Re: runit-1.0.0 release

[Richard A Downing FBCS <richard@109bean.org.uk>]

On Wed, 11 Feb 2004 20:43:12 +0000
Gerrit Pape <pape@smarden.org> wrote:

> An exit code other than zero just causes a warning, stage 2 only is
> skipped if stage 1 has crashed, e.g. due to a broken binary.  I guess
> that this happened here:
>  http://article.gmane.org/gmane.comp.sysutils.supervision.general/295
>  http://thread.gmane.org/gmane.comp.sysutils.supervision.general/296
> 
> > If I exit -1, it just reports a warning and then enters stage 2 as
> > normal.  I want to be able to say 'things got too difficult in stage
> > 1' and get straight to stage 3 and an orderly shutdown.
> 
> It would be easy to change this in the runit.c source, but I'm curious
> why you need it.  What's the advantage of running stage 3 in this
> situation?, if stage 1 cannot recover, it can possibly reboot
> immediately.

Gerrit,

I have made progress and now run a standard shell in the recovery situation,
modelling my script on the openbsd example.  I would have liked to use sulogin,
which is more secure, but as I was planning not to install sysVinit, and
sulogin is part of that package, it's out for the moment, but see below.

Next problem.  You state above that the stage 1 script can reboot immediately. 
Well, the reboot script that your openbsd example uses is also a sysVinit
script.  If I call 'init 6' (or init 0) from within stage 1, I am, of
course sending a signal to runit!  Runit doesn't accept signals in stage 1!

Catch 22?

I'm beginning to think that I'm going to have to install SysVinit after all.
(I may need wall, halt, poweroff, reboot, shutdown, and sulogin, these are all
part of sysVinit)

Cheers,
Richard.

-- 
Richard A Downing FBCS http://www.109bean.org.uk/
Anyone sending unwanted advertising e-mail to this address will be
charged $25 for network traffic and computing time. By extracting my
address from this message or its header, you agree to these terms.