<!DOCTYPE article SYSTEM "../../../dtd/article.dtd">

<article name="Connection processing methods"
         link="/en/docs/events.html"
         lang="en">

<section>

<para>
nginx supports a variety of connection processing methods.
The availability of a particular method depends on the platform used.
On platforms that support several methods nginx will normally
select the most efficient method automatically.
However, if needed, a connection processing method can be selected
explicitly with the
<link doc="ngx_core_module.xml" id="use"/> directive.
</para>

<para>
The following connection processing methods are supported:
<list type="bullet">

<listitem>
<para>
<literal>select</literal>&mdash;standard method.
The supporting module is built automatically on platforms that lack
more efficient methods.
The <literal>--with-select_module</literal> and
<literal>--without-select_module</literal> configuration parameters
can be used to forcibly enable or disable the build of this module.
</para>
</listitem>

<listitem>
<para>
<literal>poll</literal>&mdash;standard method.
The supporting module is built automatically on platforms that lack
more efficient methods.
The <literal>--with-poll_module</literal> and
<literal>--without-poll_module</literal> configuration parameters
can be used to forcibly enable or disable the build of this module.
</para>
</listitem>

<listitem>
<para>
<literal>kqueue</literal>&mdash;efficient method used on
FreeBSD 4.1+, OpenBSD 2.9+, NetBSD 2.0, and Mac OS X.
</para>
</listitem>

<listitem>
<para>
<literal>epoll</literal>&mdash;efficient method used on
Linux 2.6+.
<note>
Some older distributions like SuSE 8.2 provide patches
that add epoll support to 2.4 kernels.
</note>
</para>
</listitem>

<listitem>
<para>
<literal>rtsig</literal>&mdash;real time signals, efficient method
used on Linux 2.2.19+.
By default,
the system-wide event queue is limited by 1024 signals.
On loaded servers it may become necessary to increase this limit
by changing the <path>/proc/sys/kernel/rtsig-max</path> kernel parameter.
However, in Linux 2.6.6-mm2 this parameter is gone, and each process
now has its own event queue.
The size of each queue is limited by <literal>RLIMIT_SIGPENDING</literal>
(see also <link doc="ngx_core_module.xml" id="worker_rlimit_sigpending"/>).
</para>

<para>
On queue overflow, nginx discards the queue and falls back to
<literal>poll</literal> connection processing method until
the situation gets back to normal.
</para>
</listitem>

<listitem>
<para>
<literal>/dev/poll</literal>&mdash;efficient method used on
Solaris 7 11/99+, HP/UX 11.22+ (eventport), IRIX 6.5.15+,
and Tru64 UNIX 5.1A+.
</para>
</listitem>

<listitem>
<para>
<literal>eventport</literal>&mdash;event ports, efficient method
used on Solaris 10.
</para>
</listitem>

</list>
</para>

</section>

</article>