Interesting article on boot ordering

Bill Rugolsky Jr. brugolsky at telemetry-investments.com
Wed Sep 24 13:17:14 UTC 2003


On Wed, Sep 24, 2003 at 08:44:06AM -0400, David T Hollis wrote:
> Bryan W. Headley wrote:
> >Re: speed/quality of sysvinit versus minit or runit. The tradeoffs 
> >you're getting is that minit "knows" one of it's spawned services have 
> >stopped, and can react to that. sysvinit really doesn't know about 
> >errors...
> 
> This is one that a lot of folks are hankering for.  I'm not sure myself 
> if that is necessarily the job of the init program.  There are a great 
> number of services that don't have a tendency to die with any frequency 
> worth adding all of the overhead of an init monitoring deal and in the 
> cases where it is required, there are quite possibly better solutions 
> such as keepalived that can monitor and make more intelligent decisions 
> about the process.

The real problem with sysvinit is that it provides supervision, via
/etc/inittab, and modularity, via /etc/rc.d/init.d/*, but no mechanism
for using both.  Dynamically adding a supervised service involves editing
/etc/inittab -- ugh.

At work, I am currently in the process of deploying runit to supplement
sysvinit for services that I want to monitor.

E.g., I have USB modems for dial-in that I'd like to be able to
auto-configure with mgetty when I move the modem between servers.

Doing this with sysvinit is painful.  I suppose the alternative to runit
would be to hack up sysvinit to grok /etc/inittab.d/

Regards,

   Bill Rugolsky





More information about the fedora-devel-list mailing list