Subject: Re: [vserver] Ubuntu guest : killall5 error
From: Herbert Poetzl <herbert@13thfloor.at>
Date: Thu, 25 Sep 2008 18:55:25 +0200

On Thu, Sep 25, 2008 at 06:38:13PM +0200, Jean-François Leroux wrote:
> Sorry Herbert,
> it was late and I was forgetting the basics...
> 
> Well, I'm running gentoo with the latest 'stable' version of
> vserver-sources-2.2.0.6, and util-vserver-0.30.214.
> 
> I usually build Etch vservers but this time I wanted an Ubuntu one
> (hardy). I built it without any problem and once inside I removed the
> usual scripts (all the *mount*, *hwclock*, and so on) from startup
> with update-rc.d,  then removed klogd.
> It is booting with the 'sysv' style.

well, with the sysv init style, you won't get an
init process inside the guest, and you should
not try to send signals to the host init :)

> The vserver behaves normally but when I close it, I get 
> the message given above.

> I have tried to pass a '-o 1' option to killall5 in the the sendsigs
> script, but it doesn't seem to change anything. The two error
> messages, I think, are thrown by two calls to killall5 at the top of
> sendsigs.

either use the 'plain' init style, (then you can
signal the guest init) or avoid messing with the
blend through init

best,
Herbert

> 2008/9/25 Herbert Poetzl <herbert@13thfloor.at>:
> > On Thu, Sep 25, 2008 at 01:34:33AM +0200, Jean-François Leroux wrote:
> >> Hi all,
> >> it's all in the title: when I stop this Ubuntu vserver, I get:
> >>
> >>  * Terminating all remaining processes...
> >> killall5: can't read sid for pid 1
> >>
> >> twice, then the vs stops.
> >
> > first, what kernel, Linux-VServer patch and what
> > util-vserver version :)
> >
> >> I reckon it has something to do with sendigs, but I don't know what to
> >> do as adding pid 1 to the list of processes to omit doesn't seem to
> >> change anything.
> >
> > second, do you run an init inside the guest?
> > i.e. is the init style 'plain' and not 'sysv'?
> >
> >> What should I do ?
> >
> > provide more information :)
> >
> >> Thanks for your ideas :-)
> >
> > best,
> > Herbert
> >
> >