postgresql/doc/src/sgml/ref/postgres-ref.sgml

<!--
doc/src/sgml/ref/postgres-ref.sgml
PostgreSQL documentation
-->

<refentry id="app-postgres">
 <indexterm zone="app-postgres">
  <primary>postgres</primary>
 </indexterm>

 <refmeta>
  <refentrytitle><application>postgres</application></refentrytitle>
  <manvolnum>1</manvolnum>
  <refmiscinfo>Application</refmiscinfo>
 </refmeta>

 <refnamediv>
  <refname>postgres</refname>
  <refpurpose><productname>PostgreSQL</productname> database server</refpurpose>
 </refnamediv>

 <refsynopsisdiv>
  <cmdsynopsis>
   <command>postgres</command>
   <arg rep="repeat"><replaceable>option</replaceable></arg>
  </cmdsynopsis>
 </refsynopsisdiv>

 <refsect1>
  <title>Description</title>

  <para>
   <command>postgres</command> is the
   <productname>PostgreSQL</productname> database server.  In order
   for a client application to access a database it connects (over a
   network or locally) to a running <command>postgres</command> instance.
   The <command>postgres</command> instance then starts a separate server
   process to handle the connection.
  </para>

  <para>
   One <command>postgres</command> instance always manages the data of
   exactly one database cluster.  A database cluster is a collection
   of databases that is stored at a common file system location (the
   <quote>data area</quote>).  More than one
   <command>postgres</command> instance can run on a system at one
   time, so long as they use different data areas and different
   communication ports (see below).  When
   <command>postgres</command> starts it needs to know the location
   of the data area.  The location must be specified by the
   <option>-D</option> option or the <envar>PGDATA</envar> environment
   variable; there is no default.  Typically, <option>-D</option> or
   <envar>PGDATA</envar> points directly to the data area directory
   created by <xref linkend="app-initdb"/>.  Other possible file layouts are
   discussed in <xref linkend="runtime-config-file-locations"/>.
  </para>

  <para>
   By default <command>postgres</command> starts in the
   foreground and prints log messages to the standard error stream.  In
   practical applications <command>postgres</command>
   should be started as a background process, perhaps at boot time.
  </para>

  <para>
   The <command>postgres</command> command can also be called in
   single-user mode.  The primary use for this mode is during
   bootstrapping by <xref linkend="app-initdb"/>.  Sometimes it is used
   for debugging or disaster recovery;  note that running a single-user
   server is not truly suitable for debugging the server, since no
   realistic interprocess communication and locking will happen.
   When invoked in single-user
   mode from the shell, the user can enter queries and the results
   will be printed to the screen, but in a form that is more useful
   for developers than end users.  In the single-user mode,
   the session user will be set to the user with ID 1, and implicit
   superuser powers are granted to this user.
   This user does not actually have to exist, so the single-user mode
   can be used to manually recover from certain
   kinds of accidental damage to the system catalogs.
  </para>
 </refsect1>

 <refsect1 id="app-postgres-options">
  <title>Options</title>

   <para>
    <command>postgres</command> accepts the following command-line
    arguments.  For a detailed discussion of the options consult <xref
    linkend="runtime-config"/>.  You can save typing most of these
    options by setting up a configuration file.  Some (safe) options
    can also be set from the connecting client in an
    application-dependent way to apply only for that session.  For
    example, if the environment variable <envar>PGOPTIONS</envar> is
    set, then <application>libpq</application>-based clients will pass that
    string to the server, which will interpret it as
    <command>postgres</command> command-line options.
   </para>

   <refsect2>
    <title>General Purpose</title>

    <variablelist>
     <varlistentry>
      <term><option>-B <replaceable class="parameter">nbuffers</replaceable></option></term>
      <listitem>
       <para>
        Sets the number of shared buffers for use by the server
        processes.  The default value of this parameter is chosen
        automatically by <application>initdb</application>.
        Specifying this option is equivalent to setting the
        <xref linkend="guc-shared-buffers"/> configuration parameter.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-c <replaceable>name</replaceable>=<replaceable>value</replaceable></option></term>
      <listitem>
       <para>
        Sets a named run-time parameter. The configuration parameters
        supported by <productname>PostgreSQL</productname> are
        described in <xref linkend="runtime-config"/>. Most of the
        other command line options are in fact short forms of such a
        parameter assignment.  <option>-c</option> can appear multiple times
        to set multiple parameters.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-C <replaceable>name</replaceable></option></term>
      <listitem>
       <para>
        Prints the value of the named run-time parameter, and exits.
        (See the <option>-c</option> option above for details.)  This
        returns values from
        <filename>postgresql.conf</filename>, modified by any parameters
        supplied in this invocation.  It does not reflect parameters
        supplied when the cluster was started.
       </para>

       <para>
        This can be used on a running server for most parameters.  However,
        the server must be shut down for some runtime-computed parameters
        (e.g., <xref linkend="guc-shared-memory-size"/>,
        <xref linkend="guc-shared-memory-size-in-huge-pages"/>, and
        <xref linkend="guc-wal-segment-size"/>).
       </para>

       <para>
        This option is meant for other programs that interact with a server
        instance, such as <xref linkend="app-pg-ctl"/>, to query configuration
        parameter values.  User-facing applications should instead use <link
        linkend="sql-show"><command>SHOW</command></link> or the <structname>pg_settings</structname> view.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-d <replaceable>debug-level</replaceable></option></term>
      <listitem>
       <para>
        Sets the debug level.  The higher this value is set, the more
        debugging output is written to the server log.  Values are
        from 1 to 5.  It is also possible to pass <literal>-d
        0</literal> for a specific session, which will prevent the
        server log level of the parent <command>postgres</command> process from being
        propagated to this session.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-D <replaceable class="parameter">datadir</replaceable></option></term>
      <listitem>
       <para>
        Specifies the file system location of the database
        configuration files.  See
        <xref linkend="runtime-config-file-locations"/> for details.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-e</option></term>
      <listitem>
       <para>
        Sets the default date style to <quote>European</quote>, that is
        <literal>DMY</literal> ordering of input date fields.  This also causes
        the day to be printed before the month in certain date output formats.
        See <xref linkend="datatype-datetime"/> for more information.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-F</option></term>
      <listitem>
       <para>
        Disables <function>fsync</function> calls for improved
        performance, at the risk of data corruption in the event of a
        system crash.  Specifying this option is equivalent to
        disabling the <xref linkend="guc-fsync"/> configuration
        parameter. Read the detailed documentation before using this!
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-h <replaceable class="parameter">hostname</replaceable></option></term>
      <listitem>
       <para>
        Specifies the IP host name or address on which
        <command>postgres</command> is to listen for TCP/IP
        connections from client applications.  The value can also be a
        comma-separated list of addresses, or <literal>*</literal> to specify
        listening on all available interfaces.  An empty value
        specifies not listening on any IP addresses, in which case
        only Unix-domain sockets can be used to connect to the
        server.  Defaults to listening only on
        <systemitem class="systemname">localhost</systemitem>.
        Specifying this option is equivalent to setting the <xref
        linkend="guc-listen-addresses"/> configuration parameter.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-i</option></term>
      <listitem>
       <para>
        Allows remote clients to connect via TCP/IP (Internet domain)
        connections.  Without this option, only local connections are
        accepted.  This option is equivalent to setting
        <varname>listen_addresses</varname> to <literal>*</literal> in
        <filename>postgresql.conf</filename> or via <option>-h</option>.
       </para>
       <para>
        This option is deprecated since it does not allow access to the
        full functionality of <xref linkend="guc-listen-addresses"/>.
        It's usually better to set <varname>listen_addresses</varname> directly.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-k <replaceable class="parameter">directory</replaceable></option></term>
      <listitem>
       <para>
        Specifies the directory of the Unix-domain socket on which
        <command>postgres</command> is to listen for
        connections from client applications.  The value can also be a
        comma-separated list of directories.  An empty value
        specifies not listening on any Unix-domain sockets, in which case
        only TCP/IP sockets can be used to connect to the server.
        The default value is normally
        <filename>/tmp</filename>, but that can be changed at build time.
        Specifying this option is equivalent to setting the <xref
        linkend="guc-unix-socket-directories"/> configuration parameter.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-l</option></term>
      <listitem>
       <para>
        Enables secure connections using <acronym>SSL</acronym>.
        <productname>PostgreSQL</productname> must have been compiled with
        support for <acronym>SSL</acronym> for this option to be
        available. For more information on using <acronym>SSL</acronym>,
        refer to <xref linkend="ssl-tcp"/>.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-N <replaceable class="parameter">max-connections</replaceable></option></term>
      <listitem>
       <para>
        Sets the maximum number of client connections that this
        server will accept.  The default value of this parameter is chosen
        automatically by <application>initdb</application>.
        Specifying this option is equivalent to setting the
        <xref linkend="guc-max-connections"/> configuration parameter.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-p <replaceable class="parameter">port</replaceable></option></term>
      <listitem>
       <para>
        Specifies the TCP/IP port or local Unix domain socket file
        extension on which <command>postgres</command>
        is to listen for connections from client applications.
        Defaults to the value of the <envar>PGPORT</envar> environment
        variable, or if <envar>PGPORT</envar> is not set, then
        defaults to the value established during compilation (normally
        5432).  If you specify a port other than the default port,
        then all client applications must specify the same port using
        either command-line options or <envar>PGPORT</envar>.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-s</option></term>
      <listitem>
       <para>
        Print time information and other statistics at the end of each command.
        This is useful for benchmarking or for use in tuning the number of
        buffers.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-S</option> <replaceable class="parameter">work-mem</replaceable></term>
      <listitem>
       <para>
        Specifies the base amount of memory to be used by sorts and
        hash tables before resorting to temporary disk files.  See the
        description of the <varname>work_mem</varname> configuration
        parameter in <xref linkend="runtime-config-resource-memory"/>.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-V</option></term>
      <term><option>--version</option></term>
      <listitem>
       <para>
        Print the <application>postgres</application> version and exit.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--<replaceable>name</replaceable>=<replaceable>value</replaceable></option></term>
      <listitem>
       <para>
        Sets a named run-time parameter; a shorter form of
        <option>-c</option>.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--describe-config</option></term>
      <listitem>
       <para>
        This option dumps out the server's internal configuration variables,
        descriptions, and defaults in tab-delimited <command>COPY</command> format.
        It is designed primarily for use by administration tools.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-?</option></term>
      <term><option>--help</option></term>
      <listitem>
       <para>
        Show help about <application>postgres</application> command line
        arguments, and exit.
       </para>
      </listitem>
     </varlistentry>
    </variablelist>
   </refsect2>

   <refsect2>
    <title>Semi-Internal Options</title>

    <para>
     The options described here are used
     mainly for debugging purposes, and in some cases to assist with
     recovery of severely damaged databases. There should be no reason
     to use them in a production database setup.  They are listed
     here only for use by <productname>PostgreSQL</productname>
     system developers.  Furthermore, these options might
     change or be removed in a future release without notice.
    </para>

    <variablelist>
     <varlistentry>
      <term><option>-f</option> <literal>{ s | i | o | b | t | n | m | h }</literal></term>
      <listitem>
       <para>
        Forbids the use of particular scan and join methods:
        <literal>s</literal> and <literal>i</literal>
        disable sequential and index scans respectively,
        <literal>o</literal>, <literal>b</literal> and <literal>t</literal>
        disable index-only scans, bitmap index scans, and TID scans
        respectively, while
        <literal>n</literal>, <literal>m</literal>, and <literal>h</literal>
        disable nested-loop, merge and hash joins respectively.
       </para>

       <para>
        Neither sequential scans nor nested-loop joins can be disabled
        completely; the <literal>-fs</literal> and
        <literal>-fn</literal> options simply discourage the optimizer
        from using those plan types if it has any other alternative.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-O</option></term>
      <listitem>
       <para>
        Allows the structure of system tables to be modified.  This is
        used by <command>initdb</command>.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-P</option></term>
      <listitem>
       <para>
        Ignore system indexes when reading system tables, but still update
        the indexes when modifying the tables.  This is useful when
        recovering from damaged system indexes.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-t</option> <literal>pa[rser] | pl[anner] | e[xecutor]</literal></term>
      <listitem>
       <para>
        Print timing statistics for each query relating to each of the
        major system modules.  This option cannot be used together
        with the <option>-s</option> option.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-T</option></term>
      <listitem>
       <para>
        This option is for debugging problems that cause a server
        process to die abnormally.  The ordinary strategy in this
        situation is to notify all other server processes that they
        must terminate, by sending them <systemitem>SIGQUIT</systemitem>
        signals.  With this option, <systemitem>SIGABRT</systemitem>
        will be sent instead, resulting in production of core dump files.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-v</option> <replaceable class="parameter">protocol</replaceable></term>
      <listitem>
       <para>
        Specifies the version number of the frontend/backend protocol
        to be used for a particular session.  This option is for
        internal use only.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-W</option> <replaceable class="parameter">seconds</replaceable></term>
      <listitem>
       <para>
        A delay of this many seconds occurs when a new server process
        is started, after it conducts the authentication procedure.
        This is intended to give an opportunity to attach to the
        server process with a debugger.
       </para>
      </listitem>
     </varlistentry>
    </variablelist>
   </refsect2>

   <refsect2>
    <title>Options for Single-User Mode</title>

    <indexterm>
     <primary>single-user mode</primary>
    </indexterm>

    <para>
     The following options only apply to the single-user mode
     (see <xref linkend="app-postgres-single-user"/> below).
    </para>

    <variablelist>
     <varlistentry>
      <term><option>--single</option></term>
      <listitem>
       <para>
        Selects the single-user mode.  This must be the first argument
        on the command line.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><replaceable class="parameter">database</replaceable></term>
      <listitem>
       <para>
        Specifies the name of the database to be accessed.  This must be
        the last argument on the command line.  If it is
        omitted it defaults to the user name.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-E</option></term>
      <listitem>
       <para>
        Echo all commands to standard output before executing them.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-j</option></term>
      <listitem>
       <para>
        Use semicolon followed by two newlines, rather than just newline,
        as the command entry terminator.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-r</option> <replaceable class="parameter">filename</replaceable></term>
      <listitem>
       <para>
        Send all server log output to <replaceable
        class="parameter">filename</replaceable>.  This option is only
        honored when supplied as a command-line option.
       </para>
      </listitem>
     </varlistentry>
    </variablelist>
   </refsect2>
 </refsect1>

 <refsect1>
  <title>Environment</title>

  <variablelist>
   <varlistentry>
    <term><envar>PGCLIENTENCODING</envar></term>

    <listitem>
     <para>
      Default character encoding used by clients.  (The clients can
      override this individually.)  This value can also be set in the
      configuration file.
     </para>
    </listitem>
   </varlistentry>

   <varlistentry>
    <term><envar>PGDATA</envar></term>

    <listitem>
     <para>
      Default data directory location
     </para>
    </listitem>
   </varlistentry>

   <varlistentry>
    <term><envar>PGDATESTYLE</envar></term>

    <listitem>
     <para>
      Default value of the <xref linkend="guc-datestyle"/> run-time
      parameter.  (The use of this environment variable is deprecated.)
     </para>
    </listitem>
   </varlistentry>

   <varlistentry>
    <term><envar>PGPORT</envar></term>

    <listitem>
     <para>
      Default port number (preferably set in the configuration file)
     </para>
    </listitem>
   </varlistentry>

  </variablelist>
 </refsect1>

 <refsect1>
   <title>Diagnostics</title>

   <para>
    A failure message mentioning <literal>semget</literal> or
    <literal>shmget</literal> probably indicates you need to configure your
    kernel to provide adequate shared memory and semaphores.  For more
    discussion see <xref linkend="kernel-resources"/>.  You might be able
    to postpone reconfiguring your kernel by decreasing <xref
    linkend="guc-shared-buffers"/> to reduce the shared memory
    consumption of <productname>PostgreSQL</productname>, and/or by reducing
    <xref linkend="guc-max-connections"/> to reduce the semaphore
    consumption.
   </para>

   <para>
    A failure message suggesting that another server is already running
    should be checked carefully, for example by using the command
<screen>
<prompt>$</prompt> <userinput>ps ax | grep postgres</userinput>
</screen>
        or
<screen>
<prompt>$</prompt> <userinput>ps -ef | grep postgres</userinput>
</screen>
    depending on your system.  If you are certain that no conflicting
    server is running, you can remove the lock file mentioned in the
    message and try again.
   </para>

   <para>
    A failure message indicating inability to bind to a port might
    indicate that that port is already in use by some
    non-<productname>PostgreSQL</productname> process.  You might also
    get this error if you terminate <command>postgres</command>
    and immediately restart it using the same port; in this case, you
    must simply wait a few seconds until the operating system closes
    the port before trying again.  Finally, you might get this error if
    you specify a port number that your operating system considers to
    be reserved.  For example, many versions of Unix consider port
    numbers under 1024 to be <quote>trusted</quote> and only permit
    the Unix superuser to access them.
   </para>

 </refsect1>

 <refsect1>
  <title>Notes</title>

  <para>
   The utility command <xref linkend="app-pg-ctl"/> can be used to
   start and shut down the <command>postgres</command> server
   safely and comfortably.
  </para>

  <para>
   If at all possible, <emphasis>do not</emphasis> use
   <literal>SIGKILL</literal> to kill the main
   <command>postgres</command> server.  Doing so will prevent
   <command>postgres</command> from freeing the system
   resources (e.g., shared memory and semaphores) that it holds before
   terminating.  This might cause problems for starting a fresh
   <command>postgres</command> run.
  </para>

  <para>
   To terminate the <command>postgres</command> server normally, the
   signals <literal>SIGTERM</literal>, <literal>SIGINT</literal>, or
   <literal>SIGQUIT</literal> can be used.  The first will wait for
   all clients to terminate before quitting, the second will
   forcefully disconnect all clients, and the third will quit
   immediately without proper shutdown, resulting in a recovery run
   during restart.
  </para>

  <para>
   The <literal>SIGHUP</literal> signal will reload
   the server configuration files.  It is also possible to send
   <literal>SIGHUP</literal> to an individual server process, but that
   is usually not sensible.
  </para>

  <para>
   To cancel a running query, send the <literal>SIGINT</literal> signal
   to the process running that command. To terminate a backend process
   cleanly, send <literal>SIGTERM</literal> to that process. See
   also <function>pg_cancel_backend</function> and <function>pg_terminate_backend</function>
   in <xref linkend="functions-admin-signal"/> for the SQL-callable equivalents
   of these two actions.
  </para>

  <para>
   The <command>postgres</command> server uses <literal>SIGQUIT</literal>
   to tell subordinate server processes to terminate without normal
   cleanup.
   This signal <emphasis>should not</emphasis> be used by users.  It
   is also unwise to send <literal>SIGKILL</literal> to a server
   process &mdash; the main <command>postgres</command> process will
   interpret this as a crash and will force all the sibling processes
   to quit as part of its standard crash-recovery procedure.
  </para>
 </refsect1>

 <refsect1 id="app-postgres-bugs">
  <title>Bugs</title>
  <para>
   The <option>--</option> options will not work on <systemitem
   class="osname">FreeBSD</systemitem> or <systemitem class="osname">OpenBSD</systemitem>.
   Use <option>-c</option> instead. This is a bug in the affected operating
   systems; a future release of <productname>PostgreSQL</productname>
   will provide a workaround if this is not fixed.
  </para>
 </refsect1>

 <refsect1 id="app-postgres-single-user" xreflabel="Single-User Mode">
  <title>Single-User Mode</title>

   <para>
    To start a single-user mode server, use a command like
<screen>
<userinput>postgres --single -D /usr/local/pgsql/data <replaceable>other-options</replaceable> my_database</userinput>
</screen>
    Provide the correct path to the database directory with <option>-D</option>, or
    make sure that the environment variable <envar>PGDATA</envar> is set.
    Also specify the name of the particular database you want to work in.
   </para>

   <para>
    Normally, the single-user mode server treats newline as the command
    entry terminator; there is no intelligence about semicolons,
    as there is in <application>psql</application>.  To continue a command
    across multiple lines, you must type backslash just before each
    newline except the last one.  The backslash and adjacent newline are
    both dropped from the input command.  Note that this will happen even
    when within a string literal or comment.
   </para>

   <para>
    But if you use the <option>-j</option> command line switch, a single newline
    does not terminate command entry; instead, the sequence
    semicolon-newline-newline does.  That is, type a semicolon immediately
    followed by a completely empty line.  Backslash-newline is not
    treated specially in this mode.  Again, there is no intelligence about
    such a sequence appearing within a string literal or comment.
   </para>

   <para>
    In either input mode, if you type a semicolon that is not just before or
    part of a command entry terminator, it is considered a command separator.
    When you do type a command entry terminator, the multiple statements
    you've entered will be executed as a single transaction.
   </para>

   <para>
    To quit the session, type <acronym>EOF</acronym>
    (<keycombo action="simul"><keycap>Control</keycap><keycap>D</keycap></keycombo>, usually).
    If you've entered any text since the last command entry terminator,
    then <acronym>EOF</acronym> will be taken as a command entry terminator,
    and another <acronym>EOF</acronym> will be needed to exit.
   </para>

   <para>
    Note that the single-user mode server does not provide sophisticated
    line-editing features (no command history, for example).
    Single-user mode also does not do any background processing, such as
    automatic checkpoints or replication.
   </para>
 </refsect1>

 <refsect1 id="app-postgres-examples">
  <title>Examples</title>

  <para>
   To start <command>postgres</command> in the background
   using default values, type:

<screen>
<prompt>$</prompt> <userinput>nohup postgres &gt;logfile 2&gt;&amp;1 &lt;/dev/null &amp;</userinput>
</screen>
  </para>

  <para>
   To start <command>postgres</command> with a specific
   port, e.g., 1234:
<screen>
<prompt>$</prompt> <userinput>postgres -p 1234</userinput>
</screen>
   To connect to this server using <application>psql</application>, specify this port with the -p option:
<screen>
<prompt>$</prompt> <userinput>psql -p 1234</userinput>
</screen>
   or set the environment variable <envar>PGPORT</envar>:
<screen>
<prompt>$</prompt> <userinput>export PGPORT=1234</userinput>
<prompt>$</prompt> <userinput>psql</userinput>
</screen>
  </para>

  <para>
   Named run-time parameters can be set in either of these styles:
<screen>
<prompt>$</prompt> <userinput>postgres -c work_mem=1234</userinput>
<prompt>$</prompt> <userinput>postgres --work-mem=1234</userinput>
</screen>
   Either form overrides whatever setting might exist for
   <varname>work_mem</varname> in <filename>postgresql.conf</filename>.  Notice that
   underscores in parameter names can be written as either underscore
   or dash on the command line.  Except for short-term experiments,
   it's probably better practice to edit the setting in
   <filename>postgresql.conf</filename> than to rely on a command-line switch
   to set a parameter.
  </para>
 </refsect1>

 <refsect1>
  <title>See Also</title>

  <para>
   <xref linkend="app-initdb"/>,
   <xref linkend="app-pg-ctl"/>
  </para>
 </refsect1>
</refentry>