:: Re: [DNG] New goodies from systemd
Top Page
Delete this message
Reply to this message
Author: Steve Litt
Date:  
To: dng
Subject: Re: [DNG] New goodies from systemd
al3xu5 via Dng said on Mon, 7 Aug 2023 10:57:50 +0200

>Mon, 7 Aug 2023 03:50:41 -0400 - Steve Litt
><slitt@???>:
>
>> Martin Steigerwald said on Sun, 06 Aug 2023 10:06:21 +0200
>>
>> >Lorenzo wrote he focussed on services, not on system
>> >initialization. I expect system initialization to be hard to get
>> >right on all the kinds of systems Debian and Devuan are used on.    

>>
>> If people use a hybrid system, with sysvinit doing system
>> initialization and runit running the daemons, we'll have lots of time
>> to get system initialization right with a runit PID1 init system.
>
>
>This seems an answer to a question I have.
>
>Debian is going to drop SysV (first eliminating sysv scripts from
>individual packages and then, at a certain point, removing it
>completely) in favour of systemd :(


Let's cross that bridge when we come to it.

>
>But runit (and s6) cannot be a complete substitute here and now,
>because there are still missing scripts for system demons and for many
>applications: in fact, runit now works hybridly, using the available
>SysV demons when there are no native runit script ready to handle
>services.


True.

>
>By relying on Sysv, this same hybrid mechanism can only be a temporary
>solution: Devuan -- like all Debian-based distributions which refuse
>systend -- must have alternative init and service management systems
>with all the necessary scripts...


It's temporary, but I don't think it's temporary in days or weeks, I
think it's temporary in months to years.

>
>And here comes the question: which is the "right" init I should choose
>*now* when doing a fresh (Daedalus) installation?


If it were me, using Devuan, I'd say sysvinit.

>
>You suggest an hybrid system, with sysvinit doing system initialization
>(pid 1) and runit running the daemons, that is, start with a sysvinit
>system and then move on to runit for services only.


Yes.

>
>Ok. But I have something confused...
>
>If I understood well, it is because so the system initialization would
>be managed by sysvinit and not by runit. Right?


Right, plus the fact that at the current time, some daemons don't have
runit run scripts.

>
>And then, is this because the management of some system services is
>missing with runit (missing scripts)? Or because the hybrid mechanism
>which allows runit to use sysv scripts does not work for/during system
>initialization? Or what else?


I'm not sure of the meaning of the questions in the preceding
paragraph. Some system services run scripts *are* missing, that's true.
And right now, sysvinit can start processes not yet having a run script.

One more thing: Starting out using sysvinit as PID1 and (in the words
of runit and s6) stage 1 gives users an easy way to start learning
runit. And once you know runit, you know s6 service supervisor: They
just use different words for the same concepts.

SteveT

Steve Litt
Autumn 2022 featured book: Thriving in Tough Times
http://www.troubleshooters.com/bookstore/thrive.htm