<?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>Read your writes consistency</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="Berkeley DB Programmer's Reference Guide" />
<link rel="up" href="rep.html" title="Chapter 12. Berkeley DB Replication" />
<link rel="prev" href="rep_lease.html" title="Master Leases" />
<link rel="next" href="rep_clock_skew.html" title="Clock Skew" />
</head>
<body>
<div class="navheader">
<table width="100%" summary="Navigation header">
<tr>
<th colspan="3" align="center">Read your writes consistency</th>
</tr>
<tr>
<td width="20%" align="left"><a accesskey="p" href="rep_lease.html">Prev</a> </td>
<th width="60%" align="center">Chapter 12.
Berkeley DB Replication
</th>
<td width="20%" align="right"> <a accesskey="n" href="rep_clock_skew.html">Next</a></td>
</tr>
</table>
<hr />
</div>
<div class="sect1" lang="en" xml:lang="en">
<div class="titlepage">
<div>
<div>
<h2 class="title" style="clear: both"><a id="rep_ryw"></a>Read your writes consistency</h2>
</div>
</div>
</div>
<div class="toc">
<dl>
<dt>
<span class="sect2">
<a href="rep_ryw.html#gettoken">Getting a token</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="rep_ryw.html#tokenhandling">Token handling</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="rep_ryw.html#usingtoken">Using a token to check or wait for a transaction</a>
</span>
</dt>
</dl>
</div>
<p>
Some applications require the ability to read replicated data at a
client site, and determine whether it is consistent with data that
has been written previously at the master site.
</p>
<p>
For example, a web application may be backed by multiple database
environments, linked to form a replication group, in order to share
the workload. Web requests that update data must be served by the
replication master, but any site in the group may serve a read-only
request. Consider a work flow of a series of web requests from one
specific user at a web browser: the first request generates a
database update, but the second request merely reads data. If the
read-only request is served by a replication client database
environment, it may be important to make sure that the updated data
has been replicated to the client before performing the read (or to wait
until it has been replicated) in order to show this user a
consistent view of the data.
</p>
<p>
Berkeley DB supports this requirement through the use of transaction
"tokens". A token is a form of identification for a transaction
within the scope of the replication group. The application may
request a copy of the transaction's token at the master site during
the execution of the transaction. Later, the application running
on a client site can use a copy of the token to determine whether
the transaction has been applied at that site.
</p>
<p>
It is the application's responsibility to keep track of the token
during the interim. In the web example, the token might be sent to
the browser as a "cookie", or stored on the application server in
the user's session context.
</p>
<p>
The operations described here are supported both for Replication
Manager applications and for applications that use the replication
Base API.
</p>
<div class="sect2" lang="en" xml:lang="en">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="gettoken"></a>Getting a token</h3>
</div>
</div>
</div>
<p>
In order to get a token, the application must supply a small
memory buffer, using the <a href="../api_reference/C/txnset_commit_token.html" class="olink">DB_TXN->set_commit_token()</a> method.
</p>
<p>
Note that a token is generated only upon a successful commit
operation, and therefore the token buffer content is valid
only after a successful commit. Also, if a transaction does
not perform any update operations it does not generate a useful
token.
</p>
<p>
In the Berkeley DB Java and C# API, getting a token is simpler.
The application need only invoke the <a class="ulink" href="../java/com/sleepycat/db/Transaction.html#getCommitToken()" target="_top">Transaction.getCommitToken()</a> method,
after the transaction has committed.
</p>
</div>
<div class="sect2" lang="en" xml:lang="en">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="tokenhandling"></a>Token handling</h3>
</div>
</div>
</div>
<p>
The application should not try to interpret the content of the
token buffer, but may store and/or transmit it freely between
systems. However, since the buffer contains binary data it
may be necessary to apply some encoding for transmission (e.g.,
base 64).
</p>
<p>
The data is resilient to differences in byte order between
different systems. It does not expire: it may be retained
indefinitely for later use, even across Berkeley DB version
upgrades.
</p>
</div>
<div class="sect2" lang="en" xml:lang="en">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="usingtoken"></a>Using a token to check or wait for a transaction</h3>
</div>
</div>
</div>
<p>
The <a href="../api_reference/C/envtxn_applied.html" class="olink">DB_ENV->txn_applied()</a> method takes a copy of a token, and
determines whether the corresponding transaction is currently
applied at the local site. The timeout argument allows the
application to block for a bounded amount of time for cases
where the transaction has not yet been applied.
</p>
<p>
Depending on the transaction durability levels implemented or
configured by the application, it is sometimes possible for a
transaction to disappear from a replication group if an
original master site fails and a different site becomes the new
master without having received the transaction. When the
<a href="../api_reference/C/envtxn_applied.html" class="olink">DB_ENV->txn_applied()</a> method discovers this, it produces the
<code class="literal">DB_NOTFOUND</code> return code.
</p>
<p>
This means that the results of <a href="../api_reference/C/envtxn_applied.html" class="olink">DB_ENV->txn_applied()</a> are not guaranteed
forever. Even after a successful call to <a href="../api_reference/C/envtxn_applied.html" class="olink">DB_ENV->txn_applied()</a>, it is
possible that by the time the application tries to read the
data, the transaction and its data could have disappeared.
</p>
<p>
To avoid this problem the application should do the read
operations in the context of a transaction, and hold the
transaction handle open during the <a href="../api_reference/C/envtxn_applied.html" class="olink">DB_ENV->txn_applied()</a> call. The
<a href="../api_reference/C/envtxn_applied.html" class="olink">DB_ENV->txn_applied()</a> method itself does not actually execute in the
context of the transaction; but no rollbacks due to new master
synchronization ever occur while a transaction is active, even
a read-only transaction at a client site.
</p>
<p>
Note that the <a href="../api_reference/C/envtxn_applied.html" class="olink">DB_ENV->txn_applied()</a> method can return
<code class="literal">DB_LOCK_DEADLOCK</code>. The application should
respond to this situation just as it does for any other normal
operation: abort any existing transaction, and then pause
briefly before retrying.
</p>
</div>
</div>
<div class="navfooter">
<hr />
<table width="100%" summary="Navigation footer">
<tr>
<td width="40%" align="left"><a accesskey="p" href="rep_lease.html">Prev</a> </td>
<td width="20%" align="center">
<a accesskey="u" href="rep.html">Up</a>
</td>
<td width="40%" align="right"> <a accesskey="n" href="rep_clock_skew.html">Next</a></td>
</tr>
<tr>
<td width="40%" align="left" valign="top">Master Leases </td>
<td width="20%" align="center">
<a accesskey="h" href="index.html">Home</a>
</td>
<td width="40%" align="right" valign="top"> Clock Skew</td>
</tr>
</table>
</div>
</body>
</html>
