceLineNo">456</span> return connected;<a name="line.456"></a> |
| <span class="sourceLineNo">457</span> }<a name="line.457"></a> |
| <span class="sourceLineNo">458</span><a name="line.458"></a> |
| <span class="sourceLineNo">459</span> @Override<a name="line.459"></a> |
| <span class="sourceLineNo">460</span> public void close() throws IOException {<a name="line.460"></a> |
| <span class="sourceLineNo">461</span> LOG.info("Closing ZooKeeper Connection for Chaos Agent : " + agentName);<a name="line.461"></a> |
| <span class="sourceLineNo">462</span> try {<a name="line.462"></a> |
| <span class="sourceLineNo">463</span> zk.close();<a name="line.463"></a> |
| <span class="sourceLineNo">464</span> } catch (InterruptedException e) {<a name="line.464"></a> |
| <span class="sourceLineNo">465</span> LOG.error("Error while closing ZooKeeper Connection.");<a name="line.465"></a> |
| <span class="sourceLineNo">466</span> }<a name="line.466"></a> |
| <span class="sourceLineNo">467</span> }<a name="line.467"></a> |
| <span class="sourceLineNo">468</span><a name="line.468"></a> |
| <span class="sourceLineNo">469</span> @Override<a name="line.469"></a> |
| <span class="sourceLineNo">470</span> public void run() {<a name="line.470"></a> |
| <span class="sourceLineNo">471</span> try {<a name="line.471"></a> |
| <span class="sourceLineNo">472</span> LOG.info("Running Chaos Agent on : " + agentName);<a name="line.472"></a> |
| <span class="sourceLineNo">473</span> while (!this.isConnected()) {<a name="line.473"></a> |
| <span class="sourceLineNo">474</span> Thread.sleep(100);<a name="line.474"></a> |
| <span class="sourceLineNo">475</span> }<a name="line.475"></a> |
| <span class="sourceLineNo">476</span> this.getTasks();<a name="line.476"></a> |
| <span class="sourceLineNo">477</span> while (!stopChaosAgent.get()) {<a name="line.477"></a> |
| <span class="sourceLineNo">478</span> Thread.sleep(500);<a name="line.478"></a> |
| <span class="sourceLineNo">479</span> }<a name="line.479"></a> |
| <span class="sourceLineNo">480</span> } catch (InterruptedException e) {<a name="line.480"></a> |
| <span class="sourceLineNo">481</span> LOG.error("Error while running Chaos Agent", e);<a name="line.481"></a> |
| <span class="sourceLineNo">482</span> }<a name="line.482"></a> |
| <span class="sourceLineNo">483</span><a name="line.483"></a> |
| <span class="sourceLineNo">484</span> }<a name="line.484"></a> |
| <span class="sourceLineNo">485</span><a name="line.485"></a> |
| <span class="sourceLineNo">486</span> @Override<a name="line.486"></a> |
| <span class="sourceLineNo">487</span> public void process(WatchedEvent watchedEvent) {<a name="line.487"></a> |
| <span class="sourceLineNo">488</span> LOG.info("Processing event: " + watchedEvent.toString());<a name="line.488"></a> |
| <span class="sourceLineNo">489</span> if (watchedEvent.getType() == Event.EventType.None) {<a name="line.489"></a> |
| <span class="sourceLineNo">490</span> switch (watchedEvent.getState()) {<a name="line.490"></a> |
| <span class="sourceLineNo">491</span> case SyncConnected:<a name="line.491"></a> |
| <span class="sourceLineNo">492</span> connected = true;<a name="line.492"></a> |
| <span class="sourceLineNo">493</span> break;<a name="line.493"></a> |
| <span class="sourceLineNo">494</span> case Disconnected:<a name="line.494"></a> |
| <span class="sourceLineNo">495</span> connected = false;<a name="line.495"></a> |
| <span class="sourceLineNo">496</span> break;<a name="line.496"></a> |
| <span class="sourceLineNo">497</span> case Expired:<a name="line.497"></a> |
| <span class="sourceLineNo">498</span> connected = false;<a name="line.498"></a> |
| <span class="sourceLineNo">499</span> LOG.error("Session expired creating again");<a name="line.499"></a> |
| <span class="sourceLineNo">500</span> try {<a name="line.500"></a> |
| <span class="sourceLineNo">501</span> createZKConnection(null);<a name="line.501"></a> |
| <span class="sourceLineNo">502</span> } catch (IOException e) {<a name="line.502"></a> |
| <span class="sourceLineNo">503</span> LOG.error("Error creating Zookeeper connection", e);<a name="line.503"></a> |
| <span class="sourceLineNo">504</span> }<a name="line.504"></a> |
| <span class="sourceLineNo">505</span> default:<a name="line.505"></a> |
| <span class="sourceLineNo">506</span> LOG.error("Unknown State");<a name="line.506"></a> |
| <span class="sourceLineNo">507</span> break;<a name="line.507"></a> |
| <span class="sourceLineNo">508</span> }<a name="line.508"></a> |
| <span class="sourceLineNo">509</span> }<a name="line.509"></a> |
| <span class="sourceLineNo">510</span> }<a name="line.510"></a> |
| <span class="sourceLineNo">511</span><a name="line.511"></a> |
| <span class="sourceLineNo">512</span> private void recreateZKConnection() throws Exception {<a name="line.512"></a> |
| <span class="sourceLineNo">513</span> try {<a name="line.513"></a> |
| <span class="sourceLineNo">514</span> zk.close();<a name="line.514"></a> |
| <span class="sourceLineNo">515</span> createZKConnection(newTaskCreatedWatcher);<a name="line.515"></a> |
| <span class="sourceLineNo">516</span> createEphemeralZNode(ChaosConstants.CHAOS_AGENT_REGISTRATION_EPIMERAL_ZNODE<a name="line.516"></a> |
| <span class="sourceLineNo">517</span> + ChaosConstants.ZNODE_PATH_SEPARATOR + agentName, new byte[0]);<a name="line.517"></a> |
| <span class="sourceLineNo">518</span> } catch (IOException e) {<a name="line.518"></a> |
| <span class="sourceLineNo">519</span> LOG.error("Error creating new ZK COnnection for agent: {}", agentName + e);<a name="line.519"></a> |
| <span class="sourceLineNo">520</span> throw e;<a name="line.520"></a> |
| <span class="sourceLineNo">521</span> }<a name="line.521"></a> |
| <span class="sourceLineNo">522</span> }<a name="line.522"></a> |
| <span class="sourceLineNo">523</span><a name="line.523"></a> |
| <span class="sourceLineNo">524</span> /**<a name="line.524"></a> |
| <span class="sourceLineNo">525</span> * Executes Command locally.<a name="line.525"></a> |
| <span class="sourceLineNo">526</span> */<a name="line.526"></a> |
| <span class="sourceLineNo">527</span> protected static class LocalShell extends Shell.ShellCommandExecutor {<a name="line.527"></a> |
| <span class="sourceLineNo">528</span><a name="line.528"></a> |
| <span class="sourceLineNo">529</span> private String user;<a name="line.529"></a> |
| <span class="sourceLineNo">530</span> private String execCommand;<a name="line.530"></a> |
| <span class="sourceLineNo">531</span><a name="line.531"></a> |
| <span class="sourceLineNo">532</span> public LocalShell(String user, String execCommand) {<a name="line.532"></a> |
| <span class="sourceLineNo">533</span> super(new String[] { execCommand });<a name="line.533"></a> |
| <span class="sourceLineNo">534</span> this.user = user;<a name="line.534"></a> |
| <span class="sourceLineNo">535</span> this.execCommand = execCommand;<a name="line.535"></a> |
| <span class="sourceLineNo">536</span> }<a name="line.536"></a> |
| <span class="sourceLineNo">537</span><a name="line.537"></a> |
| <span class="sourceLineNo">538</span> @Override<a name="line.538"></a> |
| <span class="sourceLineNo">539</span> public String[] getExecString() {<a name="line.539"></a> |
| <span class="sourceLineNo">540</span> // TODO: Considering Agent is running with same user.<a name="line.540"></a> |
| <span class="sourceLineNo">541</span> if (!user.equals(ChaosConstants.DEFAULT_SHELL_USER)) {<a name="line.541"></a> |
| <span class="sourceLineNo">542</span> execCommand = String.format("su -u %1$s %2$s", user, execCommand);<a name="line.542"></a> |
| <span class="sourceLineNo">543</span> }<a name="line.543"></a> |
| <span class="sourceLineNo">544</span> return new String[] { "/usr/bin/env", "bash", "-c", execCommand };<a name="line.544"></a> |
| <span class="sourceLineNo">545</span> }<a name="line.545"></a> |
| <span class="sourceLineNo">546</span><a name="line.546"></a> |
| <span class="sourceLineNo">547</span> @Override<a name="line.547"></a> |
| <span class="sourceLineNo">548</span> public void execute() throws IOException {<a name="line.548"></a> |
| <span class="sourceLineNo">549</span> super.execute();<a name="line.549"></a> |
| <span class="sourceLineNo">550</span> }<a name="line.550"></a> |
| <span class="sourceLineNo">551</span> }<a name="line.551"></a> |
| <span class="sourceLineNo">552</span>}<a name="line.552"></a> |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| </pre> |
| </div> |
| </body> |
| </html> |