On Fri, Apr 28, 2006 at 05:48:40PM +0200, Youri LACAN-BARTLEY wrote:
> Herbert Poetzl wrote:
> >On Fri, Apr 28, 2006 at 04:20:20PM +0200, Youri LACAN-BARTLEY wrote:
> >>Hi Herbert,
> >>I tried changing the timeout as described in my initial post, the
> >>error occurred well under 120 seconds and definitely under the
> >>default 30 seconds, so what exactly sends the SIGKILL prematurely?
yes, that's what I missed in the first place ...
> >hmm, missed that one, sorry ...
> Well, what I meant was that the error I posted about occurred at
> shutdown within way under 30 seconds. If I understand correctly, a
> SIGKILL is sent to all processes that do not shutdown within a 30
> second time frame by default. I therefore gather that something other
> than the timeout mechanism is sending out the SIGKILL.
precisely, and that's why I suggested to switch to the
plain init style, where the guest runs its own init
which handles the startup and shutdown ...
> >>I'll be using strace on motion to check if that is causing the error.
> >>I might be able to solve the problem by changing init.d script
> >>handling the daemon, who knows.
> >have you tried to use a 'plain' init style on that
> >guest? it seems to me that you are using sysv, maybe
> >your scripts kill the shutdown somehow?
> I have switched to a 'plain' init style on the concerned guest, and
> to error has occurred. So I guess that solved the problem.
yes, and we know two things from this test, first
that you guest is working fine (regarding shutdown)
and second that the guest shutdown (runlevel scripts
contain a killall, which, quite naturally, kills off
the external (sysv) shutdown script too ...
> Are there any downsides to using the 'plain' init style other than
> the less 'human-readable' boot sequence?
except for that fact that you have an additional
process (init) running inside, no ..
> And is switching to 'plain' init style the solution to all timeout
> errors occurring on guest shutdown?
no, you can, now that we 'know' what causes it, look
for the script which executes the killall5 (or whatever
is used to kill of all processes) and remove it from
the proper runlevel, which will isntantly fix the kill
and thus make the sysv shutdown work as expected ...
> Thanks again to all for you for the help, and keep up the good work !
will do so,
> >>I'll keep you posted.
> >>Vserver mailing list
> Vserver mailing list
Vserver mailing list
Received on Sat Apr 29 00:37:57 2006