s6

Mirror/fork of https://skarnet.org/software/s6/
git clone https://ccx.te2000.cz/git/s6
Log | Files | Refs | README | LICENSE

scandir.html (6147B)


      1 <html>
      2   <head>
      3     <meta name="viewport" content="width=device-width, initial-scale=1.0" />
      4     <meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
      5     <meta http-equiv="Content-Language" content="en" />
      6     <title>s6: scan directories</title>
      7     <meta name="Description" content="s6: scan directory" />
      8     <meta name="Keywords" content="s6 scandir supervision svscan s6-svscan scan directory servicedir" />
      9     <!-- <link rel="stylesheet" type="text/css" href="//skarnet.org/default.css" /> -->
     10   </head>
     11 <body>
     12 
     13 <p>
     14 <a href="index.html">s6</a><br />
     15 <a href="//skarnet.org/software/">Software</a><br />
     16 <a href="//skarnet.org/">skarnet.org</a>
     17 </p>
     18 
     19 <h1> Scan directories </h1>
     20 
     21 <p>
     22  A <em>scan directory</em> is a directory containing a list of
     23 <a href="servicedir.html">service directories</a>, or symbolic links
     24 pointing to service directories.
     25 </p>
     26 
     27 <p>
     28  A scan directory represents a list of services that are supposed to
     29 be supervised. Running <a href="s6-svscan.html">s6-svscan</a> on this
     30 scan directory launches a <em>supervision tree</em>: every service
     31 listed in the scan directory will be supervised.
     32 </p>
     33 
     34 <p>
     35  There is normally only one scan directory per system, although nothing
     36 prevents a system administrator from having more.
     37 <a href="https://cr.yp.to/daemontools.html">daemontools</a> traditionally
     38 uses <tt>/service</tt>, and <a href="http://smarden.org/runit/">runit</a>
     39 traditionally uses <tt>/etc/service</tt>. s6 does not care where your
     40 scan directory is, but I would advise <tt>/service</tt> for compatibility
     41 with daemontools. Depending on your installation, <tt>/service</tt> could
     42 be a symbolic link and point to a place either in a RAM filesystem or in
     43 <tt>/var</tt>.
     44 </p>
     45 
     46 <a name="where">
     47 <h2> Where and how to build a scan directory </h2>
     48 </a>
     49 
     50 <p>
     51  Opinions and practices differ.
     52 </p>
     53 
     54 <p>
     55  It is generally accepted that the place where you store all your
     56 service directories (your "service repository") should <em>not</em> be
     57 used as a scan directory - for a simple reason: you might want to have
     58 service directories for more services than what you want to start at
     59 any given time. In other words, your scan directory will be a <em>subset</em>
     60 of your service repository, so you cannot just run s6-svscan on every
     61 service you have a service directory for. So, the first thing is to
     62 separate your <em>service repository</em>, which is just a storage place
     63 for all the services you might want to manage someday, and your <em>scan
     64 directory</em>, which is a directory representing all the services that
     65 you are <em>currently</em> managing.
     66 </p>
     67 
     68 <h3> Service repository </h3>
     69 
     70 <p>
     71  Where to store your service repository is purely a matter of personal
     72 preference. You just have to be aware that <a href="s6-supervise.html">
     73 s6-supervise</a> needs writable <tt>supervise</tt> and <tt>event</tt>
     74 subdirectories in a service directory it monitors.
     75 </p>
     76 
     77 <h3> Scan directory </h3>
     78 
     79 <p>
     80  Where and how to build your scan directory depends heavily on your boot
     81 system - and on your personal preference too.
     82 </p>
     83 
     84 <p>
     85  Standard <a href="https://cr.yp.to/daemontools.html">daemontools</a> and
     86 <a href="http://smarden.org/runit/">runit</a> installations like to have
     87 a fixed scan directory containing symlinks to service directories located
     88 in the service repository. In other words, the service repository contains
     89 the real <em>working copies</em> of the service directories. This works,
     90 as long as:
     91 </p>
     92 
     93 <ul>
     94  <li> It is possible to create writable <tt>supervise</tt> and <tt>event</tt>
     95 subdirectories in every managed service directory. This can be achieved for
     96 instance via symlinks, or by having the service repository stored on a writable
     97 filesystem. </li>
     98  <li> The scan program (<a href="s6-svscan.html">s6-svscan</a>,
     99 <a href="https://cr.yp.to/daemontools/svscan.html">svscan</a>,
    100 <a href="http://smarden.org/runit/runsvdir.8.html">runsvdir</a>...) is
    101 started late enough for all the necessary filesystems to be mounted. </li>
    102 </ul>
    103 
    104 <p>
    105  My own recommendation would be to have working copies of the service
    106 directories <em>entirely separate</em> from the service repository. The
    107 service repository can be safely stored on the root filesystem, and the
    108 needed directories copied to a RAM filesystem at boot time. The scan
    109 directory can be either the place where the working copies are written,
    110 or another directory containing symlinks to those working copies. (The
    111 latter is useful if you are using the <tt>-t</tt> option to
    112 <a href="s6-svscan.html">s6-svscan</a> with a nonzero argument: copying a
    113 directory is not atomic, but making a symlink is, so there is no risk
    114 of your scanner finding a directory while it is being copied - which
    115 could result in <a href="s6-supervise.html">s6-supervise</a> getting
    116 the wrong information and not managing the service properly.)
    117 </p>
    118 
    119 <p>
    120  An example:
    121 </p>
    122 
    123 <ul>
    124  <li> Have your service repository in <tt>/img/services</tt>, i.e. have
    125 service directories in <tt>/img/services/ftpd</tt>, <tt>/img/services/httpd</tt>,
    126 <tt>/img/services/sshd</tt>, etc. </li>
    127  <li> When booting, make <tt>/tmp</tt> a RAM filesystem, and create the
    128 directories <tt>/tmp/services</tt> and <tt>/tmp/service</tt>. </li>
    129  <li> Have s6-svscan run on <tt>/tmp/service</tt>, as early as possible in your
    130 boot sequence. This is possible whether you want to run s6-svscan
    131 <a href="s6-svscan-1.html">as process 1</a> or <a href="s6-svscan-not-1.html">not</a>. </li>
    132  <li> During the boot sequence, populate <tt>/tmp/services</tt> with copies of the
    133 service directories you need: for instance,
    134  <ul>
    135   <li> <tt>cp -a /img/services/sshd /tmp/services/sshd</tt> </li>
    136   <li> <tt>cp -a /img/services/ftpd /tmp/services/ftpd</tt> </li>
    137   <li> etc. </li>
    138  </ul> </li>
    139  <li> When you are ready to start a service, make a symlink in the
    140 <tt>/tmp/service</tt> <em>scan directory</em> pointing to the working copy of
    141 the service directory you need in <tt>/tmp/services</tt>, then notify s6-svscan.
    142 For instance, to start ftpd and httpd together:
    143 <pre> ln -s ../services/ftpd /tmp/service
    144  ln -s ../services/httpd /tmp/service
    145  s6-svscanctl -a /tmp/service</pre> </li>
    146 </ul>
    147 
    148 </body>
    149 </html>