Mailinglist Archive: opensuse-bugs (6177 mails)

[bugzilla_noreply@xxxxxxxxxx]

http://bugzilla.novell.com/show_bug.cgi?id=623460

http://bugzilla.novell.com/show_bug.cgi?id=623460#c9


--- Comment #9 from Jon Nelson <jnelson-suse@xxxxxxxxxxx> 2010-07-19 16:11:44 
UTC ---
(In reply to comment #5)
> Normally the package should not depend on the old behaviour that is
> that 5 seconds should be enough to terminate a normal server daemon.
>
> And yes I've look through out the services.   Currently only cups
> is known to have a problem for restart.
>
> The question is *why* SIGTERM takes that long for xdm od gdm or xdm
> or any other display manager?  And why was this problem discoverd that
> late as the changes was done at Mon Apr 12 17:49:46 CEST 2010.
>
> Compare with bug #595796 and bug #622058

With all due respect, I believe that there still exists an issue with lots of
packages.  A quick "grep TERM /etc/init.d/* | grep killproc" shows 36 of 110
init scripts (on this machine) make use of killproc -TERM in some way.

I recently discovered that haldaemon is *also* susceptible to this particular
issue.

To me, the core issue is the fact that many init scripts are specifying the
-TERM which means that with the new killproc behavior they do not wait _at
all_.  

I would wager that a great many of the init scripts which currently specify
-TERM would not reliably restart their packages if the system were under any
sort of load or the program took more than a fraction of a second to exit.

-- 
Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.