libdb/docs/gsg_db_rep/CXX/fwrkmasterreplica.html
2012-11-14 16:35:20 -05:00

449 lines
18 KiB
HTML
Raw Blame History

This file contains invisible Unicode characters

This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

<?xml version="1.0" encoding="UTF-8" standalone="no"?>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html xmlns="http://www.w3.org/1999/xhtml">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
<title>Chapter 4. Replica versus Master Processes</title>
<link rel="stylesheet" href="gettingStarted.css" type="text/css" />
<meta name="generator" content="DocBook XSL Stylesheets V1.73.2" />
<link rel="start" href="index.html" title="Getting Started with Replicated Berkeley DB Applications" />
<link rel="up" href="index.html" title="Getting Started with Replicated Berkeley DB Applications" />
<link rel="prev" href="heartbeats.html" title="Managing Heartbeats" />
<link rel="next" href="processingloop.html" title="Processing Loop" />
</head>
<body>
<div xmlns="" class="navheader">
<div class="libver">
<p>Library Version 11.2.5.3</p>
</div>
<table width="100%" summary="Navigation header">
<tr>
<th colspan="3" align="center">Chapter 4. Replica versus Master Processes</th>
</tr>
<tr>
<td width="20%" align="left"><a accesskey="p" href="heartbeats.html">Prev</a> </td>
<th width="60%" align="center"> </th>
<td width="20%" align="right"> <a accesskey="n" href="processingloop.html">Next</a></td>
</tr>
</table>
<hr />
</div>
<div class="chapter" lang="en" xml:lang="en">
<div class="titlepage">
<div>
<div>
<h2 class="title"><a id="fwrkmasterreplica"></a>Chapter 4. Replica versus Master Processes</h2>
</div>
</div>
</div>
<div class="toc">
<p>
<b>Table of Contents</b>
</p>
<dl>
<dt>
<span class="sect1">
<a href="fwrkmasterreplica.html#determinestate">Determining State</a>
</span>
</dt>
<dt>
<span class="sect1">
<a href="processingloop.html">Processing Loop</a>
</span>
</dt>
<dt>
<span class="sect1">
<a href="exampledoloop.html">Example Processing Loop</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="exampledoloop.html#runningit">Running It</a>
</span>
</dt>
</dl>
</dd>
</dl>
</div>
<p>
Every environment participating in a replicated application
must know whether it is a <span class="emphasis"><em>master</em></span> or
<span class="emphasis"><em>replica</em></span>. The reason for this is
because, simply, the master can modify the database while
replicas cannot. As a result, not only will you open
databases differently depended on whether the environment is
running as a master, but the environment will frequently
behave quite a bit differently depending on whether it
thinks it is operating as the read/write interface for
your database.
</p>
<p>
Moreover, an environment must also be capable of
gracefully switching between master and replica states.
This means that the environment must be able to detect when
it has switched states.
</p>
<p>
Not surprisingly, a large part of your application's code
will be tied up in knowing which state a given
environment is in and then in the logic of how to behave depending on
its state.
</p>
<p>
This chapter shows you how to determine your environment's
state, and it then shows you some sample code on how
an application might behave depending on whether it is a
master or a replica in a replicated application.
</p>
<div class="sect1" lang="en" xml:lang="en">
<div class="titlepage">
<div>
<div>
<h2 class="title" style="clear: both"><a id="determinestate"></a>Determining State</h2>
</div>
</div>
</div>
<p>
In order to determine whether your code is
running as a master or a replica, you implement a
callback whose function it is to respond to
events that happen within the DB library.
Note that these events are raised whenever the state is established.
For example, when the current environment becomes a client —
including at application startup — the
<code class="literal">DB_EVENT_REP_CLIENT</code> event is raised. Also, when an
election is held and a replica is elected to be a master, the
<code class="literal">DB_EVENT_REP_MASTER</code> event is raised on the newly elected master and the
<code class="literal">DB_EVENT_REP_NEWMASTER</code> is raised on the other replicas.
</p>
<p>
Note that this callback is usable for events beyond
those required for replication purposes. In this
section, however, we only discuss the
replication-specific events.
</p>
<p>
The callback is required to determine
which event has been passed to it, and then take
action depending on the event. For replication,
the events that we care about are:
</p>
<p>
Some of the more commonly handled events are
described below. For a complete list of events, see
the
<span>
<code class="methodname">DbEnv::set_event_notify()</code>
method in the <em class="citetitle">Berkeley DB C++ API Reference Guide</em>.
</span>
</p>
<div class="itemizedlist">
<ul type="disc">
<li>
<p>
<code class="literal">DB_EVENT_REP_CLIENT</code>
</p>
<p>
The local environment is now a replica.
</p>
</li>
<li>
<p>
<code class="literal">DB_EVENT_REP_CONNECT_BROKEN</code>
</p>
<p>
A previously established connection between
two sites in the replication group has been
broken.
</p>
</li>
<li>
<p>
<code class="literal">DB_EVENT_REP_CONNECT_ESTD</code>
</p>
<p>
A connection has been established between
two sites in the replication group.
</p>
</li>
<li>
<p>
<code class="literal">DB_EVENT_REP_CONNECT_RETRY_ESTABLISHED</code>
</p>
<p>
An attempt was made to establish a connection
to a known remote site, but the connection
attempt failed.
</p>
</li>
<li>
<p>
<code class="literal">DB_EVENT_REP_DUPMASTER</code>
</p>
<p>
A duplicate master has been discovered in the
replication group.
</p>
</li>
<li>
<p>
<code class="literal">DB_EVENT_REP_ELECTED</code>
</p>
<p>
The local site has just won an election and
is now the master. Your code should now
reconfigure itself to operation as a master
site.
</p>
</li>
<li>
<p>
<code class="literal">DB_EVENT_REP_ELECTION_FAILED</code>
</p>
<p>
The local site's attempt to initiate or
participate in a replication master election
failed, due to the lack of timely message
response from a sufficient number of remote
sites.
</p>
</li>
<li>
<p>
<code class="literal">DB_EVENT_REP_ELECTION_STARTED</code>
</p>
<p>
Replication Manager has started an election to choose a master site.
</p>
</li>
<li>
<p>
<code class="literal">DB_EVENT_REP_LOCAL_SITE_REMOVED</code>
</p>
<p>
The local site has been removed from the group.
</p>
</li>
<li>
<p>
<code class="literal">DB_EVENT_REP_NEWMASTER</code>
</p>
<p>
An election was held and a new environment was made a
master. However, the current environment <span class="emphasis"><em>is
not</em></span> the master. This event exists so that
you can cause your code to take some unique action in the
event that the replication groups switches masters.
</p>
</li>
<li>
<p>
<code class="literal">DB_EVENT_REP_MASTER</code>
</p>
<p>
The local environment is now a master.
</p>
</li>
<li>
<p>
<code class="literal">DB_EVENT_REP_MASTER_FAILURE</code>
</p>
<p>
The connection to the remote master replication site has failed.
</p>
</li>
<li>
<p>
<code class="literal">DB_EVENT_REP_PERM_FAILED</code>
</p>
<p>
The Replication Manager did not receive enough acknowledgements to
ensure the transaction's durability within the replicationg group.
The Replication Manager has therefore flushed the transaction to
the master's local disk for storage.
</p>
<p>
How the Replication Manager knows whether the acknowledgements it
has received is determined by the ack policy you have set for your
applicaton. See <a class="xref" href="fwrkpermmessage.html#fmwrkpermpolicy" title="Identifying Permanent Message Policies">Identifying Permanent Message Policies</a>
for more information.
</p>
</li>
<li>
<p>
<code class="literal">DB_EVENT_REP_SITE_ADDED</code>
</p>
<p>
A new site has joined the replication group.
</p>
</li>
<li>
<p>
<code class="literal">DB_EVENT_REP_SITE_REMOVED</code>
</p>
<p>
An existing site has been removed from the replication group.
</p>
</li>
<li>
<p>
<code class="literal">DB_EVENT_REP_STARTUPDONE</code>
</p>
<p>
The replica has completed startup
synchronization and is now processing
log records received from the master.
</p>
</li>
<li>
<p>
<code class="literal">DB_EVENT_WRITE_FAILED</code>
</p>
<p>
A Berkeley DB write to stable storage failed.
</p>
</li>
</ul>
</div>
<p>
Note that these events are raised whenever the
state is established. That is, when the current
environment becomes a replica, and that includes
at application startup, the event is raised.
Also, when an election is held and a replica is elected to be a
master, then the event occurs.
</p>
<p>
The implementation of this callback is fairly
simple. First you pass a structure to the
environment handle that you can use to record the
environment's state, and then you implement a switch
statement within the callback that you use to
record the current state, depending on the
arriving event.
</p>
<p>
For example:
</p>
<pre class="programlisting">#include &lt;db_cxx.h&gt;
/* Forward declaration */
void *event_callback(DbEnv *, u_int32_t, void *);
...
/* The structure we use to track our environment's state */
typedef struct {
int is_master;
} APP_DATA;
...
/*
* Inside our main() function, we declare an APP_DATA variable.
*/
APP_DATA my_app_data;
my_app_data.is_master = 0; /* Assume we start as a replica */
...
/*
* Now we open our environment handle and set the APP_DATA structure
* to it's app_private member.
*/
DbEnv *dbenv = new DbEnv(0);
dbenv-&gt;set_app_private(&amp;my_app_data);
/* Having done that, register the callback with the
* Berkeley DB library
*/
dbenv-&gt;set_event_notify(event_callback); </pre>
<p>
That done, we still need to implement the callback itself. This
implementation can be fairly trivial.
</p>
<pre class="programlisting">/*
* A callback used to determine whether the local environment is a
* replica or a master. This is called by the Replication Manager
* when the local environment changes state.
*/
void *
event_callback(DbEnv *dbenv, u_int32_t which, void *info)
{
APP_DATA *app = dbenv-&gt;get_app_private();
info = NULL; /* Currently unused. */
switch (which) {
case DB_EVENT_REP_MASTER:
app-&gt;is_master = 1;
break;
case DB_EVENT_REP_CLIENT:
app-&gt;is_master = 0;
break;
case DB_EVENT_REP_STARTUPDONE: /* fallthrough */
case DB_EVENT_REP_NEWMASTER:
/* Ignore. */
break;
default:
dbenv-&gt;errx(dbenv, "ignoring event %d", which);
}
} </pre>
<p>
Notice how we access the
<code class="literal">APP_DATA</code> information using the environment
handle's <code class="literal">app_private</code> data member. We also ignore
the <code class="literal">DB_EVENT_REP_NEWMASTER</code> and
<code class="literal">DB_EVENT_REP_STARTUPDONE</code> cases since these are not
relevant for simple replicated applications.
</p>
<p>
Of course, this only gives us the current state of the environment. We
still need the code that determines what to do when the environment
changes state and how to behave depending on the state (described
in the next section).
</p>
</div>
</div>
<div class="navfooter">
<hr />
<table width="100%" summary="Navigation footer">
<tr>
<td width="40%" align="left"><a accesskey="p" href="heartbeats.html">Prev</a> </td>
<td width="20%" align="center"> </td>
<td width="40%" align="right"> <a accesskey="n" href="processingloop.html">Next</a></td>
</tr>
<tr>
<td width="40%" align="left" valign="top">Managing Heartbeats </td>
<td width="20%" align="center">
<a accesskey="h" href="index.html">Home</a>
</td>
<td width="40%" align="right" valign="top"> Processing Loop</td>
</tr>
</table>
</div>
</body>
</html>