systemd alternative for Jessie?

classic Classic list List threaded Threaded
32 messages Options
12
Reply | Threaded
Open this post in threaded view
|

RE: systemd alternative for Jessie?

Arno Schuring

> Date: Wed, 14 Oct 2015 17:05:16 +0300
> From: [hidden email]
>
> On Wed, 14 Oct 2015 07:49:08 -0500
> Richard Owlett <[hidden email]> wrote:
>
>> What has the end-user, with a single machine, gained today from
>> the adoption of systemd?
>
> Speaking for myself:
> 1. It took me an hour of googling to write my own working init script.
> It takes me 10 minutes to write my own systemd unit.

The last time I wrote an init script:

- copy /etc/init.d/skeleton
- edit the DAEMON= and DESC= lines


> 2. Boot times improved.

It's been years since I've seen a system where the OS boot took longer
than the BIOS boot. Linux or Windows alike.


As additional anecdata: my systems  consistently boot to
systemd-rescue mode whenever I try systemd. This is because I make
heavy use of auto-unlockable encrypted volumes, and systemd doesn't
support auto-unlocking. I finally got around to moving all my unlock
scripts to the initramfs, only to find that the effort was futile:
even if a volume is already unlocked, systemd drops to a rescue
shell. Immediately typing 'exit' in the rescue shell boots the system
to completion without further errors.

Well, not a problem, right? Systemd was said to be modular, so I'll
just uninstall the systemd cryptsetup component. Har har.

So for all of my systems, systemd doesn't even work. I hope you
understand I don't care about boot time when the boot is unsuccesful.


Regards,
Arno

     
Reply | Threaded
Open this post in threaded view
|

Re: systemd alternative for Jessie?

Nicolas George-4
Le tridi 23 vendémiaire, an CCXXIV, Arno Schuring a écrit :
> It's been years since I've seen a system where the OS boot took longer
> than the BIOS boot. Linux or Windows alike.

You may have noticed that the operator between the time for the POST and the
time for the OS boot is +, not max().

Regards,

--
  Nicolas George

signature.asc (836 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

RE: systemd alternative for Jessie?

Arno Schuring

> Date: Wed, 14 Oct 2015 22:26:53 +0200
> From: [hidden email]
>
> Le tridi 23 vendémiaire, an CCXXIV, Arno Schuring a écrit :
>> It's been years since I've seen a system where the OS boot took longer
>> than the BIOS boot. Linux or Windows alike.
>
> You may have noticed that the operator between the time for the POST and the
> time for the OS boot is +, not max().

Of course. But when your BIOS boot time is 15 seconds, it hardly makes
any difference if your OS boots in 3 seconds or in 2.1 seconds, does it?


Regards,
Arno

     
Reply | Threaded
Open this post in threaded view
|

Re: systemd alternative for Jessie?

Nicolas George-4
Le tridi 23 vendémiaire, an CCXXIV, Arno Schuring a écrit :
> Of course. But when your BIOS boot time is 15 seconds, it hardly makes
> any difference if your OS boots in 3 seconds or in 2.1 seconds, does it?

Five percent: small but not negligible. For low-end laptops, the difference
is more significant, I experience it every day. Not even mentioning the
shutdown time.

Regards,

--
  Nicolas George

signature.asc (836 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: systemd alternative for Jessie?

Alex Moonshine-2
In reply to this post by Arno Schuring
On Wed, 14 Oct 2015 21:51:17 +0200
Arno Schuring <[hidden email]> wrote:

> The last time I wrote an init script:
>
> - copy /etc/init.d/skeleton
> - edit the DAEMON= and DESC= lines

I'll be completely honest, I just never understood sysvinit. I'm
alright with bsd-style Slackware single-rc.d-folder layout, where all
the actual scripts are and you just chmod +x them to enable/disable.
Sysv layout just confuses the hell out of me with all different folders
for each runlevel, symlinks instead of actual scripts, those weird
prefixes (S01, K01, what is this?). I always felt too intimidated by
all that to try and figure out how it all works.
Now I happily don't have to, systemd somehow was easier for me to
understand. But to each his own.


> > 2. Boot times improved.  
>
> It's been years since I've seen a system where the OS boot took longer
> than the BIOS boot. Linux or Windows alike.

Maybe you use SSDs, I don't know.
My main desktop is a fairly modern machine (core i5, z77 chipset, 8 gb
ram)
BIOS to grub > 4-5 seconds
grub to login screen with sysvinit ~ 12-14 seconds, with systemd ~ 5-7
seconds. I've never in my life seen any OS load in 2-3 seconds.

> So for all of my systems, systemd doesn't even work. I hope you
> understand I don't care about boot time when the boot is unsuccesful.
>

Fair enough. It works for all 4 of my systems where I have it. The one
where I have init (Slackware, the single-rc.d-folder one) works too.
Everything works :)

--
Time is an illusion. Lunchtime doubly so.

Best wishes,
Alex S.

Reply | Threaded
Open this post in threaded view
|

Re: systemd alternative for Jessie?

Mart van de Wege
In reply to this post by Alex Moonshine-2
Alex Moonshine <[hidden email]> writes:

> On Wed, 14 Oct 2015 07:49:08 -0500
> Richard Owlett <[hidden email]> wrote:
>
>> What has the end-user, with a single machine, gained today from
>> the adoption of systemd?
>
> Speaking for myself:
> 1. It took me an hour of googling to write my own working init script.
> It takes me 10 minutes to write my own systemd unit.
> 2. Boot times improved.
>
Autostarting programs on login is a lot simpler now, and unified across
desktop environments; I'm no longer dependent on the capabilities of the
session manager, and I can even autostart programs I need when logging in
on the console.

In my case, having a running emacs server is indispensable, so I wrote a
systemd user service unit.

Mart

--
"We will need a longer wall when the revolution comes."
    --- AJS, quoting an uncertain source.

Reply | Threaded
Open this post in threaded view
|

Re: systemd alternative for Jessie?

Dan Ritter-4
In reply to this post by Arno Schuring
On Wed, Oct 14, 2015 at 10:47:48PM +0200, Arno Schuring wrote:

>
> > Date: Wed, 14 Oct 2015 22:26:53 +0200
> > From: [hidden email]
> >
> > Le tridi 23 vendémiaire, an CCXXIV, Arno Schuring a écrit :
> >> It's been years since I've seen a system where the OS boot took longer
> >> than the BIOS boot. Linux or Windows alike.
> >
> > You may have noticed that the operator between the time for the POST and the
> > time for the OS boot is +, not max().
>
> Of course. But when your BIOS boot time is 15 seconds, it hardly makes
> any difference if your OS boots in 3 seconds or in 2.1 seconds, does it?

It can take our hardware RAID cards upwards of five minutes. We
reboot about once a month for kernel upgrades; several months if
we're lucky.

Meanwhile, a Wheezy desktop with a cheap SSD boots from cold to
usable in less than 19 seconds, which is less than the 30
seconds it takes a TCP session to consider timing out.

-dsr-

Reply | Threaded
Open this post in threaded view
|

Re: systemd alternative for Jessie?

Joel Rees-3
In reply to this post by Mart van de Wege
You say potato, I say, potato, ...

On Thu, Oct 15, 2015 at 2:56 PM, Mart van de Wege <[hidden email]> wrote:

> Alex Moonshine <[hidden email]> writes:
>
>> On Wed, 14 Oct 2015 07:49:08 -0500
>> Richard Owlett <[hidden email]> wrote:
>>
>>> What has the end-user, with a single machine, gained today from
>>> the adoption of systemd?
>>
>> Speaking for myself:
>> 1. It took me an hour of googling to write my own working init script.

It takes me varying amounts of time, depending on how much I remember
of the last time I did a particular type of script on a particular OS.

>> It takes me 10 minutes to write my own systemd unit.

For me, systemd only adds one more layer of complexity.

>> 2. Boot times improved.

Since I don't force my file systems to be unified, boot times
sometimes shoot to near infinity. (The dreaded ctrl-d, which the
systemd cabal punted on. At least, now I know not to play with that.)

> Autostarting programs on login is a lot simpler now, and unified across

Simplified according to whose definition of simple?

Not mine.

> desktop environments; I'm no longer dependent on the capabilities of the
> session manager, and I can even autostart programs I need when logging in
> on the console.

My guess is that you are now dependent, not only on the session
manager, but on systemd's handling the sesion manager properly as your
proxy. Which is just fine for you, as long as the systemd cabal is
channelling your techniques. (Or is it, as long as you are cannelling
their current favorite techniques?)

> In my case, having a running emacs server is indispensable, so I wrote a
> systemd user service unit.

emacs.

:)

I gave up trying to channel RMS and his buddies on
control-esc-whatever sequences. I understand that the control key is
effectively a pseudo-command-mode switch, but I'm comfortable with not
having to hold the switch down. (esc-colon-command).

And I don't have emacs doing shell stuff for me. I guess I'm more
comfortable with using more mnemonic names for my one-offs (perl,
python, gforth, whatever I've been playing with recently when the
shell itself isn't enough.)

Puh-taw-toe, puh-tay-toe.

> Mart
>
> --
> "We will need a longer wall when the revolution comes."
>     --- AJS, quoting an uncertain source.
>

--
Joel Rees

Be careful when you look at conspiracy.
Arm yourself with knowledge of yourself, as well:
http://reiisi.blogspot.jp/2011/10/conspiracy-theories.html

Reply | Threaded
Open this post in threaded view
|

Re: systemd alternative for Jessie?

tom arnall-2
In reply to this post by tom arnall-2
what prevents Debian from providing an alternate boot option in Jessie
which does not use systemd? My Wheezy system seems to do this.

Reply | Threaded
Open this post in threaded view
|

Re: systemd alternative for Jessie?

David Wright-3
Quoting tom arnall ([hidden email]):
> what prevents Debian from providing an alternate boot option in Jessie
> which does not use systemd? My Wheezy system seems to do this.

(Yes, this is the correct thread for this question.)

Wheezy had systemd included as a "technology preview".
I don't think Debian does technology retrospectives.

Cheers,
David.

Reply | Threaded
Open this post in threaded view
|

Re: systemd alternative for Jessie?

Lisi Reisz
In reply to this post by tom arnall-2
On Saturday 17 October 2015 00:39:18 tom arnall wrote:
> what prevents Debian from providing an alternate boot option in Jessie
> which does not use systemd? My Wheezy system seems to do this.

Nothing.  It provides one.

Lisi

Reply | Threaded
Open this post in threaded view
|

Re: systemd alternative for Jessie?

Lisi Reisz
In reply to this post by David Wright-3
On Saturday 17 October 2015 01:33:42 David Wright wrote:
> Quoting tom arnall ([hidden email]):
> > what prevents Debian from providing an alternate boot option in Jessie
> > which does not use systemd? My Wheezy system seems to do this.
>
> (Yes, this is the correct thread for this question.)
>
> Wheezy had systemd included as a "technology preview".
> I don't think Debian does technology retrospectives.

It does provide an alternative option.

https://packages.debian.org/search?keywords=sysvinit&searchon=names&suite=stable&section=all
https://packages.debian.org/search?suite=stable&section=all&arch=any&searchon=names&keywords=upstart

I didn't search for anything else.

Lisi

12