Why SIGKILL signal not terminate a process?

[Franta Hanzlík]

On my Fedora 19 i686 I see weird thing - when killing processes (by
commands as:

killall -9 kactivitymanagerd
killall -9 gam_server
killall -9 kded4
killall -9 systemd
killall -9 atril

or with PID:

kill -9 1 1322 10612 10619

), then processes stay running - they are not zombies (for PID=1 be
zombie perhaps does not make sense), but eat CPU, occupy memory etc.
I cannot say this behavior is always (I'm killing processes only when
I need it), but I saw this several times, with last Fedora distros.

It is bad glibc signal() implementation or what else?

Regards, Franta Hanzlik

[Franta Hanzlík]

Mateusz Marzantowicz wrote:
>  On 30.11.2013 00:10, Frantisek Hanzlik wrote:
> > Mateusz Marzantowicz wrote:
> >> On 29.11.2013 20:41, Frantisek Hanzlik wrote:
> >>> On my Fedora 19 i686 I see weird thing - when killing processes (by
> >>> commands as:
> >>>
> >>> killall -9 kactivitymanagerd
> >>> killall -9 gam_server
> >>> killall -9 kded4
> >>> killall -9 systemd
> >>> killall -9 atril
> >>>
> >>> or with PID:
> >>>
> >>> kill -9 1 1322 10612 10619
> >>>
> >>> ), then processes stay running - they are not zombies (for PID=1 be
> >>> zombie perhaps does not make sense), but eat CPU, occupy memory etc.
> >>> I cannot say this behavior is always (I'm killing processes only when
> >>> I need it), but I saw this several times, with last Fedora distros.
> >>>
> >>> It is bad glibc signal() implementation or what else?
> >>>
> >>> Regards, Franta Hanzlik
> >>
> >> http://unix.stackexchange.com/questions/5642/what-if-kill-9-does-not-work
> >>
> >> Mateusz Marzantowicz
> >
> > Thank for link, but I'm not a lot smarter.
> > Maybe Linux immunize system init (PID 1) against SIGKILL (but I was
> > working on Unices where SIGKILL to init caused clean & immediate
> > system halt - kernel flush buffers, unmount FSs and halt machine),
> > but what SIGKILL to other processes? I was killing them as root, and
> > as I wrote before, they was not zombies and possibly nor in
> > uninterruptible sleep - 'top' utility shows as they consumes eg.
> > several or several tens percents CPU.
> > Then, according to Your link, is only other possibility bad Linux
> > kernel?
> >
>  
>  Does it happen always regardless system load and ongoing IO operations?
>  It was mentioned in that article that the result of kill -9 might not be
>  immediate depending on what is occupying your kernel.
>  
>  Please also note, that you can't kill systemd (PID 1) on Linux because
>  it's init process and it doesn't have much sense. To reboot your system
>  'properly' use shutdown/halt/reboot or systemctl command. Simply killing
>  init is not save this days.
>  
>  Mateusz Marzantowicz 
Last time I did 'kill -9' was when I had freezed GUI, but it was possible
remotely ssh'd to this machine and working on it. I tried kill some
processes in hope that I can finish programs running in X.
IMO on these Unices 'kill -9 1' was acting as 'halt -f' - when its
man page is correct, it say:
"
Force immediate halt, power-off, reboot. Don't contact the init system.
"
Franta

[Mateusz Marzantowicz]

On 30.11.2013 00:10, Frantisek Hanzlik wrote:
>  Mateusz Marzantowicz wrote:
> > On 29.11.2013 20:41, Frantisek Hanzlik wrote:
> >> On my Fedora 19 i686 I see weird thing - when killing processes (by
> >> commands as:
> >>
> >> killall -9 kactivitymanagerd
> >> killall -9 gam_server
> >> killall -9 kded4
> >> killall -9 systemd
> >> killall -9 atril
> >>
> >> or with PID:
> >>
> >> kill -9 1 1322 10612 10619
> >>
> >> ), then processes stay running - they are not zombies (for PID=1 be
> >> zombie perhaps does not make sense), but eat CPU, occupy memory etc.
> >> I cannot say this behavior is always (I'm killing processes only when
> >> I need it), but I saw this several times, with last Fedora distros.
> >>
> >> It is bad glibc signal() implementation or what else?
> >>
> >> Regards, Franta Hanzlik
> >
> > http://unix.stackexchange.com/questions/5642/what-if-kill-9-does-not-work
> >
> > Mateusz Marzantowicz
>  
>  Thank for link, but I'm not a lot smarter.
>  Maybe Linux immunize system init (PID 1) against SIGKILL (but I was
>  working on Unices where SIGKILL to init caused clean & immediate
>  system halt - kernel flush buffers, unmount FSs and halt machine),
>  but what SIGKILL to other processes? I was killing them as root, and
>  as I wrote before, they was not zombies and possibly nor in
>  uninterruptible sleep - 'top' utility shows as they consumes eg.
>  several or several tens percents CPU.
>  Then, according to Your link, is only other possibility bad Linux
>  kernel?
>   
Does it happen always regardless system load and ongoing IO operations?
It was mentioned in that article that the result of kill -9 might not be
immediate depending on what is occupying your kernel.

Please also note, that you can't kill systemd (PID 1) on Linux because
it's init process and it doesn't have much sense. To reboot your system
'properly' use shutdown/halt/reboot or systemctl command. Simply killing
init is not save this days.



Mateusz Marzantowicz

[Franta Hanzlík]

Mateusz Marzantowicz wrote:
>  On 29.11.2013 20:41, Frantisek Hanzlik wrote:
> > On my Fedora 19 i686 I see weird thing - when killing processes (by
> > commands as:
> >
> > killall -9 kactivitymanagerd
> > killall -9 gam_server
> > killall -9 kded4
> > killall -9 systemd
> > killall -9 atril
> >
> > or with PID:
> >
> > kill -9 1 1322 10612 10619
> >
> > ), then processes stay running - they are not zombies (for PID=1 be
> > zombie perhaps does not make sense), but eat CPU, occupy memory etc.
> > I cannot say this behavior is always (I'm killing processes only when
> > I need it), but I saw this several times, with last Fedora distros.
> >
> > It is bad glibc signal() implementation or what else?
> >
> > Regards, Franta Hanzlik
>  
>  http://unix.stackexchange.com/questions/5642/what-if-kill-9-does-not-work
>  
>  Mateusz Marzantowicz 
Thank for link, but I'm not a lot smarter.
Maybe Linux immunize system init (PID 1) against SIGKILL (but I was
working on Unices where SIGKILL to init caused clean & immediate
system halt - kernel flush buffers, unmount FSs and halt machine),
but what SIGKILL to other processes? I was killing them as root, and
as I wrote before, they was not zombies and possibly nor in
uninterruptible sleep - 'top' utility shows as they consumes eg.
several or several tens percents CPU.
Then, according to Your link, is only other possibility bad Linux
kernel?

Franta

[Mateusz Marzantowicz]

On 29.11.2013 20:41, Frantisek Hanzlik wrote:
>  On my Fedora 19 i686 I see weird thing - when killing processes (by
>  commands as:
>  
>  killall -9 kactivitymanagerd
>  killall -9 gam_server
>  killall -9 kded4
>  killall -9 systemd
>  killall -9 atril
>  
>  or with PID:
>  
>  kill -9 1 1322 10612 10619
>  
>  ), then processes stay running - they are not zombies (for PID=1 be
>  zombie perhaps does not make sense), but eat CPU, occupy memory etc.
>  I cannot say this behavior is always (I'm killing processes only when
>  I need it), but I saw this several times, with last Fedora distros.
>  
>  It is bad glibc signal() implementation or what else?
>  
>  Regards, Franta Hanzlik
>   
http://unix.stackexchange.com/questions/5642/what-if-kill-9-does-not-work



Mateusz Marzantowicz

[poma]

On 29.11.2013 20:41, Frantisek Hanzlik wrote:

>  It is bad glibc signal() implementation or what else? 
What else.

Bee awesome.:!


poma