1<?xml version="1.0" encoding="UTF-8" standalone="no"?> 2<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"> 3<html xmlns="http://www.w3.org/1999/xhtml"> 4 <head> 5 <meta http-equiv="Content-Type" content="text/html; charset=UTF-8" /> 6 <title>Chapter 4. Replica versus Master Processes</title> 7 <link rel="stylesheet" href="gettingStarted.css" type="text/css" /> 8 <meta name="generator" content="DocBook XSL Stylesheets V1.62.4" /> 9 <link rel="home" href="index.html" title="Getting Started with Replicated Berkeley DB Applications" /> 10 <link rel="up" href="index.html" title="Getting Started with Replicated Berkeley DB Applications" /> 11 <link rel="previous" href="heartbeats.html" title="Managing Heartbeats" /> 12 <link rel="next" href="processingloop.html" title="Processing Loop" /> 13 </head> 14 <body> 15 <div class="navheader"> 16 <table width="100%" summary="Navigation header"> 17 <tr> 18 <th colspan="3" align="center">Chapter 4. Replica versus Master Processes</th> 19 </tr> 20 <tr> 21 <td width="20%" align="left"><a accesskey="p" href="heartbeats.html">Prev</a> </td> 22 <th width="60%" align="center"> </th> 23 <td width="20%" align="right"> <a accesskey="n" href="processingloop.html">Next</a></td> 24 </tr> 25 </table> 26 <hr /> 27 </div> 28 <div class="chapter" lang="en" xml:lang="en"> 29 <div class="titlepage"> 30 <div> 31 <div> 32 <h2 class="title"><a id="fwrkmasterreplica"></a>Chapter 4. Replica versus Master Processes</h2> 33 </div> 34 </div> 35 <div></div> 36 </div> 37 <div class="toc"> 38 <p> 39 <b>Table of Contents</b> 40 </p> 41 <dl> 42 <dt> 43 <span class="sect1"> 44 <a href="fwrkmasterreplica.html#determinestate">Determining State</a> 45 </span> 46 </dt> 47 <dt> 48 <span class="sect1"> 49 <a href="processingloop.html">Processing Loop</a> 50 </span> 51 </dt> 52 <dt> 53 <span class="sect1"> 54 <a href="exampledoloop.html">Example Processing Loop</a> 55 </span> 56 </dt> 57 <dd> 58 <dl> 59 <dt> 60 <span class="sect2"> 61 <a href="exampledoloop.html#runningit">Running It</a> 62 </span> 63 </dt> 64 </dl> 65 </dd> 66 </dl> 67 </div> 68 <p> 69 Every environment participating in a replicated application 70 must know whether it is a <span class="emphasis"><em>master</em></span> or 71 <span class="emphasis"><em>replica</em></span>. The reason for this is 72 because, simply, the master can modify the database while 73 replicas cannot. As a result, not only will you open 74 databases differently depended on whether the environment is 75 running as a master, but the environment will frequently 76 behave quit a bit differently depending on whether it 77 thinks it is operating as the read/write interface for 78 your database. 79 </p> 80 <p> 81 Moreover, an environment must also be capable of 82 gracefully switching between master and replica states. 83 This means that the environment must be able to detect when 84 it has switched states. 85 </p> 86 <p> 87 Not surprisingly, a large part of your application's code 88 will be tied up in knowing which state a given 89 environment is in and then in the logic of how to behave depending on 90 its state. 91 </p> 92 <p> 93 This chapter shows you how to determine your environment's 94 state, and it then shows you some sample code on how 95 an application might behave depending on whether it is a 96 master or a replica in a replicated application. 97 </p> 98 <div class="sect1" lang="en" xml:lang="en"> 99 <div class="titlepage"> 100 <div> 101 <div> 102 <h2 class="title" style="clear: both"><a id="determinestate"></a>Determining State</h2> 103 </div> 104 </div> 105 <div></div> 106 </div> 107 <p> 108 In order to determine whether your code is 109 running as a master or a replica, you implement a 110 callback whose function it is to respond to 111 events that happen within the DB library. 112 </p> 113 <p> 114 Note that this callback is usable for events beyond 115 those required for replication purposes. In this 116 section, however, we only discuss the 117 replication-specific events. 118 </p> 119 <p> 120 The callback is required to determine 121 which event has been passed to it, and then take 122 action depending on the event. For replication, 123 the events that we care about are: 124 </p> 125 <div class="itemizedlist"> 126 <ul type="disc"> 127 <li> 128 <p> 129 <tt class="literal">DB_EVENT_REP_MASTER</tt> 130 131 </p> 132 <p> 133 The local environment is now a master. 134 </p> 135 </li> 136 <li> 137 <p> 138 <tt class="literal">DB_EVENT_REP_CLIENT</tt> 139 140 </p> 141 <p> 142 The local environment is now a replica. 143 </p> 144 </li> 145 <li> 146 <p> 147 <tt class="literal">DB_EVENT_REP_STARTUPDONE</tt> 148 149 </p> 150 <p> 151 The replica has completed startup 152 synchronization and is now 153 processing log records received 154 from the master. 155 </p> 156 </li> 157 <li> 158 <p> 159 <tt class="literal">DB_EVENT_REP_NEWMASTER</tt> 160 161 </p> 162 <p> 163 An election was held and a new 164 environment was made a master. However, 165 the current environment <span class="emphasis"><em>is 166 not</em></span> the master. This 167 event exists so that you can 168 cause your code to take some 169 unique action in the event that the 170 replication groups switches masters. 171 </p> 172 </li> 173 </ul> 174 </div> 175 <p> 176 Note that these events are raised whenever the 177 state is established. That is, when the current 178 environment becomes a client, and that includes 179 at application startup, the event is raised. 180 Also, when an election is held and a client is elected to be a 181 master, then the event occurs. 182 </p> 183 <p> 184 The implementation of this callback is fairly 185 simple. First you pass a structure to the 186 environment handle that you can use to record the 187 environment's state, and then you implement a switch 188 statement within the callback that you use to 189 record the current state, depending on the 190 arriving event. 191 </p> 192 <p> 193 For example: 194 </p> 195 <pre class="programlisting">#include <db.h> 196/* Forward declaration */ 197void *event_callback(DB_ENV *, u_int32_t, void *); 198 199... 200 201/* The structure we use to track our environment's state */ 202typedef struct { 203 int is_master; 204} APP_DATA; 205 206... 207 208/* 209 * Inside our main() function, we declare an APP_DATA variable. 210 */ 211APP_DATA my_app_data; 212my_app_data.is_master = 0; /* Assume we start as a replica */ 213 214... 215 216/* 217 * Now we create our environment handle and set the APP_DATA structure 218 * to it's app_private member. 219 */ 220if ((ret = db_env_create(&dbenv, 0)) != 0 ) { 221 fprintf(stderr, "Error creating handles: %s\n", 222 db_strerror(ret)); 223 goto err; 224} 225dbenv->app_private = &my_app_data; 226 227/* Having done that, register the callback with the 228 * Berkeley DB library 229 */ 230dbenv->set_event_notify(dbenv, event_callback); </pre> 231 <p> 232 That done, we still need to implement the callback itself. This 233 implementation can be fairly trivial. 234</p> 235 <pre class="programlisting">/* 236 * A callback used to determine whether the local environment is a replica 237 * or a master. This is called by the replication framework 238 * when the local environment changes state. 239 */ 240void * 241event_callback(DB_ENV *dbenv, u_int32_t which, void *info) 242{ 243 APP_DATA *app = dbenv->app_private; 244 245 info = NULL; /* Currently unused. */ 246 247 switch (which) { 248 case DB_EVENT_REP_MASTER: 249 app->is_master = 1; 250 break; 251 252 case DB_EVENT_REP_CLIENT: 253 app->is_master = 0; 254 break; 255 256 case DB_EVENT_REP_STARTUPDONE: /* fallthrough */ 257 case DB_EVENT_REP_NEWMASTER: 258 /* Ignore. */ 259 break; 260 261 default: 262 dbenv->errx(dbenv, "ignoring event %d", which); 263 } 264} </pre> 265 <p> 266 Notice how we access the 267 <tt class="literal">APP_DATA</tt> information using the environment 268 handle's <tt class="literal">app_private</tt> data member. We also ignore 269 the <tt class="literal">DB_EVENT_REP_NEWMASTER</tt> and 270 <tt class="literal">DB_EVENT_REP_STARTUPDONE</tt> cases since these are not 271 relevant for simple replicated applications. 272</p> 273 <p> 274 Of course, this only gives us the current state of the environment. We 275 still need the code that determines what to do when the environment 276 changes state and how to behave depending on the state (described 277 in the next section). 278</p> 279 </div> 280 </div> 281 <div class="navfooter"> 282 <hr /> 283 <table width="100%" summary="Navigation footer"> 284 <tr> 285 <td width="40%" align="left"><a accesskey="p" href="heartbeats.html">Prev</a> </td> 286 <td width="20%" align="center"> 287 <a accesskey="u" href="index.html">Up</a> 288 </td> 289 <td width="40%" align="right"> <a accesskey="n" href="processingloop.html">Next</a></td> 290 </tr> 291 <tr> 292 <td width="40%" align="left" valign="top">Managing Heartbeats </td> 293 <td width="20%" align="center"> 294 <a accesskey="h" href="index.html">Home</a> 295 </td> 296 <td width="40%" align="right" valign="top"> Processing Loop</td> 297 </tr> 298 </table> 299 </div> 300 </body> 301</html> 302