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

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

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

 <refnamediv>
  <refname>vacuumdb</refname>
  <refpurpose>garbage-collect and analyze a <productname>PostgreSQL</productname> database</refpurpose>
 </refnamediv>

 <refsynopsisdiv>
  <cmdsynopsis>
   <command>vacuumdb</command>
   <arg rep="repeat"><replaceable>connection-option</replaceable></arg>
   <arg rep="repeat"><replaceable>option</replaceable></arg>

   <arg choice="plain" rep="repeat">
    <arg choice="opt">
     <group choice="plain">
      <arg choice="plain"><option>-t</option></arg>
      <arg choice="plain"><option>--table</option></arg>
     </group>
     <replaceable>table</replaceable>
     <arg choice="opt">( <replaceable class="parameter">column</replaceable> [,...] )</arg>
    </arg>
   </arg>

   <arg choice="opt">
    <group choice="plain">
     <arg choice="plain"><replaceable>dbname</replaceable></arg>
     <arg choice="plain"><option>-a</option></arg>
     <arg choice="plain"><option>--all</option></arg>
    </group>
   </arg>
  </cmdsynopsis>

  <cmdsynopsis>
   <command>vacuumdb</command>
   <arg rep="repeat"><replaceable>connection-option</replaceable></arg>
   <arg rep="repeat"><replaceable>option</replaceable></arg>

   <arg choice="plain" rep="repeat">
    <arg choice="opt">
     <group choice="plain">
      <arg choice="plain"><option>-n</option></arg>
      <arg choice="plain"><option>--schema</option></arg>
     </group>
     <replaceable>schema</replaceable>
    </arg>
   </arg>

   <arg choice="opt">
    <group choice="plain">
     <arg choice="plain"><replaceable>dbname</replaceable></arg>
     <arg choice="plain"><option>-a</option></arg>
     <arg choice="plain"><option>--all</option></arg>
    </group>
   </arg>
  </cmdsynopsis>

  <cmdsynopsis>
   <command>vacuumdb</command>
   <arg rep="repeat"><replaceable>connection-option</replaceable></arg>
   <arg rep="repeat"><replaceable>option</replaceable></arg>

   <arg choice="plain" rep="repeat">
    <arg choice="opt">
     <group choice="plain">
      <arg choice="plain"><option>-N</option></arg>
      <arg choice="plain"><option>--exclude-schema</option></arg>
     </group>
     <replaceable>schema</replaceable>
    </arg>
   </arg>

   <arg choice="opt">
    <group choice="plain">
     <arg choice="plain"><replaceable>dbname</replaceable></arg>
     <arg choice="plain"><option>-a</option></arg>
     <arg choice="plain"><option>--all</option></arg>
    </group>
   </arg>
  </cmdsynopsis>
 </refsynopsisdiv>

 <refsect1>
  <title>Description</title>

  <para>
   <application>vacuumdb</application> is a utility for cleaning a
   <productname>PostgreSQL</productname> database.
   <application>vacuumdb</application> will also generate internal statistics
   used by the <productname>PostgreSQL</productname> query optimizer.
  </para>

  <para>
   <application>vacuumdb</application> is a wrapper around the SQL
   command <link linkend="sql-vacuum"><command>VACUUM</command></link>.
   There is no effective difference between vacuuming and analyzing
   databases via this utility and via other methods for accessing the
   server.
  </para>

 </refsect1>


 <refsect1>
  <title>Options</title>

   <para>
    <application>vacuumdb</application> accepts the following command-line arguments:
    <variablelist>
     <varlistentry>
      <term><option>-a</option></term>
      <term><option>--all</option></term>
      <listitem>
       <para>
        Vacuum all databases.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--buffer-usage-limit <replaceable class="parameter">size</replaceable></option></term>
      <listitem>
       <para>
        Specifies the
        <glossterm linkend="glossary-buffer-access-strategy">Buffer Access Strategy</glossterm>
        ring buffer size for a given invocation of <application>vacuumdb</application>.
        This size is used to calculate the number of shared buffers which will
        be reused as part of this strategy.  See <xref linkend="sql-vacuum"/>.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option><optional>-d</optional> <replaceable class="parameter">dbname</replaceable></option></term>
      <term><option><optional>--dbname=</optional><replaceable class="parameter">dbname</replaceable></option></term>
      <listitem>
       <para>
        Specifies the name of the database to be cleaned or analyzed,
        when <option>-a</option>/<option>--all</option> is not used.
        If this is not specified, the database name is read
        from the environment variable <envar>PGDATABASE</envar>.  If
        that is not set, the user name specified for the connection is
        used.  The <replaceable>dbname</replaceable> can be a <link
        linkend="libpq-connstring">connection string</link>.  If so,
        connection string parameters will override any conflicting command
        line options.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--disable-page-skipping</option></term>
      <listitem>
       <para>
        Disable skipping pages based on the contents of the visibility map.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-e</option></term>
      <term><option>--echo</option></term>
      <listitem>
       <para>
        Echo the commands that <application>vacuumdb</application> generates
        and sends to the server.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-f</option></term>
      <term><option>--full</option></term>
      <listitem>
       <para>
        Perform <quote>full</quote> vacuuming.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-F</option></term>
      <term><option>--freeze</option></term>
      <listitem>
       <para>
        Aggressively <quote>freeze</quote> tuples.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--force-index-cleanup</option></term>
      <listitem>
       <para>
        Always remove index entries pointing to dead tuples.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-j <replaceable class="parameter">njobs</replaceable></option></term>
      <term><option>--jobs=<replaceable class="parameter">njobs</replaceable></option></term>
      <listitem>
       <para>
        Execute the vacuum or analyze commands in parallel by running
        <replaceable class="parameter">njobs</replaceable>
        commands simultaneously.  This option may reduce the processing time
        but it also increases the load on the database server.
       </para>
       <para>
        <application>vacuumdb</application> will open
        <replaceable class="parameter">njobs</replaceable> connections to the
        database, so make sure your <xref linkend="guc-max-connections"/>
        setting is high enough to accommodate all connections.
       </para>
       <para>
        Note that using this mode together with the <option>-f</option>
        (<literal>FULL</literal>) option might cause deadlock failures if
        certain system catalogs are processed in parallel.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--min-mxid-age <replaceable class="parameter">mxid_age</replaceable></option></term>
      <listitem>
       <para>
        Only execute the vacuum or analyze commands on tables with a multixact
        ID age of at least <replaceable class="parameter">mxid_age</replaceable>.
        This setting is useful for prioritizing tables to process to prevent
        multixact ID wraparound (see
        <xref linkend="vacuum-for-multixact-wraparound"/>).
       </para>
       <para>
        For the purposes of this option, the multixact ID age of a relation is
        the greatest of the ages of the main relation and its associated
        <acronym>TOAST</acronym> table, if one exists.  Since the commands
        issued by <application>vacuumdb</application> will also process the
        <acronym>TOAST</acronym> table for the relation if necessary, it does
        not need to be considered separately.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--min-xid-age <replaceable class="parameter">xid_age</replaceable></option></term>
      <listitem>
       <para>
        Only execute the vacuum or analyze commands on tables with a
        transaction ID age of at least
        <replaceable class="parameter">xid_age</replaceable>.  This setting
        is useful for prioritizing tables to process to prevent transaction
        ID wraparound (see <xref linkend="vacuum-for-wraparound"/>).
       </para>
       <para>
        For the purposes of this option, the transaction ID age of a relation
        is the greatest of the ages of the main relation and its associated
        <acronym>TOAST</acronym> table, if one exists.  Since the commands
        issued by <application>vacuumdb</application> will also process the
        <acronym>TOAST</acronym> table for the relation if necessary, it does
        not need to be considered separately.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--missing-stats-only</option></term>
      <listitem>
       <para>
        Only analyze relations that are missing statistics for a column, index
        expression, or extended statistics object.  When used with
        <option>--analyze-in-stages</option>, this option prevents
        <application>vacuumdb</application> from temporarily replacing existing
        statistics with ones generated with lower statistics targets, thus
        avoiding transiently worse query optimizer choices.
       </para>
       <para>
        This option can only be used in conjunction with
        <option>--analyze-only</option> or <option>--analyze-in-stages</option>.
       </para>
       <para>
        Note that <option>--missing-stats-only</option> requires
        <literal>SELECT</literal> privileges on
        <link linkend="catalog-pg-statistic"><structname>pg_statistic</structname></link>
        and
        <link linkend="catalog-pg-statistic-ext-data"><structname>pg_statistic_ext_data</structname></link>,
        which are restricted to superusers by default.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-n <replaceable class="parameter">schema</replaceable></option></term>
      <term><option>--schema=<replaceable class="parameter">schema</replaceable></option></term>
      <listitem>
       <para>
        Clean or analyze all tables in
        <replaceable class="parameter">schema</replaceable> only.  Multiple
        schemas can be vacuumed by writing multiple <option>-n</option> switches.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-N <replaceable class="parameter">schema</replaceable></option></term>
      <term><option>--exclude-schema=<replaceable class="parameter">schema</replaceable></option></term>
      <listitem>
       <para>
        Do not clean or analyze any tables in
        <replaceable class="parameter">schema</replaceable>.  Multiple schemas
        can be excluded by writing multiple <option>-N</option> switches.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--no-index-cleanup</option></term>
      <listitem>
       <para>
        Do not remove index entries pointing to dead tuples.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--no-process-main</option></term>
      <listitem>
       <para>
        Skip the main relation.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--no-process-toast</option></term>
      <listitem>
       <para>
        Skip the TOAST table associated with the table to vacuum, if any.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--no-truncate</option></term>
      <listitem>
       <para>
        Do not truncate empty pages at the end of the table.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-P <replaceable class="parameter">parallel_workers</replaceable></option></term>
      <term><option>--parallel=<replaceable class="parameter">parallel_workers</replaceable></option></term>
      <listitem>
       <para>
        Specify the number of parallel workers for <firstterm>parallel vacuum</firstterm>.
        This allows the vacuum to leverage multiple CPUs to process indexes.
        See <xref linkend="sql-vacuum"/>.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-q</option></term>
      <term><option>--quiet</option></term>
      <listitem>
       <para>
        Do not display progress messages.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--skip-locked</option></term>
      <listitem>
       <para>
        Skip relations that cannot be immediately locked for processing.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-t <replaceable class="parameter">table</replaceable> [ (<replaceable class="parameter">column</replaceable> [,...]) ]</option></term>
      <term><option>--table=<replaceable class="parameter">table</replaceable> [ (<replaceable class="parameter">column</replaceable> [,...]) ]</option></term>
      <listitem>
       <para>
        Clean or analyze <replaceable class="parameter">table</replaceable> only.
        Column names can be specified only in conjunction with
        the <option>--analyze</option> or <option>--analyze-only</option> options.
        Multiple tables can be vacuumed by writing multiple
        <option>-t</option> switches.
       </para>
       <tip>
        <para>
         If you specify columns, you probably have to escape the parentheses
         from the shell.  (See examples below.)
        </para>
       </tip>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-v</option></term>
      <term><option>--verbose</option></term>
      <listitem>
       <para>
        Print detailed information during processing.
       </para>
      </listitem>
     </varlistentry>

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

     <varlistentry>
      <term><option>-z</option></term>
      <term><option>--analyze</option></term>
      <listitem>
       <para>
        Also calculate statistics for use by the optimizer.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-Z</option></term>
      <term><option>--analyze-only</option></term>
      <listitem>
       <para>
        Only calculate statistics for use by the optimizer (no vacuum).
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--analyze-in-stages</option></term>
      <listitem>
       <para>
        Only calculate statistics for use by the optimizer (no vacuum),
        like <option>--analyze-only</option>.  Run three
        stages of analyze; the first stage uses the lowest possible statistics
        target (see <xref linkend="guc-default-statistics-target"/>)
        to produce usable statistics faster, and subsequent stages build the
        full statistics.
       </para>

       <para>
        This option is only useful to analyze a database that currently has
        no statistics or has wholly incorrect ones, such as if it is newly
        populated from a restored dump or by <command>pg_upgrade</command>.
        Be aware that running with this option in a database with existing
        statistics may cause the query optimizer choices to become
        transiently worse due to the low statistics targets of the early
        stages.
       </para>
      </listitem>
     </varlistentry>

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

    </variablelist>
   </para>

   <para>
    <application>vacuumdb</application> also accepts
    the following command-line arguments for connection parameters:
    <variablelist>
     <varlistentry>
      <term><option>-h <replaceable class="parameter">host</replaceable></option></term>
      <term><option>--host=<replaceable class="parameter">host</replaceable></option></term>
      <listitem>
       <para>
        Specifies the host name of the machine on which the server
        is running.  If the value begins with a slash, it is used
        as the directory for the Unix domain socket.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-p <replaceable class="parameter">port</replaceable></option></term>
      <term><option>--port=<replaceable class="parameter">port</replaceable></option></term>
      <listitem>
       <para>
        Specifies the TCP port or local Unix domain socket file
        extension on which the server
        is listening for connections.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-U <replaceable class="parameter">username</replaceable></option></term>
      <term><option>--username=<replaceable class="parameter">username</replaceable></option></term>
      <listitem>
       <para>
        User name to connect as.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-w</option></term>
      <term><option>--no-password</option></term>
      <listitem>
       <para>
        Never issue a password prompt.  If the server requires
        password authentication and a password is not available by
        other means such as a <filename>.pgpass</filename> file, the
        connection attempt will fail.  This option can be useful in
        batch jobs and scripts where no user is present to enter a
        password.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-W</option></term>
      <term><option>--password</option></term>
      <listitem>
       <para>
        Force <application>vacuumdb</application> to prompt for a
        password before connecting to a database.
       </para>

       <para>
        This option is never essential, since
        <application>vacuumdb</application> will automatically prompt
        for a password if the server demands password authentication.
        However, <application>vacuumdb</application> will waste a
        connection attempt finding out that the server wants a password.
        In some cases it is worth typing <option>-W</option> to avoid the extra
        connection attempt.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--maintenance-db=<replaceable class="parameter">dbname</replaceable></option></term>
      <listitem>
       <para>
        When the <option>-a</option>/<option>--all</option> is used, connect
        to this database to gather the list of databases to vacuum.
        If not specified, the <literal>postgres</literal> database will be used,
        or if that does not exist, <literal>template1</literal> will be used.
        This can be a <link linkend="libpq-connstring">connection
        string</link>.  If so, connection string parameters will override any
        conflicting command line options.  Also, connection string parameters
        other than the database name itself will be re-used when connecting
        to other databases.
       </para>
      </listitem>
     </varlistentry>
    </variablelist>
   </para>
 </refsect1>


 <refsect1>
  <title>Environment</title>

  <variablelist>
   <varlistentry>
    <term><envar>PGDATABASE</envar></term>
    <term><envar>PGHOST</envar></term>
    <term><envar>PGPORT</envar></term>
    <term><envar>PGUSER</envar></term>

    <listitem>
     <para>
      Default connection parameters
     </para>
    </listitem>
   </varlistentry>

   <varlistentry>
    <term><envar>PG_COLOR</envar></term>
    <listitem>
     <para>
      Specifies whether to use color in diagnostic messages. Possible values
      are <literal>always</literal>, <literal>auto</literal> and
      <literal>never</literal>.
     </para>
    </listitem>
   </varlistentry>
  </variablelist>

  <para>
   This utility, like most other <productname>PostgreSQL</productname> utilities,
   also uses the environment variables supported by <application>libpq</application>
   (see <xref linkend="libpq-envars"/>).
  </para>

 </refsect1>


 <refsect1>
  <title>Diagnostics</title>

  <para>
   In case of difficulty, see <xref linkend="sql-vacuum"/>
   and <xref linkend="app-psql"/> for
   discussions of potential problems and error messages.
   The database server must be running at the
   targeted host.  Also, any default connection settings and environment
   variables used by the <application>libpq</application> front-end
   library will apply.
  </para>

 </refsect1>

 <refsect1>
  <title>Examples</title>

   <para>
    To clean the database <literal>test</literal>:
<screen>
<prompt>$ </prompt><userinput>vacuumdb test</userinput>
</screen>
   </para>

   <para>
    To clean and analyze for the optimizer a database named
    <literal>bigdb</literal>:
<screen>
<prompt>$ </prompt><userinput>vacuumdb --analyze bigdb</userinput>
</screen>
   </para>

   <para>
    To clean a single table
    <literal>foo</literal> in a database named
    <literal>xyzzy</literal>, and analyze a single column
    <literal>bar</literal> of the table for the optimizer:
<screen>
<prompt>$ </prompt><userinput>vacuumdb --analyze --verbose --table='foo(bar)' xyzzy</userinput>
</screen></para>

   <para>
    To clean all tables in the <literal>foo</literal> and <literal>bar</literal> schemas
    in a database named <literal>xyzzy</literal>:
<screen>
<prompt>$ </prompt><userinput>vacuumdb --schema='foo' --schema='bar' xyzzy</userinput>
</screen></para>


 </refsect1>

 <refsect1>
  <title>See Also</title>

  <simplelist type="inline">
   <member><xref linkend="sql-vacuum"/></member>
  </simplelist>
 </refsect1>

</refentry>