topic Re: signals in Operating System - HP-UX

signals

khilari — Tue, 05 Dec 2006 11:55:26 GMT

How does SIGTERM terminate a process in a orderly fashion? And how is it different to SIGHUP? Are there any other signal types recommended which would work better then SIGKILL which would just kill the process by force?

Re: signals

Patrick Wallek — Tue, 05 Dec 2006 11:57:56 GMT

Info on signals can be found from:

# man 5 signal

Re: signals

James R. Ferguson — Tue, 05 Dec 2006 12:07:38 GMT

Hi:

In order for a signal to terminate a process in an orderly fashion, the signal must be catchable and it must be caught.

In the shell, arming a 'trap' constitutes the signal handler. When caught, the signal handler code runs and does *only* what you tell it to do. Different actions might be taken upon receipt of different signals. Removing temporary files, flushing file buffers, and releasing shared memory segments are common actions handled.

Some signals are *not* catchable. A kill -9 is the classic case. No signal handler that you write will run when your program receives a SIGKILL. Given that a SIGKILL cannot be caught (trapped) nothing can run to cleanup and hence things like shared memory segments can be left orphaned by your process.

Regards!

...JRF...

Re: signals

Pete Randall — Tue, 05 Dec 2006 12:09:22 GMT

A SIGHUP is a hangup signal, quite often used to get a daemon to re-read it's configuration file. A SIGKILL is a hard kill - it does not give a process a chance to shut down in an orderly fashion, release memory and resources, etc. A SIGTERM attempts to have things shut down gracefully, releasing resources owned by the process.

Pete

Re: signals

spex — Tue, 05 Dec 2006 12:17:38 GMT

> Are there any other signal types recommended which would work better then SIGKILL which would just kill the process by force?

SIGKILL (signum 9) is the "deadliest" signal of all.

PCS

Re: signals

Pete Randall — Tue, 05 Dec 2006 12:24:49 GMT

I should also mention that there is more information available in "man kill".

Pete

Re: signals

James R. Ferguson — Tue, 05 Dec 2006 12:32:22 GMT

Hi (again):

I should add that there are times when a SIGKILL will have no effect. If a process is waiting on an I/O, or has already died but is waiting to be removed from the process table by its parent process (a "zombie"); then a 'kill -9' will have no effect.

Regards!

...JRF...