Re: cfengine-die help...

[Christian Pearce]

On Wed, 2004-09-15 at 02:58, Predrag Zecevic - Solaris System
Administrator wrote:
> Have you tried w/o echo command?

Yes it hangs.  Even with background set to true.

> Did you tried to debug it?

Yes.  There isn't much debugging statements for that part of the code. 
The cfengine-die causes a SIGALRM which causes the process to
terminate.  a nohup only blocks SIGHUP.  We would have to rewrite the
code.

> 
> Regards ;-)
> 
> Christian Pearce wrote:
> 
> >I mistakenly put that before.  I want the shutdown to go into the
> >background.  If I did && it would wait for shutdown to finish then it
> >would echo "cfengine-die".  Which leaves me where I was before.  I think
> >the shell command is correct.  I am puzzled as to why shutdown gets
> >killed.  hmm maybe nohup doesn't block the SIGALRM.  Only the SIGHUP. 
> >Maybe we need to change this behavior in cfengine.  Considering we
> >aren't using the alarm timer for anything other than 0.  Which is
> >immediate.  This way we could block the hup with a nohup.  Unless I am
> >misunderstanding this.
> >
> >On Tue, 2004-09-14 at 11:03, Predrag Zecevic - Solaris System
> >Administrator wrote:
> >  
> >
> >>Hi Christian,
> >>
> >>you are missing one '&' in command bellow: shouldn't this look like: ... 
> >>&& echo "cfengine-die" ?
> >>Regards ;-)
> >>
> >>Christian Pearce wrote:
> >>
> >>    
> >>
> >>>   # Reboot if scheduled automation occurs
> >>>   reboot::
> >>>       '$(nohup) $(shutdown) -y -g 300 -i 6 "SysNav is rebooting the
> >>>server to install patch Cluster(s)." & echo "cfengine-die"'
> >>>           background=true
> >>>           useshell=true
> >>>
> >>>Can someone explain to me the process under which cfengine kills a
> >>>forked script?  This is the last statement I tried.  I call shutdown
> >>>with nohup, but that doesn't seem to keep cfengine from killing it after
> >>>I echo the cfengine die.  Unfortunately if I don't use cfengine-die
> >>>cfengine will not continue executing.
> >>>
> >>>I assume this is the code that does it
> >>>  
> >>>  if (ptr->timeout != 0)
> >>>     {
> >>>     alarm(0);
> >>>     signal(SIGALRM,SIG_DFL);
> >>>     }
> >>>
> >>>alarm(0) sets the seconds for the signal to be generated to zero.
> >>>signal(SIGALRM, SIG_DFL) executes Terminate by default with the SIG_DFL
> >>>handle.  
> >>>
> >>>But shutdown still gets killed off.
> >>> 
> >>>
> >>>      
> >>>
-- 
Christian Pearce
http://www.commnav.com
http://www.perfectorder.com