[gentoo-dev] /var/run or /run for init scripts?

[gentoo-dev] /var/run or /run for init scripts?

[Markos Chandras]

Hi all,

I see a lot of inconsistencies in the init scripts. Some of them are
using /var/run, others use /run. I just checked my box, and there is
no /var/run although some init scripts are trying to use it. Which one
is the correct one and why do we have both? The following guide only
mentions /var/run

http://www.gentoo.org/doc/en/handbook/handbook-x86.xml?part=2&chap=4

-- 
Regards,
Markos Chandras - Gentoo Linux Developer
http://dev.gentoo.org/~hwoarang

Re: [gentoo-dev] /var/run or /run for init scripts?

[Diego Elio Pettenò]

On 06/02/2013 14:52, Markos Chandras wrote:
> I see a lot of inconsistencies in the init scripts. Some of them are
> using /var/run, others use /run. I just checked my box, and there is
> no /var/run although some init scripts are trying to use it. Which one
> is the correct one and why do we have both? The following guide only
> mentions /var/run

We should be migrating to /run at this point.

http://goo.gl/hrWcN

-- 
Diego Elio Pettenò — Flameeyes
flameeyes@flameeyes.eu — http://blog.flameeyes.eu/

Re: [gentoo-dev] /var/run or /run for init scripts?

[Ian Stakenvicius]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

On 06/02/13 08:55 AM, Diego Elio Pettenò wrote:
> On 06/02/2013 14:52, Markos Chandras wrote:
>> I see a lot of inconsistencies in the init scripts. Some of them
>> are using /var/run, others use /run. I just checked my box, and
>> there is no /var/run although some init scripts are trying to use
>> it. Which one is the correct one and why do we have both? The
>> following guide only mentions /var/run
> 
> We should be migrating to /run at this point.
> 
> http://goo.gl/hrWcN
> 

I had thought /var/run was supposed to be compatibility a symlink to
/run now ?  Is that the case or should /var/run be eliminated from
systems?

(I realize this is a tangent to the original post)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.19 (GNU/Linux)

iF4EAREIAAYFAlESYUIACgkQ2ugaI38ACPBUyQEArLfQfeUnW510aizYrjlHcGRl
fjHoZOnHdC10O2Sam6UBALOedX+jKZb5wig4xI/l6lEtFOUVOLNcPU69kk3cpEsU
=aueY
-----END PGP SIGNATURE-----

Re: [gentoo-dev] /var/run or /run for init scripts?

[Markos Chandras]

On 6 February 2013 13:55, Diego Elio Pettenò <flameeyes@flameeyes.eu> wrote:
> On 06/02/2013 14:52, Markos Chandras wrote:
>> I see a lot of inconsistencies in the init scripts. Some of them are
>> using /var/run, others use /run. I just checked my box, and there is
>> no /var/run although some init scripts are trying to use it. Which one
>> is the correct one and why do we have both? The following guide only
>> mentions /var/run
>
> We should be migrating to /run at this point.
>
> http://goo.gl/hrWcN
>
> --
> Diego Elio Pettenò — Flameeyes
> flameeyes@flameeyes.eu — http://blog.flameeyes.eu/
>

Thanks. Would it made sense to symlink /var/run -> /run so we don't
end up with stable entries in /var/run directory? Some of my init
scripts appear to reported as "crashed" whereas the process is running
just fine. I suspect this is because a stale entry is in /var/run
directory (or /run I am not sure)

-- 
Regards,
Markos Chandras - Gentoo Linux Developer
http://dev.gentoo.org/~hwoarang

Re: [gentoo-dev] /var/run or /run for init scripts?

[Diego Elio Pettenò]

On 06/02/2013 14:58, Markos Chandras wrote:
> Thanks. Would it made sense to symlink /var/run -> /run so we don't
> end up with stable entries in /var/run directory? Some of my init
> scripts appear to reported as "crashed" whereas the process is running
> just fine. I suspect this is because a stale entry is in /var/run
> directory (or /run I am not sure)

I would say that we should have that symlink, and I told WilliamH so
before. I think he was going to ask Mike (vapier) about adding the
symlink to baselayout itself, so that it doesn't get reaped away.

I agree that the symlink should stay there for compatibility at least,
which should also answer Ian's question.

-- 
Diego Elio Pettenò — Flameeyes
flameeyes@flameeyes.eu — http://blog.flameeyes.eu/

[gentoo-dev] The /run migration

[Ian Stakenvicius]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

On 06/02/13 09:02 AM, Diego Elio Pettenò wrote:
> On 06/02/2013 14:58, Markos Chandras wrote:
>> Thanks. Would it made sense to symlink /var/run -> /run so we
>> don't end up with stable entries in /var/run directory? Some of
>> my init scripts appear to reported as "crashed" whereas the
>> process is running just fine. I suspect this is because a stale
>> entry is in /var/run directory (or /run I am not sure)
> 
> I would say that we should have that symlink, and I told WilliamH
> so before. I think he was going to ask Mike (vapier) about adding
> the symlink to baselayout itself, so that it doesn't get reaped
> away.
> 
> I agree that the symlink should stay there for compatibility at
> least, which should also answer Ian's question.
> 

OK - so I've noticed some issues with the way the /run migration has
gone down; it seems that a lot of systems do not have a consistent
migration and as I don't remember (and my initial look couldn't find)
the details for its implementation, I'm hoping someone can chime in
and (A) describe the process as it is supposed to work, and (B) point
out where it's implemented so that this implementation can be adjusted
(or how it's been adjusted can be reviewed) to get everyone's system
to a consistent state.

Normally i'd just ask WilliamH on irc, but since this is more of a
forensic discussion i thought it better to do it here..

So, *my* systems do have /var/run -> /run , which means at some point
the /run migration did happen and compatibility symlinks were created.
 If hwoarang's systems don't have this, there must be an issue somewhere.

Now, with /run migration itself -- WilliamH and I discussed this issue
where /run (when the initial change was to be made if upgrading to
openrc-0.11.x) would always have a symlink to /lib/something/openrc/
in the directory itself (that is, underneath the tmpfs mount).
However when I investigated, I discovered that two of my systems had
actual dead directories and temp files in /run (and no symlink).  Code
has been added to /etc/init.d/bootmisc to clean out anything in /run
underneath the tmpfs, but it might be pertinent to figure out why this
happened in the first place, as it might be related to why the
/var/run symlink might not have been created (and relate to other
inconsistencies we haven't found yet).


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.19 (GNU/Linux)

iF0EAREIAAYFAlESZhsACgkQ2ugaI38ACPC9UgD/c0orcDKEHWpuneSI9ZZ3ZuYo
QwQZipdv/bryXly1Ek8A+KrameA53aeF4srZywxS+vjCucIkcS3Rm9x3Yq0O700=
=x0lk
-----END PGP SIGNATURE-----

Re: [gentoo-dev] The /run migration

[Ian Stakenvicius]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

On 06/02/13 09:18 AM, Ian Stakenvicius wrote:
> On 06/02/13 09:02 AM, Diego Elio Pettenò wrote:
>> On 06/02/2013 14:58, Markos Chandras wrote:
>>> Thanks. Would it made sense to symlink /var/run -> /run so we 
>>> don't end up with stable entries in /var/run directory? Some
>>> of my init scripts appear to reported as "crashed" whereas the 
>>> process is running just fine. I suspect this is because a
>>> stale entry is in /var/run directory (or /run I am not sure)
> 
>> I would say that we should have that symlink, and I told
>> WilliamH so before. I think he was going to ask Mike (vapier)
>> about adding the symlink to baselayout itself, so that it doesn't
>> get reaped away.
> 
>> I agree that the symlink should stay there for compatibility at 
>> least, which should also answer Ian's question.
> 
> 
> OK - so I've noticed some issues with the way the /run migration
> has gone down; it seems that a lot of systems do not have a
> consistent migration and as I don't remember (and my initial look
> couldn't find) the details for its implementation, I'm hoping
> someone can chime in and (A) describe the process as it is supposed
> to work, and (B) point out where it's implemented so that this
> implementation can be adjusted (or how it's been adjusted can be
> reviewed) to get everyone's system to a consistent state.
> 
> Normally i'd just ask WilliamH on irc, but since this is more of a 
> forensic discussion i thought it better to do it here..
> 
> So, *my* systems do have /var/run -> /run , which means at some
> point the /run migration did happen and compatibility symlinks were
> created. If hwoarang's systems don't have this, there must be an
> issue somewhere.
> 
> Now, with /run migration itself -- WilliamH and I discussed this
> issue where /run (when the initial change was to be made if
> upgrading to openrc-0.11.x) would always have a symlink to
> /lib/something/openrc/ in the directory itself (that is, underneath
> the tmpfs mount). However when I investigated, I discovered that
> two of my systems had actual dead directories and temp files in
> /run (and no symlink).  Code has been added to /etc/init.d/bootmisc
> to clean out anything in /run underneath the tmpfs, but it might be
> pertinent to figure out why this happened in the first place, as it
> might be related to why the /var/run symlink might not have been
> created (and relate to other inconsistencies we haven't found
> yet).
> 


As a follow up -- there are probably two cases to look at; #1 are
stable-only users (so just what happens when the stable bumps
occurred), and #2 are ~arch users and/or ~arch keyworders.  I'm going
to guess that the majority of inconsistencies with the planned results
will come from #2 as the different implementations of /run migration
would have taken place differently with each ~arch bump depending on
who installed what when..


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.19 (GNU/Linux)

iF4EAREIAAYFAlESZy8ACgkQ2ugaI38ACPDtpAD+N8OuOAL44dwCL2EE72IzGuxX
Bmz9YFagyEqauI2da5UA/RIAfpPDdd0of0LDxH1T/9C5stnbovzlRqnQCKi66e+1
=SWm7
-----END PGP SIGNATURE-----

Re: [gentoo-dev] The /run migration

[Markos Chandras]

On 6 February 2013 14:18, Ian Stakenvicius <axs@gentoo.org> wrote:
>
> So, *my* systems do have /var/run -> /run , which means at some point
> the /run migration did happen and compatibility symlinks were created.
>  If hwoarang's systems don't have this, there must be an issue somewhere.
>

My system is a brand new ~testing installation with a
stage3-amd64-20130110.tar.bz2. I am not sure who is responsible for
creating this symlink. I see the symlink is present on that stage3
tarball so somehow it must have been removed from my system. Even if
it
was a user error, then shouldn't there be a mechanism of recreating it
on every boot if it's gone missing? At least until all init scripts
migrate to
/run.

-- 
Regards,
Markos Chandras - Gentoo Linux Developer
http://dev.gentoo.org/~hwoarang

Re: [gentoo-dev] The /run migration

[Ian Stakenvicius]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

On 06/02/13 09:53 AM, Markos Chandras wrote:
> On 6 February 2013 14:18, Ian Stakenvicius <axs@gentoo.org> wrote:
>> 
>> So, *my* systems do have /var/run -> /run , which means at some
>> point the /run migration did happen and compatibility symlinks
>> were created. If hwoarang's systems don't have this, there must
>> be an issue somewhere.
>> 
> 
> My system is a brand new ~testing installation with a 
> stage3-amd64-20130110.tar.bz2. I am not sure who is responsible
> for creating this symlink. I see the symlink is present on that
> stage3 tarball so somehow it must have been removed from my system.
> Even if it was a user error, then shouldn't there be a mechanism of
> recreating it on every boot if it's gone missing? At least until
> all init scripts migrate to /run.
> 

..there was a discussion a week or two back about portage cleaning up
symlinks, or something that needs to be done to keep portage warning
about symlinks, or something.  Anyways, I'm wondering if a change was
made related to that and for whatever reason portage is now cleaning
/var/run


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.19 (GNU/Linux)

iF4EAREIAAYFAlEScMgACgkQ2ugaI38ACPAl7wEAj/n6Euiq/8gNn2tb8LjdJB7E
07yk78xCMZJudAHI/NEA/jHR5BoQIHZu2Tm5PRBN3BiK3Fe1miak3Z4UGVuSRudx
=j+bI
-----END PGP SIGNATURE-----

Re: [gentoo-dev] The /run migration

[Mike Gilbert]

On Wed, Feb 6, 2013 at 10:03 AM, Ian Stakenvicius <axs@gentoo.org> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA256
>
> On 06/02/13 09:53 AM, Markos Chandras wrote:
>> On 6 February 2013 14:18, Ian Stakenvicius <axs@gentoo.org> wrote:
>>>
>>> So, *my* systems do have /var/run -> /run , which means at some
>>> point the /run migration did happen and compatibility symlinks
>>> were created. If hwoarang's systems don't have this, there must
>>> be an issue somewhere.
>>>
>>
>> My system is a brand new ~testing installation with a
>> stage3-amd64-20130110.tar.bz2. I am not sure who is responsible
>> for creating this symlink. I see the symlink is present on that
>> stage3 tarball so somehow it must have been removed from my system.
>> Even if it was a user error, then shouldn't there be a mechanism of
>> recreating it on every boot if it's gone missing? At least until
>> all init scripts migrate to /run.
>>
>
> ..there was a discussion a week or two back about portage cleaning up
> symlinks, or something that needs to be done to keep portage warning
> about symlinks, or something.  Anyways, I'm wondering if a change was
> made related to that and for whatever reason portage is now cleaning
> /var/run
>

Portage will "cleanup" the /var/run symlink after unmerging the last
package that installed files under /var/run.

I think an early init script (bootmisc?) needs to create the /var/run
symlink if it is missing.

Re: [gentoo-dev] The /run migration

[William Hubbs]

[-- Attachment #1: Type: text/plain, Size: 1725 bytes --]

On Wed, Feb 06, 2013 at 10:23:06AM -0500, Mike Gilbert wrote:
> On Wed, Feb 6, 2013 at 10:03 AM, Ian Stakenvicius <axs@gentoo.org> wrote:
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA256
> >
> > On 06/02/13 09:53 AM, Markos Chandras wrote:
> >> On 6 February 2013 14:18, Ian Stakenvicius <axs@gentoo.org> wrote:
> >>>
> >>> So, *my* systems do have /var/run -> /run , which means at some
> >>> point the /run migration did happen and compatibility symlinks
> >>> were created. If hwoarang's systems don't have this, there must
> >>> be an issue somewhere.
> >>>
> >>
> >> My system is a brand new ~testing installation with a
> >> stage3-amd64-20130110.tar.bz2. I am not sure who is responsible
> >> for creating this symlink. I see the symlink is present on that
> >> stage3 tarball so somehow it must have been removed from my system.
> >> Even if it was a user error, then shouldn't there be a mechanism of
> >> recreating it on every boot if it's gone missing? At least until
> >> all init scripts migrate to /run.
> >>
> >
> > ..there was a discussion a week or two back about portage cleaning up
> > symlinks, or something that needs to be done to keep portage warning
> > about symlinks, or something.  Anyways, I'm wondering if a change was
> > made related to that and for whatever reason portage is now cleaning
> > /var/run
> >
> 
> Portage will "cleanup" the /var/run symlink after unmerging the last
> package that installed files under /var/run.
> 
> I think an early init script (bootmisc?) needs to create the /var/run
> symlink if it is missing.

The only problem with this approach is it doesn't solve the issue for
people who are not using OpenRc.

William


[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-dev] The /run migration

[Cyprien Nicolas]

[-- Attachment #1: Type: text/plain, Size: 388 bytes --]

On Wed, Feb 06, 2013 at 10:23:06AM -0500, Mike Gilbert wrote:
> Portage will "cleanup" the /var/run symlink after unmerging the last
> package that installed files under /var/run.

And if you install back that package, portage will create the /var/run
directory, and not the symlink. openssh-6.1_p1 has: keepdir /var/run

-- 
Cyprien Nicolas (Fulax)
Gentoo Lisp project contrib

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 490 bytes --]

Re: [gentoo-dev] /var/run or /run for init scripts?

[Philip Webb]

130206 Diego Elio Pettenò wrote:
> On 06/02/2013 14:58, Markos Chandras wrote:
>> Would it made sense to symlink /var/run -> /run
>> so we don't end up with stable entries in /var/run directory?
> I would say that we should have that symlink
> and I told WilliamH so before.

FWIW I have :

  root:505 ~> ls -l /var/run
    lrwxrwxrwx 1 root root 4 Jun 20  2012 /var/run -> /run
  root:506 ~> equery b /var/run
    * Searching for /var/run ...
    sys-apps/dbus-1.6.8 (/var/run)

-- 
========================,,============================================
SUPPORT     ___________//___,   Philip Webb
ELECTRIC   /] [] [] [] [] []|   Cities Centre, University of Toronto
TRANSIT    `-O----------O---'   purslowatchassdotutorontodotca

Re: [gentoo-dev] The /run migration

[Stefan Ehret]

On Wed, 2013-02-06 at 10:23 -0500, Mike Gilbert wrote:
> On Wed, Feb 6, 2013 at 10:03 AM, Ian Stakenvicius <axs@gentoo.org> wrote:
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA256
> >
> > On 06/02/13 09:53 AM, Markos Chandras wrote:
> >> On 6 February 2013 14:18, Ian Stakenvicius <axs@gentoo.org> wrote:
> >>>
> >>> So, *my* systems do have /var/run -> /run , which means at some
> >>> point the /run migration did happen and compatibility symlinks
> >>> were created. If hwoarang's systems don't have this, there must
> >>> be an issue somewhere.
> >>>
> >>
> >> My system is a brand new ~testing installation with a
> >> stage3-amd64-20130110.tar.bz2. I am not sure who is responsible
> >> for creating this symlink. I see the symlink is present on that
> >> stage3 tarball so somehow it must have been removed from my system.
> >> Even if it was a user error, then shouldn't there be a mechanism of
> >> recreating it on every boot if it's gone missing? At least until
> >> all init scripts migrate to /run.
> >>
> >
> > ..there was a discussion a week or two back about portage cleaning up
> > symlinks, or something that needs to be done to keep portage warning
> > about symlinks, or something.  Anyways, I'm wondering if a change was
> > made related to that and for whatever reason portage is now cleaning
> > /var/run
> >
> 
> Portage will "cleanup" the /var/run symlink after unmerging the last
> package that installed files under /var/run.
> 
> I think an early init script (bootmisc?) needs to create the /var/run
> symlink if it is missing.
> 

My disision by stage3-tarball is to wait a few weeks and will read the
gentoo mails.

Best regards

Stefan Ehret

Re: [gentoo-dev] The /run migration

[Markos Chandras]

On 6 February 2013 15:30, William Hubbs <williamh@gentoo.org> wrote:
> On Wed, Feb 06, 2013 at 10:23:06AM -0500, Mike Gilbert wrote:
>> On Wed, Feb 6, 2013 at 10:03 AM, Ian Stakenvicius <axs@gentoo.org> wrote:
>> > -----BEGIN PGP SIGNED MESSAGE-----
>> > Hash: SHA256
>> >
>> > On 06/02/13 09:53 AM, Markos Chandras wrote:
>> >> On 6 February 2013 14:18, Ian Stakenvicius <axs@gentoo.org> wrote:
>> >>>
>> >>> So, *my* systems do have /var/run -> /run , which means at some
>> >>> point the /run migration did happen and compatibility symlinks
>> >>> were created. If hwoarang's systems don't have this, there must
>> >>> be an issue somewhere.
>> >>>
>> >>
>> >> My system is a brand new ~testing installation with a
>> >> stage3-amd64-20130110.tar.bz2. I am not sure who is responsible
>> >> for creating this symlink. I see the symlink is present on that
>> >> stage3 tarball so somehow it must have been removed from my system.
>> >> Even if it was a user error, then shouldn't there be a mechanism of
>> >> recreating it on every boot if it's gone missing? At least until
>> >> all init scripts migrate to /run.
>> >>
>> >
>> > ..there was a discussion a week or two back about portage cleaning up
>> > symlinks, or something that needs to be done to keep portage warning
>> > about symlinks, or something.  Anyways, I'm wondering if a change was
>> > made related to that and for whatever reason portage is now cleaning
>> > /var/run
>> >
>>
>> Portage will "cleanup" the /var/run symlink after unmerging the last
>> package that installed files under /var/run.
>>
>> I think an early init script (bootmisc?) needs to create the /var/run
>> symlink if it is missing.
>
> The only problem with this approach is it doesn't solve the issue for
> people who are not using OpenRc.
>
> William
>

You mean the issue for people not having /var/run->/run and not
running openrc? Do all rc systems expect a /var/run->/run? If so, let
other rc system developers to fix it properly in their code. Since we
do have the ability to fix it in openrc I would say please do it just
to make sure systems will remain functional until all init scripts are
properly fixed.

-- 
Regards,
Markos Chandras - Gentoo Linux Developer
http://dev.gentoo.org/~hwoarang