[Glass] How bad is terminating stones by their own kill signal handler?

Mariano Martinez Peck via Glass glass at lists.gemtalksystems.com
Fri Feb 20 11:40:16 PST 2015

On Fri, Feb 20, 2015 at 4:32 PM, Martin McClure <
martin.mcclure at gemtalksystems.com> wrote:

> On 02/20/2015 11:18 AM, Mariano Martinez Peck via Glass wrote:
> > Hi guys,
> >
> > We are migrating to CentOS 7 and after lots of hours I cannot have
> > either a Sys V init.d or systemd script that correctly start and stop
> > all my stones and gems...... I have to give up. Cannot loose more
> > time... (I even opened this
> >
> http://serverfault.com/questions/670176/centos7-systemctl-calling-execstop-on-reboot-after-my-procceses-are-already-kill
> )
> >
> > The start side is working correct. So I wonder...is it risky not to
> > explicitly shutdown stone using shutdown method? I see kill signal is
> > captured correctly and the stone still seems to be doing its sutff
> > correct. But I wonder, is this assumption 100% safe?  I ask regarding:
> >
> > 1) I don't want to manually have to start processes in the next boot ,
> > neither manually or even worst using -R or .. whatever... in other
> > words, I want that the stone starts up without any human action.
> >
> > 2) Of course, I don't want to loose transactions.
> >
> > Thanks in advance and sorry for the bad mood....
> Hi Mariano,
> Killing the stone with SIGTERM (the default for the kill command) is
> generally safe. But there are things to watch out for. For example,
> Linux systems I've worked with often send SIGTERM to running processes
> at shutdown, wait a short time (a second or so, in some cases) then send
> SIGKILL (kill -9) to anything still running.
> If the stone has not completed an orderly shutdown at the time that
> SIGKILL is received, the next startup will be unclean and startup will
> recover from transaction logs and will take longer. But you shouldn't
> lose any transactions.
Thanks for the explanation Martin. In fact, I have experienced exactly
that. Most of the times, it seems the time gemstone needs to shutdown
cleanly is less than whatever timeout the OS waits before really killing
it. But a couple of times happened to me what you said, that the next boot
took longer and then watching the logs gemstone was recovering from logs....

But good to know it is safe and I don't loose transactions.

> Personally, I've been steering well clear of systemd, though I may have
> to learn it eventually... Good luck with it!
Pufff.....can you believe I am since yesterday trying to make it work? This
CentOS 7 changed so many things...first the firewall (now firewalld rather
than iptables, now systemd rather than Sys V......
At least, I do get the stopstone called during shutdown, but it seems too
late because I already received the kill signal. I searched every possible
post and tried every single combination of systemd.... but nothing worked.
So I guess I will have to stay this way (I really don't like to shutdown
things uncleanly...but if gemstone works)..

Thanks Martin,

> Regards,
> -Martin

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gemtalksystems.com/mailman/private/glass/attachments/20150220/3f3f7c1f/attachment.html>

More information about the Glass mailing list