<?xml version="1.0" encoding="utf-8" ?>
<!DOCTYPE chapter SYSTEM "chapter.dtd">
<chapter>
<header>
<copyright>
<year>2004</year><year>2016</year>
<holder>Ericsson AB. All Rights Reserved.</holder>
</copyright>
<legalnotice>
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions 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 section describes how the Erlang runtime system is configured
for IP communication. It also explains how you can configure it
for your needs by a configuration file.
The information is primarily intended for users with special
configuration needs or problems. There is normally no need
for specific settings for Erlang to function properly on a correctly
IP-configured platform.</p>
<p>When Erlang starts up it reads the <c>Kernel</c> variable
<c><![CDATA[inetrc]]></c>, which, if defined, is to specify the location
and name of a user configuration file. Example:</p>
<code type="none"><![CDATA[
% erl -kernel inetrc '"./cfg_files/erl_inetrc"']]></code>
<p>Notice that the use of an <c><![CDATA[.inetrc]]></c> file, which was
supported in earlier Erlang/OTP versions, is now obsolete.</p>
<p>A second way to specify the configuration file is to set
environment variable <c><![CDATA[ERL_INETRC]]></c> to the full name of
the file. Example (bash):</p>
<code type="none"><![CDATA[
% export ERL_INETRC=./cfg_files/erl_inetrc]]></code>
<p>Notice that the <c>Kernel</c> 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 uses
default configuration settings and a native lookup method that
works correctly under most circumstances. Erlang reads no
information from system <c>inet</c> configuration files (such as
<c>/etc/host.conf</c> and <c>/etc/nsswitch.conf</c>) in these modes,
except for <c>/etc/resolv.conf</c> and <c>/etc/hosts</c> that is read and
monitored for changes on Unix platforms for the internal DNS client
<seealso marker="kernel:inet_res"><c>kernel:inet_res</c></seealso>.</p>
<p>If Erlang is started in long name distributed mode, it needs to
get the domain name from somewhere and reads system <c>inet</c>
configuration files for this information. Any hosts and resolver
information found 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
priority order).</p>
<p>On Windows platforms, Erlang searches the system registry rather than
looks 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
<c>inet</c> configuration files (or system registry):</p>
<list type="bulleted">
<item>Hostnames and host addresses</item>
<item>Domain name</item>
<item>Nameservers</item>
<item>Search domains</item>
<item>Lookup method</item>
</list>
<p>This data can also be specified explicitly in the user
configuration file. This file is to contain lines
of configuration parameters (each terminated with a full stop).
Some parameters add data to the configuration (such as host
and nameserver), others overwrite any previous settings
(such as 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
<c>inet</c> configuration database.</p>
<p>The valid configuration parameters are as follows:</p>
<taglist>
<tag><c><![CDATA[{file, Format, File}.]]></c></tag>
<item>
<p><c><![CDATA[Format = atom()]]></c></p>
<p><c><![CDATA[File = string()]]></c></p>
<p>Specify a system file that Erlang is to read configuration data from.
<c><![CDATA[Format]]></c> tells the parser how the file is to be
interpreted:</p>
<list type="bulleted">
<item><c><![CDATA[resolv]]></c> (Unix resolv.conf)</item>
<item><c><![CDATA[host_conf_freebsd]]></c> (FreeBSD host.conf)</item>
<item><c><![CDATA[host_conf_bsdos]]></c> (BSDOS host.conf)</item>
<item><c><![CDATA[host_conf_linux]]></c> (Linux host.conf)</item>
<item><c><![CDATA[nsswitch_conf]]></c> (Unix nsswitch.conf)</item>
<item><c><![CDATA[hosts]]></c> (Unix hosts)</item>
</list>
<p><c><![CDATA[File]]></c> is to specify the filename with full
path.</p>
</item>
<tag><c><![CDATA[{resolv_conf, File}.]]></c></tag>
<item>
<p><c><![CDATA[File = string()]]></c></p>
<p>Specify a system file that Erlang is to read resolver
configuration from for the internal DNS client
<seealso marker="kernel:inet_res"><c>kernel:inet_res</c></seealso>,
and monitor for changes, even if it does not exist.
The path must be absolute.</p>
<p>This can override the configuration parameters
<c><![CDATA[nameserver]]></c> and
<c><![CDATA[search]]></c> depending on the contents
of the specified file. They can also change any time in the future
reflecting the file contents.</p>
<p>If the file is specified as an empty string <c>""</c>,
no file is read or monitored in the future. This emulates
the old behavior 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 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>
</item>
<tag><c><![CDATA[{hosts_file, File}.]]></c></tag>
<item>
<p><c><![CDATA[File = string()]]></c></p>
<p>Specify a system file that Erlang is to 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 lookup option
<c>file</c> is used.</p>
<p>If the file is specified as an empty string <c>""</c>,
no file is read or monitored in the future. This emulates
the old behavior 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 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>
</item>
<tag><c><![CDATA[{registry, Type}.]]></c></tag>
<item>
<p><c><![CDATA[Type = atom()]]></c></p>
<p>Specify a system registry that Erlang is to read configuration
data from. <c><![CDATA[win32]]></c> is the only valid option.</p>
</item>
<tag><c><![CDATA[{host, IP, Aliases}.]]></c></tag>
<item>
<p><c><![CDATA[IP = tuple()]]></c></p>
<p><c><![CDATA[Aliases = [string()]]]></c></p>
<p>Add host entry to the hosts table.</p>
</item>
<tag><c><![CDATA[{domain, Domain}.]]></c></tag>
<item>
<p><c><![CDATA[Domain = string()]]></c></p>
<p>Set domain name.</p>
</item>
<tag><c><![CDATA[{nameserver, IP [,Port]}.]]></c></tag>
<item>
<p><c><![CDATA[IP = tuple()]]></c></p>
<p><c><![CDATA[Port = integer()]]></c></p>
<p>Add address (and port, if other than default) of the primary
nameserver to use for
<seealso marker="kernel:inet_res"><c>kernel:inet_res</c></seealso>.
</p>
</item>
<tag><c><![CDATA[{alt_nameserver, IP [,Port]}.]]></c></tag>
<item>
<p><c><![CDATA[IP = tuple()]]></c></p>
<p><c><![CDATA[Port = integer()]]></c></p>
<p>Add address (and port, if other than default) of the secondary
nameserver for
<seealso marker="kernel:inet_res"><c>kernel:inet_res</c></seealso>.
</p>
</item>
<tag><c><![CDATA[{search, Domains}.]]></c></tag>
<item>
<p><c><![CDATA[Domains = [string()]]]></c></p>
<p>Add search domains for
<seealso marker="kernel:inet_res"><c>kernel:inet_res</c></seealso>.
</p>
</item>
<tag><c><![CDATA[{lookup, Methods}.]]></c></tag>
<item>
<p><c><![CDATA[Methods = [atom()]]]></c></p>
<p>Specify lookup methods and in which order to try them.
The valid methods are as follows:</p>
<list type="bulleted">
<item><c><![CDATA[native]]></c> (use system calls)</item>
<item><c><![CDATA[file]]></c> (use host data retrieved from system
configuration files and/or the user configuration file)</item>
<item><c><![CDATA[dns]]></c> (use the Erlang DNS client
<seealso marker="kernel:inet_res"><c>kernel:inet_res</c></seealso>
for nameserver queries)</item>
</list>
<p>The lookup method <c><![CDATA[string]]></c> tries to
parse the hostname as an 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>
</item>
<tag><c><![CDATA[{cache_size, Size}.]]></c></tag>
<item>
<p><c><![CDATA[Size = integer()]]></c></p>
<p>Set the resolver cache size. Defaults to 100 DNS records.</p>
</item>
<tag><c><![CDATA[{cache_refresh, Time}.]]></c></tag>
<item>
<p><c><![CDATA[Time = integer()]]></c></p>
<p>Set how often (in milliseconds) the resolver cache for
<seealso marker="kernel:inet_res"><c>kernel:inet_res</c></seealso>
is refreshed (that is, expired DNS records are deleted).
Defaults to 1 hour.</p>
</item>
<tag><c><![CDATA[{timeout, Time}.]]></c></tag>
<item>
<p><c><![CDATA[Time = integer()]]></c></p>
<p>Set the time to wait until retry (in milliseconds) for DNS queries
made by
<seealso marker="kernel:inet_res"><c>kernel;inet_res</c></seealso>.
Defaults to 2 seconds.</p>
</item>
<tag><c><![CDATA[{retry, N}.]]></c></tag>
<item>
<p><c><![CDATA[N = integer()]]></c></p>
<p>Set the number of DNS queries
<seealso marker="kernel:inet_res"><c>kernel:inet_res</c></seealso>
will try before giving up. Defaults to 3.</p>
</item>
<tag><c><![CDATA[{inet6, Bool}.]]></c></tag>
<item>
<p><c><![CDATA[Bool = true | false]]></c></p>
<p>Tells the DNS client
<seealso marker="kernel:inet_res"><c>kernel:inet_res</c></seealso>
to look up IPv6 addresses. Defaults to <c>false</c>.</p>
</item>
<tag><c><![CDATA[{usevc, Bool}.]]></c></tag>
<item>
<p><c><![CDATA[Bool = true | false]]></c></p>
<p>Tells the DNS client
<seealso marker="kernel:inet_res"><c>kernel:inet_res</c></seealso>
to use TCP (Virtual Circuit) instead of UDP. Defaults to
<c>false</c>.</p>
</item>
<tag><c><![CDATA[{edns, Version}.]]></c></tag>
<item>
<p><c><![CDATA[Version = false | 0]]></c></p>
<p>Sets the EDNS version that
<seealso marker="kernel:inet_res"><c>kernel:inet_res</c></seealso>
will use. The only allowed version is zero. Defaults to <c>false</c>,
which means not to use EDNS.</p>
</item>
<tag><c><![CDATA[{udp_payload_size, Size}.]]></c></tag>
<item>
<p><c><![CDATA[N = integer()]]></c></p>
<p>Sets the allowed UDP payload size
<seealso marker="kernel:inet_res"><c>kernel:inet_res</c></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; this is not entirely
correct, as the advertised UDP payload size of the
individual nameserver is what is to be used,
but this simple strategy will do until a more intelligent
(probing, caching) algorithm needs to be implemented.
Default to 1280, which stems from the standard Ethernet MTU size.</p>
</item>
<tag><c><![CDATA[{udp, Module}.]]></c></tag>
<item>
<p><c><![CDATA[Module = atom()]]></c></p>
<p>Tell Erlang to use another primitive UDP module than
<c>inet_udp</c>.</p>
</item>
<tag><c><![CDATA[{tcp, Module}.]]></c></tag>
<item>
<p><c><![CDATA[Module = atom()]]></c></p>
<p>Tell Erlang to use another primitive TCP module than
<c>inet_tcp</c>.</p>
</item>
<tag><c><![CDATA[clear_hosts.]]></c></tag>
<item>
<p>Clear the hosts table.</p>
</item>
<tag><c><![CDATA[clear_ns.]]></c></tag>
<item>
<p>Clear the list of recorded nameservers (primary and secondary).</p>
</item>
<tag><c><![CDATA[clear_search.]]></c></tag>
<item>
<p>Clear the list of search domains.</p>
</item>
</taglist>
</section>
<section>
<title>User Configuration Example</title>
<p>Assume that 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. If lookup fails, Erlang
is to request the data from a nameserver (using the Erlang
DNS client, set to use EDNS allowing larger responses).
The resolver configuration is updated when
its configuration file changes. Also, DNS records
are never to be cached. The user configuration file
(in this example named <c><![CDATA[erl_inetrc]]></c>, stored
in directory <c><![CDATA[./cfg_files]]></c>) can then look as follows
(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 can, for example, be started as follows:</p>
<code type="none"><![CDATA[
% erl -sname my_node -kernel inetrc '"./cfg_files/erl_inetrc"']]></code>
</section>
</chapter>
