Jeremy Huntwork wrote:
> TheOldFellow wrote:
>> I hesitate to say this, but: A supervising init
like runit would fix
>> that
>> in a trice. Upstart might be a candidate
eventually too.
>
> At this point, I'd be happy to try something like that
- though I'm not
> sure how much authority I have anymore...
>
> The only problem I see with that is since Belg likes to
just drop random
> processes, what happens if it drops our monitoring
init? I'm wondering
> if I can write something on another server that can
monitor what
> processes Belg is currently running remotely.
What happens on runit, I can't speak for the others, is that
there is a
process 1, called as init by the kernel at start up. If
init fails
you're screwed, but that's true of every other init. Init
starts, and
will respawn if necessary, runsvdir, that actually starts
the monitoring
processes. Each service has a monitoring runsv process that
restarts
the service if it fails, ditto for the log processes.
I wouldn't implement it on belg, too much of a revolution,
but I think
some kind of process supervision is a necessary part of a
high
reliability system.
You can run runit under init, BTW, so it can be tried out
without too
much commitment - in which case it's like daemontools
without the
(expletive deleted) Bernstein.
R.
--
http://linuxfromscratch.org/mailman/listinfo/alfs-discu
ss
FAQ: http://www.linux
fromscratch.org/faq/
Unsubscribe: See the above information page
|