When the system is booting up, init first runs the lines with the action type "sysinit". At this time, the system console may not yet be fully initialized, so the output from these actions must be restricted to a minimum. These are executed in the order they appear in the inittab file. These actions handle very basic initialization, such as /sbin/ioinitrc, and the stty command to initialize the console to a known state.
Once the sysinit actions have been completed, the next step is the actions of type "boot" and "bootwait", again in the order they appear in the inittab file. These actions can expect the console to be basically functional, so more complex output is possible.
The difference between "boot" and "bootwait" would be that "boot" is left to run in the background while init goes forward in the boot process, and "bootwait" causes init to wait for the action to complete before it does anything else. The most important bootwait action is /sbin/bcheckrc, which runs the file system checks, changes the root filesystem to a read/write mode and mounts other filesystems.
At this point, init begins to transition the system to the runlevel defined by the initdefault entry (usually runlevel 3 in HP-UX). This means yet another pass through the inittab file. At this time, it runs the actions with types "once" or "respawn" - and only if the runlevel field is empty or matches the target runlevel.
When you moved your Oracle CRS inittab lines before the /sbin/rc line, the system probably attempted to execute them at this point. But because the system is still in a very "bare" state at this point, these scripts probably failed, possibly without leaving any trace to any logs.
(For further investigation: Read the init.evmd, init.cssd and init.crsd scripts, looking for any tests that would cause the scripts to terminate silently. Would the test conditions be true when the system has not yet completed booting up?)
The /sbin/rc line is the main workhorse of HP-UX startup and shutdown. It runs the /sbin/rc?.d scripts as appropriate. These scripts (and *only* them) have the additional benefit of logging their output to /etc/rc.log.
Once the main boot-up pass through the inittab file is complete, init settles to its normal run-time role of monitoring all its child processes. If one of init's children dies, and it was started by a "respawn" action, init will start that process again. (In modern Unix-style systems, there is usually a safety check that delays the restart if the process has failed for more than X times within the last minute or so.)
Because your Oracle CRS lines are using the "respawn" action type, they probably failed the first time if they were run before /sbin/rc (probably because the system was not yet functional enough for them), then succeeded after init restarted them on its first monitoring pass.
MK
MK
