rustit/mirror_runit/runit-2.1.2/doc/benefits.html
2024-03-24 23:51:57 +00:00

187 lines
8 KiB
HTML

<!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML//EN">
<html>
<head>
<title>runit - benefits</title>
</head>
<body>
<a href="http://smarden.org/pape/">G. Pape</a><br>
<a href="index.html">runit</a><br>
<hr>
<h1>runit - benefits</h1>
<hr>
<a href="#supervision">Service supervision</a><br>
<a href="#state">Clean process state</a><br>
<a href="#log">Reliable logging facility</a><br>
<a href="#fast">Fast system boot up and shutdown</a><br>
<a href="#portability">Portability</a><br>
<a href="#packaging">Packaging friendly</a><br>
<a href="#smallcode">Small code size</a>
<hr>
<a name="supervision"><h3>Service supervision</h3></a>
Each service is associated with a <i>service directory</i>, and each
service daemon runs as a child process of a supervising
<a href="runsv.8.html">runsv</a> process running in this directory.
The <a href="runsv.8.html">runsv</a> program provides a reliable interface
for signalling the service daemon and controlling the service and
supervisor.
Normally the <a href="sv.8.html">sv</a> program is used to send commands
through this interface, and to query status informations about the service.
<p>
The <a href="runsv.8.html">runsv</a> program supervises the corresponding
service daemon.
By default a service is defined to be up, that means, if the service daemon
dies, it will be restarted.
Of course you can <a href="sv.8.html">tell runsv</a> otherwise.
<p>
This reliable interface to control daemons and supervisors obsoletes
pid-guessing programs, such as <tt>pidof</tt>, <tt>killall</tt>,
<tt>start-stop-daemon</tt>, which, due to guessing, are prone to failures
by design.
It also obsoletes so called <tt>pid-files</tt>, no need for each and every
service daemon to include code to daemonize, to write the new process id
into a file, and to take care that the file is removed properly on shutdown,
which might be very difficult in case of a crash.
<hr>
<a name="state"><h3>Clean process state</h3></a>
<i>runit</i> guarantees each service a clean process state, no matter if the
service is activated for the first time or automatically at boot time,
reactivated, or simply restarted.
This means that the service always is started with the same environment,
resource limits, open file descriptors, and controlling terminals.
<p>
You don't necessarily have that with <i>sysv init</i> scripts for example.
It requires a carefully written init script that reliably cleans up and sets
the process state before starting the service daemon.
This adds even more complexity to the init script in comparison with a run
script used by <i>runit</i>.
Many of today's init scripts don't provide a clean process state, here is
an example on what could happen:
<pre>
# /etc/init.d/foo-daemon start
Starting foo daemon: food.
#
</pre>
Fine.
Everything works, nothing to worry about.
After rebooting the system this shows up on the screen:
<pre>
...
Starting foo daemon: food: command not found
failed.
...
</pre>
The <tt>food</tt> program is installed in <tt>/opt/foo/bin/</tt>.
When starting the service for the first time using the init script, the
<tt>PATH</tt> environment variable contained <tt>/opt/foo/bin</tt>.
After reboot <tt>init</tt> started the service using the init script, but
with a different value for the <tt>PATH</tt> variable, not containing
<tt>/opt/foo/bin</tt>.
Of course the init script should have set <tt>PATH</tt> before starting the
daemon; the problem is that it worked in the first place, and that the error
didn't show up until system reboot.
<p>
With bad init scripts miraculous things could also happen when just doing
<pre>
# /etc/init.d/foo-daemon restart
</pre>
at the command line.
<p>
The clean process state includes open file descriptors, obsoleting the
widely used hack in many service daemons to force-close all file descriptors
that might be open, up to the limit of available file descriptors for the
daemon process (often results in 1024 unnecessary close() system calls in a
great number of service daemon implementations).
<hr>
<a name="log"><h3>Reliable logging facility</h3></a>
The <a href="runsv.8.html">runsv</a> program provides a reliable logging
facility for the service daemon.
If configured, <a href="runsv.8.html">runsv</a> creates a pipe, starts and
supervises an additional log service, redirects the log daemon's standard
input to read from the pipe, and redirects the service daemon's standard
output to write to the pipe.
Restarting the service does not require restarting the log service, and vice
versa.
A good choice for a log daemon is <i>runit</i>'s service logging daemon
<a href="svlogd.8.html">svlogd</a>.
<p>
The service daemon and the log daemon can run with different process states,
and under different user id's.
<i>runit</i> supports easy and reliable logging for service daemons running
chroot'ed.
<p>
If <a href="runsv.8.html">runsv</a> is told to shutdown a service, e.g. at
system shutdown, it ensures that the log service stays up as long as the
corresponding service daemon is running and possibly writing to the log.
<hr>
<a name="fast"><h3>Fast system boot up and shutdown</h3></a>
After the system's one time tasks (stage 1) are done, the system services
are started up in parallel.
The operating system's process scheduler takes care of having the services
available as soon as possible.
<p>
On system shutdown, stage 3 uses <a href="runsv.8.html">runsv</a>'s control
interface to wait until each service daemon is terminated and all logs are
written.
Again, services are taken down in parallel.
As soon as all services are down, system halt or system reboot is initiated.
<hr>
<a name="portability"><h3>Portability</h3></a>
<i>runit</i> comes ready-to-run for Debian GNU/Linux and BSD systems, and
can easily be configured to run on other UNIX systems.
The UNIX system's one time initialization tasks and tasks to shutdown the
system must be identified and <i>runit</i>'s stages 1 and 3 configured
accordingly.
<p>
Stages 1 and 3 handle one time tasks.
They only run for short and exit soon.
Stage 2 handles the system's uptime tasks (via the
<a href="runsvdir.8.html">runsvdir</a> program) and is running the whole
system's uptime.
<p>
<i>runit</i>'s stage 2 is portable across UNIX systems.
<i>runit</i> is well suited for server systems and embedded systems, and
also does its job well on desktop systems.
<hr>
<a name="packaging"><h3>Packaging friendly</h3></a>
<i>runit</i>'s stages 1 and 3 are distribution specific.
They normally are shell scripts, and an operating system distribution with
software package management should adapt these scripts if they need support
for their package management.
The
<a href="http://packages.debian.org/unstable/admin/runit-run.html">
runit-run</a>
Debian package is an attempt to integrate <i>runit</i> into
<a href="http://www.debian.org/">Debian GNU/Linux</a> as an alternative to
the default
<a href="http://packages.debian.org/unstable/base/sysvinit.html">
sysvinit</a>.
<p>
Stage 2 is packaging friendly:
all a software package that provides a service needs to do is to include
a <i>service directory</i> in the package, and to provide a symbolic link
to this directory in <tt>/service/</tt>.
The service will be started within five seconds, and automatically at boot
time.
The package's install and update scripts can use the reliable control
interface to stop, start, restart, or send signals to the service.
On package removal, the symbolic link simply is removed.
The service will be taken down automatically.
<hr>
<a name="smallcode"><h3>Small code size</h3></a>
One of the <i>runit</i> project's principles is to keep the code size
small.
As of version 1.0.0 of <i>runit</i>, the <tt>runit.c</tt> source contains
330 lines of code; the <tt>runsvdir.c</tt> source is 274 lines of code, the
<tt>runsv.c</tt> source 509.
This minimizes the possibility of bugs introduced by programmer's fault,
and makes it more easy for security related people to proofread the source
code.
<p>
The <i>runit</i> core programs have a very small memory footprint and do not
allocate memory dynamically.
<hr>
<address><a href="mailto:pape@smarden.org">
Gerrit Pape &lt;pape@smarden.org&gt;
</a></address>
</body>
</html>