<?xml version="1.0" encoding="latin1" ?>
<!DOCTYPE chapter SYSTEM "chapter.dtd">
<chapter>
<header>
<copyright>
<year>2004</year><year>2009</year>
<holder>Ericsson AB. All Rights Reserved.</holder>
</copyright>
<legalnotice>
The contents of this file are subject to the Erlang Public License,
Version 1.1, (the "License"); you may not use this file except in
compliance with the License. You should have received a copy of the
Erlang Public License along with this software. If not, it can be
retrieved online at http://www.erlang.org/.
Software distributed under the License is distributed on an "AS IS"
basis, WITHOUT WARRANTY OF ANY KIND, either express or implied. See
the License for the specific language governing rights and limitations
under the License.
</legalnotice>
<title>Inet configuration</title>
<prepared>Peter Andersson</prepared>
<docno></docno>
<date>2004-03-02</date>
<rev>PA1</rev>
<file>inet_cfg.xml</file>
</header>
<section>
<title>Introduction</title>
<p>This chapter tells you how the Erlang runtime system is configured
for IP communication. It also explains how you may configure it
for your own particular needs by means of a configuration file.
The information here is mainly intended for users with special
configuration needs or problems. There should normally be no need
for specific settings for Erlang to function properly on a correctly
IP configured platform. </p>
<p>When Erlang starts up it will read the kernel variable
<c><![CDATA[inetrc]]></c> which, if defined, should specify the location and
name of a user configuration file. Example:</p>
<p><c><![CDATA[% erl -kernel inetrc '"./cfg_files/erl_inetrc"']]></c></p>
<p>Note that the usage of a <c><![CDATA[.inetrc]]></c> file, which was
supported in earlier Erlang versions, is now obsolete.</p>
<p>A second way to specify the configuration file is to set the
environment variable <c><![CDATA[ERL_INETRC]]></c> to the full name of the file. Example (bash):</p>
<p><c><![CDATA[% export ERL_INETRC=./cfg_files/erl_inetrc]]></c></p>
<p>Note that the kernel variable <c><![CDATA[inetrc]]></c> overrides this environment variable.</p>
<p>If no user configuration file is specified and Erlang is started
in non-distributed or short name distributed mode, Erlang will use
default configuration settings and a native lookup method that should
work correctly under most circumstances. Erlang
will not read any information from system inet configuration files
(like /etc/host.conf, /etc/nsswitch.conf, etc) in these modes,
except for /etc/resolv.conf and /etc/hosts that is read and monitored
for changes on Unix platforms for the internal DNS client
<seealso marker="kernel:inet_res">inet_res</seealso>.</p>
<p>If Erlang is started in long name distributed mode, it needs to
get the domain name from somewhere and will read system inet
configuration files for this information. Any hosts and resolver
information found then is also recorded, but not
used as long as Erlang is configured for native lookups. (The
information becomes useful if the lookup method is changed to
<c><![CDATA['file']]></c> or <c><![CDATA['dns']]></c>, see below).</p>
<p>Native lookup (system calls) is always the default resolver method. This
is true for all platforms except VxWorks and OSE Delta where <c><![CDATA['file']]></c>
or <c><![CDATA['dns']]></c> is used (in that order of priority).</p>
<p>On Windows platforms, Erlang will search the system registry rather than
look for configuration files when started in long name distributed mode. </p>
</section>
<section>
<title>Configuration Data</title>
<p>Erlang records the following data in a local database if found in system
inet configuration files (or system registry):</p>
<list type="bulleted">
<item>Host names and addresses</item>
<item>Domain name</item>
<item>Nameservers</item>
<item>Search domains</item>
<item>Lookup method</item>
</list>
<p>This data may also be specified explicitly in the user
configuration file. The configuration file should contain lines
of configuration parameters (each terminated with a full
stop). Some parameters add data to the configuration (e.g. host
and nameserver), others overwrite any previous settings
(e.g. domain and lookup). The user configuration file is always
examined last in the configuration process, making it possible
for the user to override any default values or previously made
settings. Call <c><![CDATA[inet:get_rc()]]></c> to view the state of the
inet configuration database.</p>
<p>These are the valid configuration parameters:</p>
<p></p>
<taglist>
<tag><em><c><![CDATA[{file, Format, File}.]]></c></em></tag>
<item>
<p></p>
<p><c><![CDATA[Format = atom()]]></c></p>
<p><c><![CDATA[File = string()]]></c></p>
<p></p>
<p>Specify a system file that Erlang should read configuration
data from. <c><![CDATA[Format]]></c> tells the parser how the file should be
interpreted: <c><![CDATA[resolv]]></c> (Unix resolv.conf), <c><![CDATA[host_conf_freebsd]]></c>
(FreeBSD host.conf), <c><![CDATA[host_conf_bsdos]]></c> (BSDOS host.conf),
<c><![CDATA[host_conf_linux]]></c> (Linux host.conf), <c><![CDATA[nsswitch_conf]]></c>
(Unix nsswitch.conf) or <c><![CDATA[hosts]]></c> (Unix hosts). <c><![CDATA[File]]></c> should
specify the name of the file with full path.</p>
<p></p>
</item>
<tag><em><c><![CDATA[{resolv_conf, File}.]]></c></em></tag>
<item>
<p></p>
<p><c><![CDATA[File = string()]]></c></p>
<p></p>
<p>Specify a system file that Erlang should read resolver
configuration from for the internal DNS client
<seealso marker="kernel:inet_res">inet_res</seealso>,
and monitor for changes, even if it does not exist.
The path must be absolute.</p>
<p>This may override the configuration parameters
<c><![CDATA[nameserver]]></c> and
<c><![CDATA[search]]></c> depending on the contents
of the specified file. They may also change any time in the future
reflecting the file contents.</p>
<p>If the file is specified as an empty string "",
no file is read nor monitored in the future. This emulates
the old behaviour of not configuring the DNS client when
the node is started in short name distributed mode.</p>
<p>If this parameter is not specified it defaults to
<c><![CDATA[/etc/resolv.conf]]></c> unless the environment variable
<c><![CDATA[ERL_INET_ETC_DIR]]></c> is set which defines
the directory for this file to some maybe other than
<c><![CDATA[/etc]]></c>.</p>
<p></p>
</item>
<tag><em><c><![CDATA[{hosts_file, File}.]]></c></em></tag>
<item>
<p></p>
<p><c><![CDATA[File = string()]]></c></p>
<p></p>
<p>Specify a system file that Erlang should read resolver
configuration from for the internal hosts file resolver
and monitor for changes, even if it does not exist.
The path must be absolute.</p>
<p>These host entries are searched after all added with
<c>{file, hosts, File}</c> above or
<c>{host, IP, Aliases}</c> below when the lookup option
<c>file</c> is used.</p>
<p>If the file is specified as an empty string "",
no file is read nor monitored in the future. This emulates
the old behaviour of not configuring the DNS client when
the node is started in short name distributed mode.</p>
<p>If this parameter is not specified it defaults to
<c><![CDATA[/etc/hosts]]></c> unless the environment variable
<c><![CDATA[ERL_INET_ETC_DIR]]></c> is set which defines
the directory for this file to some maybe other than
<c><![CDATA[/etc]]></c>.</p>
<p></p>
</item>
<tag><em><c><![CDATA[{registry, Type}.]]></c></em></tag>
<item>
<p></p>
<p><c><![CDATA[Type = atom()]]></c></p>
<p></p>
<p>Specify a system registry that Erlang should read configuration
data from. Currently, <c><![CDATA[win32]]></c> is the only valid option.</p>
<p></p>
</item>
<tag><em><c><![CDATA[{host, IP, Aliases}.]]></c></em></tag>
<item>
<p></p>
<p><c><![CDATA[IP = tuple()]]></c></p>
<p><c><![CDATA[Aliases = [string()]]]></c></p>
<p></p>
<p>Add host entry to the hosts table.</p>
<p></p>
</item>
<tag><em><c><![CDATA[{domain, Domain}.]]></c></em></tag>
<item>
<p></p>
<p><c><![CDATA[Domain = string()]]></c></p>
<p></p>
<p>Set domain name.</p>
<p></p>
</item>
<tag><em><c><![CDATA[{nameserver, IP [,Port]}.]]></c></em></tag>
<item>
<p></p>
<p><c><![CDATA[IP = tuple()]]></c></p>
<p><c><![CDATA[Port = integer()]]></c></p>
<p></p>
<p>Add address (and port, if other than default) of primary
nameserver to use for
<seealso marker="kernel:inet_res">inet_res</seealso>.</p>
<p></p>
</item>
<tag><em><c><![CDATA[{alt_nameserver, IP [,Port]}.]]></c></em></tag>
<item>
<p></p>
<p><c><![CDATA[IP = tuple()]]></c></p>
<p><c><![CDATA[Port = integer()]]></c></p>
<p></p>
<p>Add address (and port, if other than default) of secondary
nameserver for
<seealso marker="kernel:inet_res">inet_res</seealso>.</p>
<p></p>
</item>
<tag><em><c><![CDATA[{search, Domains}.]]></c></em></tag>
<item>
<p></p>
<p><c><![CDATA[Domains = [string()]]]></c></p>
<p></p>
<p>Add search domains for
<seealso marker="kernel:inet_res">inet_res</seealso>.</p>
<p></p>
</item>
<tag><em><c><![CDATA[{lookup, Methods}.]]></c></em></tag>
<item>
<p></p>
<p><c><![CDATA[Methods = [atom()]]]></c></p>
<p></p>
<p>Specify lookup methods and in which order to try them.
The valid methods are: <c><![CDATA[native]]></c> (use system calls),
<c><![CDATA[file]]></c> (use host data retrieved from
system configuration files and/or
the user configuration file) or <c><![CDATA[dns]]></c>
(use the Erlang DNS client
<seealso marker="kernel:inet_res">inet_res</seealso>
for nameserver queries).</p>
<p>The lookup method <c><![CDATA[string]]></c> tries to
parse the hostname as a IPv4 or IPv6 string and return
the resulting IP address. It is automatically tried
first when <c><![CDATA[native]]></c> is <em>not</em>
in the <c><![CDATA[Methods]]></c> list. To skip it in this case
the pseudo lookup method <c><![CDATA[nostring]]></c> can be
inserted anywhere in the <c><![CDATA[Methods]]></c> list.</p>
<p></p>
</item>
<tag><em><c><![CDATA[{cache_size, Size}.]]></c></em></tag>
<item>
<p></p>
<p><c><![CDATA[Size = integer()]]></c></p>
<p></p>
<p>Set size of resolver cache. Default is 100 DNS records.</p>
<p></p>
</item>
<tag><em><c><![CDATA[{cache_refresh, Time}.]]></c></em></tag>
<item>
<p></p>
<p><c><![CDATA[Time = integer()]]></c></p>
<p></p>
<p>Set how often (in millisec)
the resolver cache for
<seealso marker="kernel:inet_res">inet_res</seealso>.
is refreshed (i.e. expired DNS records are deleted).
Default is 1 h.</p>
<p></p>
</item>
<tag><em><c><![CDATA[{timeout, Time}.]]></c></em></tag>
<item>
<p></p>
<p><c><![CDATA[Time = integer()]]></c></p>
<p></p>
<p>Set the time to wait until retry (in millisec) for DNS queries
made by
<seealso marker="kernel:inet_res">inet_res</seealso>.
Default is 2 sec.</p>
<p></p>
</item>
<tag><em><c><![CDATA[{retry, N}.]]></c></em></tag>
<item>
<p></p>
<p><c><![CDATA[N = integer()]]></c></p>
<p></p>
<p>Set the number of DNS queries
<seealso marker="kernel:inet_res">inet_res</seealso>
will try before giving up.
Default is 3.</p>
<p></p>
</item>
<tag><em><c><![CDATA[{inet6, Bool}.]]></c></em></tag>
<item>
<p></p>
<p><c><![CDATA[Bool = true | false]]></c></p>
<p></p>
<p>Tells the DNS client
<seealso marker="kernel:inet_res">inet_res</seealso>
to look up IPv6 addresses. Default is false.</p>
<p></p>
</item>
<tag><em><c><![CDATA[{usevc, Bool}.]]></c></em></tag>
<item>
<p></p>
<p><c><![CDATA[Bool = true | false]]></c></p>
<p></p>
<p>Tells the DNS client
<seealso marker="kernel:inet_res">inet_res</seealso>
to use TCP (Virtual Circuit) instead of UDP. Default is false.</p>
<p></p>
</item>
<tag><em><c><![CDATA[{edns, Version}.]]></c></em></tag>
<item>
<p></p>
<p><c><![CDATA[Version = false | 0]]></c></p>
<p></p>
<p>Sets the EDNS version that
<seealso marker="kernel:inet_res">inet_res</seealso>
will use. The only allowed is zero. Default is false
which means to not use EDNS.</p>
<p></p>
</item>
<tag><em><c><![CDATA[{udp_payload_size, Size}.]]></c></em></tag>
<item>
<p></p>
<p><c><![CDATA[N = integer()]]></c></p>
<p></p>
<p>Sets the allowed UDP payload size
<seealso marker="kernel:inet_res">inet_res</seealso>
will advertise in EDNS queries. Also sets the limit
when the DNS query will be deemed too large for UDP
forcing a TCP query instead, which is not entirely
correct since the advertised UDP payload size of the
individual nameserver is what should be used,
but this simple strategy will do until a more intelligent
(probing, caching) algorithm need be implemented.
The default is 1280 which stems from the
standard Ethernet MTU size.</p>
<p></p>
</item>
<tag><em><c><![CDATA[{udp, Module}.]]></c></em></tag>
<item>
<p></p>
<p><c><![CDATA[Module = atom()]]></c></p>
<p></p>
<p>Tell Erlang to use other primitive UDP module than inet_udp.</p>
<p></p>
</item>
<tag><em><c><![CDATA[{tcp, Module}.]]></c></em></tag>
<item>
<p></p>
<p><c><![CDATA[Module = atom()]]></c></p>
<p></p>
<p>Tell Erlang to use other primitive TCP module than inet_tcp.</p>
<p></p>
</item>
<tag><em><c><![CDATA[clear_hosts.]]></c></em></tag>
<item>
<p></p>
<p>Clear the hosts table.</p>
<p></p>
</item>
<tag><em><c><![CDATA[clear_ns.]]></c></em></tag>
<item>
<p></p>
<p>Clear the list of recorded nameservers (primary and secondary).</p>
<p></p>
</item>
<tag><em><c><![CDATA[clear_search.]]></c></em></tag>
<item>
<p></p>
<p>Clear the list of search domains.</p>
<p></p>
</item>
</taglist>
</section>
<section>
<title>User Configuration Example</title>
<p>Here follows a user configuration example.</p>
<p>Assume a user does not want Erlang to use the native lookup method,
but wants Erlang to read all information necessary from start and use
that for resolving names and addresses. In case lookup fails, Erlang
should request the data from a nameserver (using the Erlang
DNS client, set to use EDNS allowing larger responses).
The resolver configuration will be updated when
its configuration file changes, furthermore, DNS records
should never be cached. The user configuration file
(in this example named <c><![CDATA[erl_inetrc]]></c>, stored
in directory <c><![CDATA[./cfg_files]]></c>) could then look like this
(Unix):</p>
<pre>
%% -- ERLANG INET CONFIGURATION FILE --
%% read the hosts file
{file, hosts, "/etc/hosts"}.
%% add a particular host
{host, {134,138,177,105}, ["finwe"]}.
%% do not monitor the hosts file
{hosts_file, ""}.
%% read and monitor nameserver config from here
{resolv_conf, "/usr/local/etc/resolv.conf"}.
%% enable EDNS
{edns,0}.
%% disable caching
{cache_size, 0}.
%% specify lookup method
{lookup, [file, dns]}.</pre>
<p>And Erlang could, for example, be started like this:</p>
<p><c><![CDATA[% erl -sname my_node -kernel inetrc '"./cfg_files/erl_inetrc"']]></c></p>
</section>
</chapter>
