1<!--$Id: rep_elect.so,v 1.28 2007/12/03 21:01:23 sue Exp $--> 2<!--Copyright (c) 1997,2008 Oracle. All rights reserved.--> 3<!--See the file LICENSE for redistribution information.--> 4<html> 5<head> 6<title>Berkeley DB: DB_ENV->rep_elect</title> 7<meta name="description" content="Berkeley DB: An embedded database programmatic toolkit."> 8<meta name="keywords" content="embedded,database,programmatic,toolkit,btree,hash,hashing,transaction,transactions,locking,logging,access method,access methods,Java,C,C++"> 9</head> 10<body bgcolor=white> 11<table width="100%"><tr valign=top> 12<td> 13<b>DB_ENV->rep_elect</b> 14</td> 15<td align=right> 16<a href="../api_c/api_core.html"><img src="../images/api.gif" alt="API"></a> 17<a href="../ref/toc.html"><img src="../images/ref.gif" alt="Ref"></a></td> 18</tr></table> 19<hr size=1 noshade> 20<tt> 21<b><pre> 22#include <db.h> 23<p> 24int 25DB_ENV->rep_elect(DB_ENV *env, 26 u_int32_t nsites, u_int32_t nvotes, u_int32_t flags); 27</pre></b> 28<hr size=1 noshade> 29<b>Description: DB_ENV->rep_elect</b> 30<p>The DB_ENV->rep_elect method holds an election for the master of a replication 31group.</p> 32<p>The DB_ENV->rep_elect method is not called by most replication 33applications. It should only be called by applications implementing 34their own network transport layer, explicitly holding replication group 35elections and handling replication messages outside of the replication 36manager framework.</p> 37<p>If the election is successful, Berkeley DB will notify the application of 38the results of the election by means of either the 39<a href="../api_c/env_event_notify.html#DB_EVENT_REP_ELECTED">DB_EVENT_REP_ELECTED</a> or <a href="../api_c/env_event_notify.html#DB_EVENT_REP_NEWMASTER">DB_EVENT_REP_NEWMASTER</a> events 40(see <a href="../api_c/env_event_notify.html">DB_ENV->set_event_notify</a> method for more information). The 41application is responsible for adjusting its relationship to the other 42database environments in the replication group, including directing 43all database updates to the newly selected master, in accordance with 44the results of the election.</p> 45<p>The thread of control that calls the DB_ENV->rep_elect method must not be the 46thread of control that processes incoming messages; processing the 47incoming messages is necessary to successfully complete an election.</p> 48<b>Parameters</b> <br> 49 <b>nsites</b><ul compact><li>The <b>nsites</b> parameter specifies the number of replication sites 50expected to participate in the election. Once the current site has 51election information from that many sites, it will short-circuit the 52election and immediately cast its vote for a new master. The 53<b>nsites</b> parameter must be no less than 54<b>nvotes</b>, or 0 if the election should use the value previously set 55using the <a href="../api_c/rep_nsites.html">DB_ENV->rep_set_nsites</a> method. If an application is using master 56leases, then the value <b>must</b> be 0 and the value from 57<a href="../api_c/rep_nsites.html">DB_ENV->rep_set_nsites</a> method must be used.</ul> 58 <b>nvotes</b><ul compact><li>The <b>nvotes</b> parameter specifies the minimum number of replication 59sites from which the current site must have election information, before 60the current site will cast a vote for a new master. The <b>nvotes</b> 61parameter must be no greater than <b>nsites</b>, 62or 0 if the election should use the value ((<b>nsites</b> / 2) + 1) as 63the <b>nvotes</b> argument.</ul> 64 <b>flags</b><ul compact><li>The <b>flags</b> parameter is currently unused, and must be set to 0.</ul> 65<br> 66<p>Elections are done in two parts: first, replication sites collect 67information from the other replication sites they know about, and 68second, replication sites cast their votes for a new master. The second 69phase is triggered by one of two things: either the replication site 70gets election information from <b>nsites</b> sites, or the election 71<b>timeout</b> expires. Once the second phase is triggered, the 72replication site will cast a vote for the new master of its choice if, 73and only if, the site has election information from at least 74<b>nvotes</b> sites. If a site receives <b>nvotes</b> votes for it 75to become the new master, then it will become the new master.</p> 76<p>We recommend <b>nvotes</b> be set to at least:</p> 77<blockquote><pre>(sites participating in the election / 2) + 1</pre></blockquote> 78<p>to ensure there are never more than two masters active at the same time 79even in the case of a network partition. When a network partitions, the 80side of the partition with more than half the environments will elect a 81new master and continue, while the environments communicating with fewer 82than half of the environments will fail to find a new master, as no site 83can get <b>nvotes</b> votes.</p> 84<p>We recommend <b>nsites</b> be set to:</p> 85<blockquote><pre>number of sites in the replication group - 1</pre></blockquote> 86<p>when choosing a new master after a current master fails. This allows 87the group to reach a consensus without having to wait for the timeout 88to expire.</p> 89<p>When choosing a master from among a group of client 90sites all restarting at the same time, it makes more sense to set 91<b>nsites</b> to the total number of sites in the group, since there is no 92known missing site. Furthermore, in order to ensure the best choice 93from among sites that may take longer to boot than the local site, 94setting <b>nvotes</b> also to this same total number of sites will 95guarantee that every site in the group is considered. Alternatively, 96using the special timeout for full elections allows full participation 97on restart but allows election of a master if one site does not 98reboot and rejoin the group in a reasonable amount of time. (See the 99<a href="../ref/rep/elect.html">Elections</a> 100section in the Berkeley DB Reference Guide for more information.)</p> 101<p>Setting <b>nsites</b> to lower values can increase the speed of an 102election, but can also result in election failure, and is usually not 103recommended.</p> 104<a name="2"><!--meow--></a> 105<br><b>Errors</b> 106<p>The DB_ENV->rep_elect method 107may fail and return one of the following non-zero errors:</p> 108<br> 109<b>DB_REP_UNAVAIL</b><ul compact><li>The replication group was unable to elect a master, or was unable to 110complete the election in the election timeout period (see 111<a href="../api_c/rep_timeout.html">DB_ENV->rep_set_timeout</a> method for more information).</ul> 112<br> 113<hr size=1 noshade> 114<br><b>Class</b> 115<a href="../api_c/env_class.html">DB_ENV</a> 116<br><b>See Also</b> 117<a href="../api_c/rep_list.html">Replication and Related Methods</a> 118</tt> 119<table width="100%"><tr><td><br></td><td align=right> 120<a href="../api_c/api_core.html"><img src="../images/api.gif" alt="API"></a><a href="../ref/toc.html"><img src="../images/ref.gif" alt="Ref"></a> 121</td></tr></table> 122<p><font size=1>Copyright (c) 1996,2008 Oracle. All rights reserved.</font> 123</body> 124</html> 125
