doc/src/sgml/archive-modules.sgml

   1 <!-- doc/src/sgml/archive-modules.sgml -->
   2
   3 <chapter id="archive-modules">
   4  <title>Archive Modules</title>
   5  <indexterm zone="archive-modules">
   6   <primary>Archive Modules</primary>
   7  </indexterm>
   8
   9  <para>
  10   PostgreSQL provides infrastructure to create custom modules for continuous
  11   archiving (see <xref linkend="continuous-archiving"/>).  While archiving via
  12   a shell command (i.e., <xref linkend="guc-archive-command"/>) is much
  13   simpler, a custom archive module will often be considerably more robust and
  14   performant.
  15  </para>
  16
  17  <para>
  18   When a custom <xref linkend="guc-archive-library"/> is configured, PostgreSQL
  19   will submit completed WAL files to the module, and the server will avoid
  20   recycling or removing these WAL files until the module indicates that the files
  21   were successfully archived.  It is ultimately up to the module to decide what
  22   to do with each WAL file, but many recommendations are listed at
  23   <xref linkend="backup-archiving-wal"/>.
  24  </para>
  25
  26  <para>
  27   Archiving modules must at least consist of an initialization function (see
  28   <xref linkend="archive-module-init"/>) and the required callbacks (see
  29   <xref linkend="archive-module-callbacks"/>).  However, archive modules are
  30   also permitted to do much more (e.g., declare GUCs and register background
  31   workers).
  32  </para>
  33
  34  <para>
  35   The <filename>contrib/basic_archive</filename> module contains a working
  36   example, which demonstrates some useful techniques.
  37  </para>
  38
  39  <sect1 id="archive-module-init">
  40   <title>Initialization Functions</title>
  41   <indexterm zone="archive-module-init">
  42    <primary>_PG_archive_module_init</primary>
  43   </indexterm>
  44   <para>
  45    An archive library is loaded by dynamically loading a shared library with the
  46    <xref linkend="guc-archive-library"/>'s name as the library base name.  The
  47    normal library search path is used to locate the library.  To provide the
  48    required archive module callbacks and to indicate that the library is
  49    actually an archive module, it needs to provide a function named
  50    <function>_PG_archive_module_init</function>.  The result of the function
  51    must be a pointer to a struct of type
  52    <structname>ArchiveModuleCallbacks</structname>, which contains everything
  53    that the core code needs to know to make use of the archive module.  The
  54    return value needs to be of server lifetime, which is typically achieved by
  55    defining it as a <literal>static const</literal> variable in global scope.
  56
  57 <programlisting>
  58 typedef struct ArchiveModuleCallbacks
  59 {
  60     ArchiveStartupCB startup_cb;
  61     ArchiveCheckConfiguredCB check_configured_cb;
  62     ArchiveFileCB archive_file_cb;
  63     ArchiveShutdownCB shutdown_cb;
  64 } ArchiveModuleCallbacks;
  65 typedef const ArchiveModuleCallbacks *(*ArchiveModuleInit) (void);
  66 </programlisting>
  67
  68    Only the <function>archive_file_cb</function> callback is required.  The
  69    others are optional.
  70   </para>
  71  </sect1>
  72
  73  <sect1 id="archive-module-callbacks">
  74   <title>Archive Module Callbacks</title>
  75   <para>
  76    The archive callbacks define the actual archiving behavior of the module.
  77    The server will call them as required to process each individual WAL file.
  78   </para>
  79
  80   <sect2 id="archive-module-startup">
  81    <title>Startup Callback</title>
  82    <para>
  83     The <function>startup_cb</function> callback is called shortly after the
  84     module is loaded.  This callback can be used to perform any additional
  85     initialization required.  If the archive module has any state, it can use
  86     <structfield>state->private_data</structfield> to store it.
  87
  88 <programlisting>
  89 typedef void (*ArchiveStartupCB) (ArchiveModuleState *state);
  90 </programlisting>
  91    </para>
  92   </sect2>
  93
  94   <sect2 id="archive-module-check">
  95    <title>Check Callback</title>
  96    <para>
  97     The <function>check_configured_cb</function> callback is called to determine
  98     whether the module is fully configured and ready to accept WAL files (e.g.,
  99     its configuration parameters are set to valid values).  If no
 100     <function>check_configured_cb</function> is defined, the server always
 101     assumes the module is configured.
 102
 103 <programlisting>
 104 typedef bool (*ArchiveCheckConfiguredCB) (ArchiveModuleState *state);
 105 </programlisting>
 106
 107     If <literal>true</literal> is returned, the server will proceed with
 108     archiving the file by calling the <function>archive_file_cb</function>
 109     callback.  If <literal>false</literal> is returned, archiving will not
 110     proceed, and the archiver will emit the following message to the server log:
 111 <screen>
 112 WARNING:  archive_mode enabled, yet archiving is not configured
 113 </screen>
 114     In the latter case, the server will periodically call this function, and
 115     archiving will proceed only when it returns <literal>true</literal>.
 116    </para>
 117
 118    <note>
 119     <para>
 120      When returning <literal>false</literal>, it may be useful to append some
 121      additional information to the generic warning message.  To do that, provide
 122      a message to the <function>arch_module_check_errdetail</function> macro
 123      before returning <literal>false</literal>.  Like
 124      <function>errdetail()</function>, this macro accepts a format string
 125      followed by an optional list of arguments.  The resulting string will be
 126      emitted as the <literal>DETAIL</literal> line of the warning message.
 127     </para>
 128    </note>
 129   </sect2>
 130
 131   <sect2 id="archive-module-archive">
 132    <title>Archive Callback</title>
 133    <para>
 134     The <function>archive_file_cb</function> callback is called to archive a
 135     single WAL file.
 136
 137 <programlisting>
 138 typedef bool (*ArchiveFileCB) (ArchiveModuleState *state, const char *file, const char *path);
 139 </programlisting>
 140
 141     If <literal>true</literal> is returned, the server proceeds as if the file
 142     was successfully archived, which may include recycling or removing the
 143     original WAL file.  If <literal>false</literal> is returned or an error is thrown, the server will
 144     keep the original WAL file and retry archiving later.
 145     <replaceable>file</replaceable> will contain just the file name of the WAL
 146     file to archive, while <replaceable>path</replaceable> contains the full
 147     path of the WAL file (including the file name).
 148    </para>
 149
 150    <note>
 151     <para>
 152      The <function>archive_file_cb</function> callback is called in a
 153      short-lived memory context that will be reset between invocations.  If you
 154      need longer-lived storage, create a memory context in the module's
 155      <function>startup_cb</function> callback.
 156     </para>
 157    </note>
 158   </sect2>
 159
 160   <sect2 id="archive-module-shutdown">
 161    <title>Shutdown Callback</title>
 162    <para>
 163     The <function>shutdown_cb</function> callback is called when the archiver
 164     process exits (e.g., after an error) or the value of
 165     <xref linkend="guc-archive-library"/> changes.  If no
 166     <function>shutdown_cb</function> is defined, no special action is taken in
 167     these situations.  If the archive module has any state, this callback should
 168     free it to avoid leaks.
 169
 170 <programlisting>
 171 typedef void (*ArchiveShutdownCB) (ArchiveModuleState *state);
 172 </programlisting>
 173    </para>
 174   </sect2>
 175  </sect1>
 176 </chapter>