Init scripts

Much of the assortment of processes on a running UNIX system is started by shell scripts invoked via /etc/inittab; there are two major schemes for these scripts. All usually start with some sort of global "rc" script.

BSD-style

• /etc/rc (or maybe /sbin/rc) invokes other scripts like /etc/rc.net, /etc/rc.local; it gets the desired run level as an argument which it uses to decide what to start
• An /etc/rc.conf (or maybe /etc/rc.config file provides configuration parameters used by other scripts
• Most local customization (to start up locally-installed software) is done in /etc/rc.local
• /etc/rc.local can get pretty big and complicated

System V-style

• init invokes a global rc script passing the run level as an argument, or maybe selects an rc script (rc1, rc2, rc3 ...) based on run level
• /etc/init.d contains an assortment of scripts, each one of which is concerned only with a particular subsystem; each takes standard arguments like "start" or "stop" that tell it what to do with that subsystem
• /etc/rcN.d (one for each run level N) contains links to /etc/init.d scripts which are run when run level N is selected
• There is a standard naming scheme for the /etc/rcN.d scripts to sequence them properly; "S" or "K" says whether to run the script with its "start" or "stop" argument, a two-digit sequence number sorts them into order, and then a symbolic name follows to indicate the subsystem involved, usually the same as used in the init.d script
• The rc.d "K" scripts are run first, then "S" scripts
• This is much more complicated than the BSD system, but also much more modular, and easier for automated installers to work with; as a sysadmin, rather than editing and rearranging a global rc.local, you instead create, edit, and rename individual init.d scripts and rc.d links

Next ->