On Mon, 17 Jan 2011 21:54:18 +0100
Lennart Poettering <mzerqung(a)0pointer.de> wrote:
On Mon, 17.01.11 14:31, Petr Pisar (ppisar(a)redhat.com) wrote:
>
> On 2011-01-17, Richard W.M. Jones <rjones(a)redhat.com> wrote:
> >
> > Now when you send a signal to the watchdog daemon, it can take
> > several seconds to shut down.
>
> I hope systemd and similar inits running daemons on foreground solve
> this problem simply by waiting for the child's PID.
Yes, that's what happens in systemd. However we currently use SIGKILL
after a timeout, if the normal kill signal (SIGTERM usually) didn't
work. I have now added to my TODO list to make this SIGKILL step
optional.
SIGKILLing databases and similar stuff is not a good idea indeed, they
may be committing data to disk, and you don't want to corrupt data if
it can be helped.
Simo.
--
Simo Sorce * Red Hat, Inc * New York