Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.debian.user > #195317 > unrolled thread

Removing libsystemd0 from a non-systemd system

Back to article view | Back to linux.debian.user

Contents

  Removing libsystemd0 from a non-systemd system David Griffith <dave@661.org> - 2018-05-06 05:20 +0200
    Re: Removing libsystemd0 from a non-systemd system <tomas@tuxteam.de> - 2018-05-06 09:00 +0200
    Re: Removing libsystemd0 from a non-systemd system Patrick Bartek <nemommxiv@gmail.com> - 2018-05-06 19:20 +0200
      Re: Removing libsystemd0 from a non-systemd system David Griffith <dave@661.org> - 2018-05-07 03:50 +0200
        Re: Removing libsystemd0 from a non-systemd system The Wanderer <wanderer@fastmail.fm> - 2018-05-07 04:30 +0200
          Re: Removing libsystemd0 from a non-systemd system David Griffith <dave@661.org> - 2018-05-07 08:40 +0200
            Re: Removing libsystemd0 from a non-systemd system Ansgar Burchardt <"Ansgar Burchardt"@43-1.org> - 2018-05-07 10:00 +0200
        Re: Removing libsystemd0 from a non-systemd system Patrick Bartek <nemommxiv@gmail.com> - 2018-05-07 22:00 +0200
          Re: Removing libsystemd0 from a non-systemd system Abdullah Ramazanoğlu <ar018@yahoo.com> - 2018-05-08 01:50 +0200
            Re: Removing libsystemd0 from a non-systemd system <tomas@tuxteam.de> - 2018-05-08 06:40 +0200
              Re: Removing libsystemd0 from a non-systemd system David Griffith <dave@661.org> - 2018-05-08 09:00 +0200
              Re: Removing libsystemd0 from a non-systemd system Abdullah Ramazanoğlu <ar018@yahoo.com> - 2018-05-08 11:30 +0200

#195317 — Removing libsystemd0 from a non-systemd system

Have any advances been made in figuring out just how to remove libsystemd0 
from a Debian 9 machine that's running sysvinit?  The ongoing presence of 
libsystemd0 has caused slowly-progressing trouble with several machines of 
mine culminating in complete failure a couple days ago.  Initially I 
thought this was unrelated to systemd, but now I tracked it down to 
systemd's remnants and the problem is progressing much faster with 
freshly-installed machines.


-- 
David Griffith
dave@661.org

A: Because it fouls the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?

[toc] | [next] | [standalone]

#195319

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Sun, May 06, 2018 at 02:44:16AM +0000, David Griffith wrote:
> 
> Have any advances been made in figuring out just how to remove
> libsystemd0 from a Debian 9 machine that's running sysvinit?  The
> ongoing presence of libsystemd0 has caused slowly-progressing
> trouble with several machines of mine culminating in complete
> failure a couple days ago.  Initially I thought this was unrelated
> to systemd, but now I tracked it down to systemd's remnants and the
> problem is progressing much faster with freshly-installed machines.

Could you be a bit more precise?

At least one example where libsystemd0 is causing problems would
be enlightening. Pointers welcome.

Cheers
- -- tomás
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)

iEYEARECAAYFAlruogQACgkQBcgs9XrR2katOgCfUJNU4QWrXkzsn4AJWEsNITfX
x8wAn0sszJyo1Qnxze12dVd83An17598
=wLyA
-----END PGP SIGNATURE-----

[toc] | [prev] | [next] | [standalone]

#195356

On Sun, 6 May 2018 02:44:16 +0000 (UTC) David Griffith <dave@661.org>
wrote:

> 
> Have any advances been made in figuring out just how to remove
> libsystemd0 from a Debian 9 machine that's running sysvinit?  The
> ongoing presence of libsystemd0 has caused slowly-progressing trouble
> with several machines of mine culminating in complete failure a
> couple days ago.  Initially I thought this was unrelated to systemd,
> but now I tracked it down to systemd's remnants and the problem is
> progressing much faster with freshly-installed machines.
> 
> 
> -- 
> David Griffith
> dave@661.org

First, how exactly did you convert to sysvinit, etc? And what kind of
trouble?

I've been running Stretch with sysvinit for almost a year -- as
a personal machine, not a server -- and have had absolutely NO
problems.  Here's the link I used:

http://without-systemd.org/wiki/index.php/How_to_remove_systemd_from_a_Debian_Stretch_installation

I used the very first conversion steps, the simplest one, and none
of the optional ones.  No pinning.  No third-party systemdless repos,
etc. I still have systemd libraries including libsystemd0 for those apps
that have systemd as a dependenciy.  No problems.  Totally removing
systemd is a pain requiring third-party systemdless repos and keeping a
wary eye out for problems. I did it a couple times as part of my
experiments, and always had glitches.

One thing did just occur to me: Are you using the GNOME desktop?  I've
heard stories about it and systemd.  It is VERY dependent on it. I
haven't used GNOME whatever version for about 7 years. I use only a
window manager Openbox.

B

[toc] | [prev] | [next] | [standalone]

#195377

On Sun, 6 May 2018, Patrick Bartek wrote:

> On Sun, 6 May 2018 02:44:16 +0000 (UTC) David Griffith <dave@661.org>
> wrote:
>
>> Have any advances been made in figuring out just how to remove 
>> libsystemd0 from a Debian 9 machine that's running sysvinit?  The 
>> ongoing presence of libsystemd0 has caused slowly-progressing trouble 
>> with several machines of mine culminating in complete failure a couple 
>> days ago.  Initially I thought this was unrelated to systemd, but now I 
>> tracked it down to systemd's remnants and the problem is progressing 
>> much faster with freshly-installed machines.
>
> First, how exactly did you convert to sysvinit, etc? And what kind of
> trouble?
>
> I've been running Stretch with sysvinit for almost a year -- as
> a personal machine, not a server -- and have had absolutely NO
> problems.  Here's the link I used:
>
> http://without-systemd.org/wiki/index.php/How_to_remove_systemd_from_a_Debian_Stretch_installation
>
> I used the very first conversion steps, the simplest one, and none
> of the optional ones.  No pinning.  No third-party systemdless repos,
> etc. I still have systemd libraries including libsystemd0 for those apps
> that have systemd as a dependenciy.  No problems.  Totally removing
> systemd is a pain requiring third-party systemdless repos and keeping a
> wary eye out for problems. I did it a couple times as part of my
> experiments, and always had glitches.
>
> One thing did just occur to me: Are you using the GNOME desktop?  I've
> heard stories about it and systemd.  It is VERY dependent on it. I
> haven't used GNOME whatever version for about 7 years. I use only a
> window manager Openbox.

I followed that same thing you did as soon as the machine was installed. 
I also did optional steps 2 and 3.  I didn't do 1 because all the machines 
in question are headless.  I stopped using GNOME when version 3 came out 
and switched to MATE for most of my desktop needs.

One of the symptoms that made me think libsystemd0 had something to do 
with it was the output of "apt-get upgrade".  It would always report "1 
not upgraded" or "2 not upgraded".

The trouble manifested in dependency hell and networking that would 
mysteriously stop for no readily apparent reason (on reboot after kernel 
upgrade or out of the blue).  Usually networking could be regained by 
doing a LISH login and manually turning on the network interfaces.  Then 
interface names started changing randomly.  This was after names like 
"eth0" and friends were abandoned.  Servers died by way of networking only 
working halfway, no matter what I did.  I was able to ssh in and do scp 
and rsync transfers, but that was about it.

What's the point of allowing libsystemd0 to exist when systemd has been 
purged?

Is anyone working on a mechanism to allow for install-time selection of a 
desired init?  I brought this up a few times since systemd came to Debian, 
but I've never heard anything more on this.


-- 
David Griffith
dave@661.org

A: Because it fouls the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?

[toc] | [prev] | [next] | [standalone]

#195380

[Multipart message — attachments visible in raw view] — view raw

On 2018-05-06 at 21:25, David Griffith wrote:

> What's the point of allowing libsystemd0 to exist when systemd has
> been purged?

So that programs which interface with systemd can detect whether or not
systemd is present, and fall back to alternate code paths when it's not.

As I understand matters (without having actually dug into the code),
that detection code literally is what libsystemd0 *is*; when systemd is
present, it passes through function calls to be handled in appropriate
places, and when systemd is not present, it returns an appropriate
default or failure value.

-- 
   The Wanderer

The reasonable man adapts himself to the world; the unreasonable one
persists in trying to adapt the world to himself. Therefore all
progress depends on the unreasonable man.         -- George Bernard Shaw

[toc] | [prev] | [next] | [standalone]

#195390

On Sun, 6 May 2018, The Wanderer wrote:

> On 2018-05-06 at 21:25, David Griffith wrote:
>
>> What's the point of allowing libsystemd0 to exist when systemd has
>> been purged?
>
> So that programs which interface with systemd can detect whether or not
> systemd is present, and fall back to alternate code paths when it's not.
>
> As I understand matters (without having actually dug into the code),
> that detection code literally is what libsystemd0 *is*; when systemd is
> present, it passes through function calls to be handled in appropriate
> places, and when systemd is not present, it returns an appropriate
> default or failure value.

I was under the impression that systemd-shim provided this functionality. 
When I look at https://packages.debian.org/stretch/systemd-shim and 
https://packages.debian.org/stretch/libsystemd0, their functions appear to 
be identical except that the latter actually does talk to systemd if 
systemd is present.

-- 
David Griffith
dave@661.org

A: Because it fouls the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?

[toc] | [prev] | [next] | [standalone]

#195394

David Griffith writes:
> On Sun, 6 May 2018, The Wanderer wrote:
>> On 2018-05-06 at 21:25, David Griffith wrote:
>>
>>> What's the point of allowing libsystemd0 to exist when systemd has
>>> been purged?
>>
>> So that programs which interface with systemd can detect whether or not
>> systemd is present, and fall back to alternate code paths when it's not.
>>
>> As I understand matters (without having actually dug into the code),
>> that detection code literally is what libsystemd0 *is*; when systemd is
>> present, it passes through function calls to be handled in appropriate
>> places, and when systemd is not present, it returns an appropriate
>> default or failure value.
>
> I was under the impression that systemd-shim provided this
> functionality.

No, that's wrong.

> When I look at
> https://packages.debian.org/stretch/systemd-shim and
> https://packages.debian.org/stretch/libsystemd0, their functions
> appear to be identical except that the latter actually does talk to
> systemd if systemd is present.

systemd-shim provides a daemon that allows some parts of systemd to work
even when systemd does not provide pid-1.  For desktop systems that
means systemd-logind will be able to run.  (logind wants to start/stop
some processes using an interface provided by systemd's pid-1 to manage
cgroups; there is a fork (elogind) that just merged the cgroup
management into the logind binary.)

libsystemd0 is a shared library providing some functions to applications
linking against the library.  Some parts work without any daemons
installed, some parts want to talk to some daemons provided by systemd
services, depending on which functionality one uses.

A daemon with a Dbus interface and a shared library are quite different
things, especially when taking about why some package dependencies
exist.

Ansgar

[toc] | [prev] | [next] | [standalone]

#195455

On Mon, 7 May 2018 01:25:15 +0000 (UTC) David Griffith <dave@661.org>
wrote:

> On Sun, 6 May 2018, Patrick Bartek wrote:
> 
> > On Sun, 6 May 2018 02:44:16 +0000 (UTC) David Griffith
> > <dave@661.org> wrote:
> >
> >> Have any advances been made in figuring out just how to remove 
> >> libsystemd0 from a Debian 9 machine that's running sysvinit?  The 
> >> ongoing presence of libsystemd0 has caused slowly-progressing
> >> trouble with several machines of mine culminating in complete
> >> failure a couple days ago.  Initially I thought this was unrelated
> >> to systemd, but now I tracked it down to systemd's remnants and
> >> the problem is progressing much faster with freshly-installed
> >> machines.
> >
> > First, how exactly did you convert to sysvinit, etc? And what kind
> > of trouble?
> >
> > I've been running Stretch with sysvinit for almost a year -- as
> > a personal machine, not a server -- and have had absolutely NO
> > problems.  Here's the link I used:
> >
> > http://without-systemd.org/wiki/index.php/How_to_remove_systemd_from_a_Debian_Stretch_installation
> >
> > I used the very first conversion steps, the simplest one, and none
> > of the optional ones.  No pinning.  No third-party systemdless
> > repos, etc. I still have systemd libraries including libsystemd0
> > for those apps that have systemd as a dependenciy.  No problems.
> > Totally removing systemd is a pain requiring third-party
> > systemdless repos and keeping a wary eye out for problems. I did it
> > a couple times as part of my experiments, and always had glitches.
> >
> > One thing did just occur to me: Are you using the GNOME desktop?
> > I've heard stories about it and systemd.  It is VERY dependent on
> > it. I haven't used GNOME whatever version for about 7 years. I use
> > only a window manager Openbox.
> 
> I followed that same thing you did as soon as the machine was
> installed. I also did optional steps 2 and 3.  I didn't do 1 because
> all the machines in question are headless.  I stopped using GNOME
> when version 3 came out and switched to MATE for most of my desktop
> needs.

I started with a basic terminal-only netinstall system. No X. No GUI,
etc. Once I had that initial system running, I converted it to sysvinit,
rebooted, and then added the rest leaving the systemd libraries
installed treating them like any other dependency. I left the system to
boot to a terminal where I log in and run startx, if I need the GUI.
So, no login manager or session manager.

I decided after numerous experiments with and without systemd
installed, and with and without third-party "no systemd" repos that
overall leaving the systemd libraries installed did nothing except take
up space, and they were there if some app needed them. I tried never to
install anything that had systemd or any part of it as a direct
dependency.  So far, I've had no systemd issues.

> One of the symptoms that made me think libsystemd0 had something to
> do with it was the output of "apt-get upgrade".  It would always
> report "1 not upgraded" or "2 not upgraded".

I experienced issues with trying to totally remove systemd.  That's one
reason I went with the just leaving systemd installed philosophy.  Plus,
I was wary of the third-party "no systemd" repos.  Would the owner keep
them up-to-date, etc.? I wanted a stable, unbreakable system.

That "not upgraded" thing is the default to prevent, for stability's
sake, major number upgrades of installed files.  Just do an "apt-get
dist-upgrade" to override.  So, you installed 3.2.0.24 of something. A
standard upgrade will only install 3.2.X.XX.  A dist-upgrade will
install 3.4.X.XX for example, but will not install 4.X.X.XX.

> The trouble manifested in dependency hell and networking that would 
> mysteriously stop for no readily apparent reason (on reboot after
> kernel upgrade or out of the blue).  Usually networking could be
> regained by doing a LISH login and manually turning on the network
> interfaces.  Then interface names started changing randomly.  This
> was after names like "eth0" and friends were abandoned.  Servers died
> by way of networking only working halfway, no matter what I did.  I
> was able to ssh in and do scp and rsync transfers, but that was about
> it.

Can't help here.  Probably, removing some systemd library that some
utility needs is the culprit. I've discovered that sometimes something
that doesn't have a direct systemd dependency may depend on
something that does. That's why among other reasons I abandoned the "no
systemd" approach.

> What's the point of allowing libsystemd0 to exist when systemd has
> been purged?

That was answered by others.
 
> Is anyone working on a mechanism to allow for install-time selection
> of a desired init?  I brought this up a few times since systemd came
> to Debian, but I've never heard anything more on this.

I asked that question when Jessie debuted.  No one is.  Certainly not
Debian. The only suggestion I got was to build my own preseeded install
disk, but that's not the answer -- You still have to install systemd
first and convert.

B

[toc] | [prev] | [next] | [standalone]

#195470

On Mon, 7 May 2018 12:40:13 -0700 Patrick Bartek said:
> On Mon, 7 May 2018 01:25:15 +0000 (UTC) David Griffith <dave@661.org>
> wrote:
  
>> Is anyone working on a mechanism to allow for install-time selection
>> of a desired init?  I brought this up a few times since systemd came
>> to Debian, but I've never heard anything more on this.  
> 
> I asked that question when Jessie debuted.  No one is.  Certainly not
> Debian. The only suggestion I got was to build my own preseeded install
> disk, but that's not the answer -- You still have to install systemd
> first and convert.

This is probably because Debian is commited to systemd and going that way
further and further. The fact that today it is possible to convert to sysvinit
is because systemd migration is still in transient. Upcoming release or the
next one possibly won't even allow switching to sysvinit easily (or at all).

Regards
-- 
Abdullah Ramazanoğlu

[toc] | [prev] | [next] | [standalone]

#195479

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Tue, May 08, 2018 at 02:23:57AM +0300, Abdullah Ramazanoğlu wrote:

[...]

> >> Is anyone working on a mechanism to allow for install-time selection
> >> of a desired init?  I brought this up a few times since systemd came
> >> to Debian, but I've never heard anything more on this.  
> > 
> > I asked that question when Jessie debuted.  No one is.  Certainly not
> > Debian. The only suggestion I got was to build my own preseeded install
> > disk, but that's not the answer -- You still have to install systemd
> > first and convert.
> 
> This is probably because Debian is commited to systemd and going that way
> further and further. The fact that today it is possible to convert to sysvinit
> is because systemd migration is still in transient. Upcoming release or the
> next one possibly won't even allow switching to sysvinit easily (or at all).

I think that will depend on whether there are enough people willing
to put in the necessary work to maintain two inits or not.

Cheers
- -- tomás
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)

iEYEARECAAYFAlrxJeIACgkQBcgs9XrR2kZM1ACcDEG2teh4BTbMJfnft7RtoaHX
W9cAn11008ZYHaPbOWO6tywtL+JlNfR3
=VIa1
-----END PGP SIGNATURE-----

[toc] | [prev] | [next] | [standalone]

#195481

[Multipart message — attachments visible in raw view] — view raw

On Tue, 8 May 2018, tomas@tuxteam.de wrote:
> On Tue, May 08, 2018 at 02:23:57AM +0300, Abdullah Ramazanoğlu wrote:
>>>> Is anyone working on a mechanism to allow for install-time selection
>>>> of a desired init?  I brought this up a few times since systemd came
>>>> to Debian, but I've never heard anything more on this.
>>>
>>> I asked that question when Jessie debuted.  No one is.  Certainly not
>>> Debian. The only suggestion I got was to build my own preseeded install
>>> disk, but that's not the answer -- You still have to install systemd
>>> first and convert.
>>
>> This is probably because Debian is commited to systemd and going that way
>> further and further. The fact that today it is possible to convert to sysvinit
>> is because systemd migration is still in transient. Upcoming release or the
>> next one possibly won't even allow switching to sysvinit easily (or at all).
>
> I think that will depend on whether there are enough people willing
> to put in the necessary work to maintain two inits or not.

I'm willing to put in the time and effort.


-- 
David Griffith
dave@661.org

A: Because it fouls the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?

[toc] | [prev] | [next] | [standalone]

#195484

On Tue, 8 May 2018 06:21:54 +0200 tomas@tuxteam.de said:
> On Tue, May 08, 2018 at 02:23:57AM +0300, Abdullah Ramazanoğlu wrote:

--8<--

> > This is probably because Debian is commited to systemd and going that way
> > further and further. The fact that today it is possible to convert to
> > sysvinit is because systemd migration is still in transient. Upcoming
> > release or the next one possibly won't even allow switching to sysvinit
> > easily (or at all).  
> 
> I think that will depend on whether there are enough people willing
> to put in the necessary work to maintain two inits or not.

Quite right, but I would also add the user preferences factor to it.

This really boils down to natural selection algorithm that dominates the OSS
world: Components compete for adoption by peers and distros, which in turn
compete for the contributor and user base (i.e. us).

Regards
-- 
Abdullah Ramazanoğlu

[toc] | [prev] | [standalone]

Back to top | Article view | linux.debian.user

csiph-web