[gentoo-user] Custom ebuilds for CoreOS

[gentoo-user] Custom ebuilds for CoreOS

[James]

Anyone know anything about coreos?

Lookie lookie, they have "ebuilds"? 

python-oem-2.7.6-r1.ebuild [1] 


It clams to be 100% open source. It runs on "bare metal", linux systems,
clusters and clouds. It claims to have a much small footprint ~114 MB and
boots very very fast via pxi(boot).

Very interesting....

It does look  commercial too?:
https://coreos.com/


I guess my take is that eventually, linux will be very small, embedded
and a cluster/cloud environment is where most systems will plug in,
kinda like most modern cell phones. Hopefully, there'll be a systemd centric
version so that enables individuals and small companies can remain "in the
game".

Surely there will be a openrc version(s) that survives, adapts and remains
relevant.

To me, it appears that some forward looking folks have forked (stolen the
best parts?) gentoo, made some fundamental (long overdue changes) and are
all about creating a source_to_cluster platform.  (hmmmm, vaguely sounds
familiar...scratching head). It is a natural evilution for linux to take; or
are we going to embrace some much needed change (new ideas) into gentoo?


James


[1] https://github.com/coreos/coreos-overlay/tree/master/dev-lang/python-oem

[2] https://github.com/coreos/coreos-overlay/blob/master/eclass/git.eclass

https://github.com/coreos

https://coreos.com/products/

https://coreos.com/blog/rocket/

https://coreos.com/docs/

Re: [gentoo-user] Custom ebuilds for CoreOS

[Rich Freeman]

On Mon, Dec 1, 2014 at 3:46 PM, James <wireless@tampabay.rr.com> wrote:
>
> It clams to be 100% open source. It runs on "bare metal", linux systems,
> clusters and clouds. It claims to have a much small footprint ~114 MB and
> boots very very fast via pxi(boot).

The whole idea of CoreOS is to be the host for a bunch of containers.
The host is completely generic - other than maybe configuring things
like the network or hardware or things actually related to hosting
(what containers to run/how/etc) you aren't suppose to really touch
it.  You don't install packages on the host.  All the stuff you care
about goes into the containers.

Think of it like VMWare on bare metal, except it is linux and you're
running containers and not VMs (so much more efficient, and less
secure).

>
> Surely there will be a openrc version(s) that survives, adapts and remains
> relevant.

Again, the point of CoreOS is that you don't care how the host works.
You won't add/remove services from the host.  As such you won't care
what init implementation it runs.

The containers are a completely different beast.  You might just run
your application in the container as PID 1.  Or, maybe you run
something like sysvinit+openrc or systemd inside a container.  You
could have one of each running on the same host.

>
> To me, it appears that some forward looking folks have forked (stolen the
> best parts?) gentoo, made some fundamental (long overdue changes) and are
> all about creating a source_to_cluster platform.  (hmmmm, vaguely sounds
> familiar...scratching head). It is a natural evilution for linux to take; or
> are we going to embrace some much needed change (new ideas) into gentoo?

I have no idea if CoreOS is Gentoo-derived, but it is very much a
special-purpose distro.  The whole concept is that you put all the
value-add in the containers, and then you just want a really standard
and lightweight distro to host your containers in.  Maybe you run
CentOS in one container, and Gentoo in another container, and Debian
in another container.

--
Rich

Re: [gentoo-user] Custom ebuilds for CoreOS

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 617 bytes --]

On Mon, 1 Dec 2014 20:46:54 +0000 (UTC), James wrote:

> I guess my take is that eventually, linux will be very small, embedded
> and a cluster/cloud environment is where most systems will plug in,
> kinda like most modern cell phones. Hopefully, there'll be a systemd
> centric version so that enables individuals and small companies can
> remain "in the game".

Given that CoreOS have sponsored some systemd development
(systemd-networkd), I think it is reasonable to assume they plan to stick
with systemd for the foreseeable future.


-- 
Neil Bothwick

Top Oxymorons Number 11: Terribly pleased

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 181 bytes --]

[gentoo-user] Re: Custom ebuilds for CoreOS

[James]

Rich Freeman <rich0 <at> gentoo.org> writes:


> > To me, it appears that some forward looking folks have forked 
> > (stolen the best parts?) gentoo, made some fundamental 
> > (long overdue changes) and are > > all about creating a 
> > source_to_cluster platform.  (hmmmm, vaguely sounds 
> > familiar...scratching head). It is a natural evilution for linux to 
> > take; or are we going to embrace some much needed change 
> > (new ideas) into gentoo?



> I have no idea if CoreOS is Gentoo-derived, but it is very much a
> special-purpose distro.  The whole concept is that you put all the
> value-add in the containers, and then you just want a really standard
> and lightweight distro to host your containers in.  Maybe you run
> CentOS in one container, and Gentoo in another container, and Debian
> in another container.

Your  first points are understood; and centos appear to be focused on the
commercial "cloud" mentality of don't buy hareware, rent containers
from us crowd.  That, to me, is a fool's path.

What I'm hoping for is that with the (gentoo) past of revolving devs,
Hasufell ideas for distributed development by reducing the gentoo core;
Flameyes takedown of tinderbox, my pursuit of clustering and many other
issues (pid1) all seem to inidcate that many distros are fundamentally
examining their path(s) forward. So, I think gentoo can have a minimize
version that achieves what CoreOS is doing, but it is gentoo-bare-metal
centric.  I think Gentoo can robustly support systemd and openrc, containers
and other key areas and new technologies, in a fundamentally 
unique way.


I do think a fundamental "update" to the entire gentoo environment is a
healthy ares for discussion. I do appreciate your insights on coreOS. I see
it as a minimized embedded effort to bring resources into a cluster that is
exclusively controlled by the owner. I have "zero" interest in the "cloud"
as beside being a very dumb idea for too many reasons to innumerate, it
removes folks from gaining knowledge of direct hardware experiences.

I do love the way the "cloud" vendors find and collect up the very best
ideas. I hate how the "cloud" vendors want to offer those best ideas,
as a transient benefit via time-rented binaries.

I strongly believe we are at a nexus (a vergence in the force) as many
new technologies are converging very rapidly. Call it what you like,
but, we are at the crossroads of some very unique opportunites, imho.
If we had a gentoo cluster right now, something like tinderbox would
have been running there all along. YMMV.


James

Re: [gentoo-user] Custom ebuilds for CoreOS

[Canek Peláez Valdés]

On Mon, Dec 1, 2014 at 3:56 PM, Neil Bothwick <neil@digimed.co.uk> wrote:
> On Mon, 1 Dec 2014 20:46:54 +0000 (UTC), James wrote:
>
>> I guess my take is that eventually, linux will be very small, embedded
>> and a cluster/cloud environment is where most systems will plug in,
>> kinda like most modern cell phones. Hopefully, there'll be a systemd
>> centric version so that enables individuals and small companies can
>> remain "in the game".
>
> Given that CoreOS have sponsored some systemd development
> (systemd-networkd), I think it is reasonable to assume they plan to stick
> with systemd for the foreseeable future.

More importantly, CoreOS uses systemd to monitor/control the instances
inside containers like systemd-nspawn does, only in a more general and
powerful way.

I don't think you can currently run the CoreOS host with anything
other than systemd, and to make it so it would be a lot of work. From
[2]:

"""
Within the CoreOS world, you will almost exclusively use systemd to
manage the lifecycle of your Docker containers.
"""

Regards.

[2] https://coreos.com/docs/launching-containers/launching/getting-started-with-systemd/
-- 
Canek Peláez Valdés
Profesor de asignatura, Facultad de Ciencias
Universidad Nacional Autónoma de México

Re: [gentoo-user] Custom ebuilds for CoreOS

[Andreas K. Huettel]

Am Montag 01 Dezember 2014, 20:46:54 schrieb James:
> Anyone know anything about coreos?
> 
> Lookie lookie, they have "ebuilds"?
> 

According to wikipedia, CoreOS is a fork of ChromeOS [1].

ChromeOS is most definitely a Gentoo derivative [2,3,4], even though that fact 
is not really well known (and not really publicised).

[1] https://en.wikipedia.org/wiki/CoreOS
[2] https://de.wikipedia.org/wiki/Google_Chrome_OS, see infobox
[3] http://www.chromium.org/chromium-os
[4] https://chromium.googlesource.com/chromiumos/overlays/portage/

-- 
Andreas K. Huettel
Gentoo Linux developer
kde, council

Re: [gentoo-user] Custom ebuilds for CoreOS

[Rich Freeman]

On Tue, Dec 2, 2014 at 7:36 AM, Andreas K. Huettel <dilfridge@gentoo.org> wrote:
> Am Montag 01 Dezember 2014, 20:46:54 schrieb James:
>> Anyone know anything about coreos?
>>
>> Lookie lookie, they have "ebuilds"?
>>
>
> According to wikipedia, CoreOS is a fork of ChromeOS [1].
>
> ChromeOS is most definitely a Gentoo derivative [2,3,4], even though that fact
> is not really well known (and not really publicised).

Interesting.  Talk about a march of init systems.  You have Gentoo
which defaults to openrc and supports systemd, to ChromeOS which only
supports upstart, to CoreOS which uses systemd.

In any case, the whole point of both ChromeOS and CoreOS is that
they're hosts for running applications completely outside of the usual
unix-y approach of sticking stuff in /usr.  Applications on ChromeOS
are Chrome extensions and the like, and applications on CoreOS are
containers.  The whole point of both is to abstract away all the guts
of how the OS operates, so the choice of init really shouldn't matter
much to anybody using either.  If you really want to stick stuff in
/usr and interact with host processes directly, then you really should
find a distro which isn't designed to be a black box in this regard.

--
Rich

[gentoo-user] Re: Custom ebuilds for CoreOS

[James]

Rich Freeman <rich0 <at> gentoo.org> writes:


> > ChromeOS is most definitely a Gentoo derivative [2,3,4], even though  
> > that fact is not really well known (and not really publicised).

Thanks for the links. I did not see that bit of history...


> Interesting.  Talk about a march of init systems.  You have Gentoo
> which defaults to openrc and supports systemd, to ChromeOS which only
> supports upstart, to CoreOS which uses systemd.

I have always maintained that init system, is only critical choices,
because the cloud/cluster technologies are not yet mature. What you
get from these init system, can easily be established with a few
files at boot time (PXE and many others methods), that is what CoreOS
is doing, just in an updated fashion.


> In any case, the whole point of both ChromeOS and CoreOS is that
> they're hosts for running applications completely outside of the usual
> unix-y approach of sticking stuff in /usr.  Applications on ChromeOS
> are Chrome extensions and the like, and applications on CoreOS are
> containers.  The whole point of both is to abstract away all the guts
> of how the OS operates, so the choice of init really shouldn't matter
> much to anybody using either.  If you really want to stick stuff in
> /usr and interact with host processes directly, then you really should
> find a distro which isn't designed to be a black box in this regard.

Yes, your are correct, that is what the "cloud" vendors are striving for.
I see a  much deeper future, that leverages their ideas to invigorate Gentoo.


In the good old days, it was very common for folks to build up minimized
gentoo sytems, by starting of with "-*" in the USE settings of make.conf.
Sure now that sort of thing is frowned upon by the devs, but it was and 
is a very valid method to minimize the size and complexity of a system.
I still have running gentoo sytems with just a few flags set and with "-*"
in the USE settings. I do not sync them, but update them selectively in a
manual process. This yeilds a linux system, usually for a special purpose,
that is so minimized it's very close to a stipped/optimized embedded system.


WE seem to have lost the "embedded" focus here at gentoo. Gentoo-embedded,
as a discussion/sharing group seems to be dead; but I think it is
because most are slobberingly working on 64 bit arm offerings. Minimized
sytems can be delivered via something like CoreOS and then the other
codes (binaries whatever) can be added dynamically to yield a very
focused target system, a replacement system, or a parallel system to
handle a dynamic resource loading problem, jus to name a few reason
for the CoreOS approach to building up a cluster.

I do not see the "/usr" types of systems (like a current gentoo workstation
or server) going away any time soon. What I hope WE can pull off at Gentoo
is integration of the best of the CoreOS ideas into "Gentoo proper". I was
just very surprised to stumble across CoreOS; as it is what my hopes
(vision?) of Gentoo are to be, beside continuing the traditional linux
progression (/var/usr/local/home/etc..... )type of unix derivative OS. I do
believe that this traditional linux (what's left of unix) belongs to the
masses and "force feeding" of systemd was a very, very poorly made decision.

I do see the "Cloud vendors" eating away at the Microsoft and Apple user
base and large companies with masses of "clerical" employees.

So, to sum this up, in my view, is to say that CoreOS (ideas) offers us a
pathway to be able to build (via dynamic downloads) any system we want from
a minimized linux state machine on a 16 bit core, to a "-*" minimized gentoo
system or a full gentoo linux workstataion or server. Substitute "cluster
controlled by user" for "cloud" and I really like the CoreOS vision. I think
they have stolen the "Gentoo Grand Unification Theory" from us, because we
are "napping" here at Gentoo.

Just so folk know, a minimize system is far easier to keep secure, and
replace dynamically for whatever the failure reason is. I guess that CoreOS
is just building up  clusters from derivatives ot TFTPboot...... That is
what's old (farts) is new again, as it appears we are returning full circle.



YiPeeeeeeeeeeeee!

James

Re: [gentoo-user] Re: Custom ebuilds for CoreOS

[Rich Freeman]

On Tue, Dec 2, 2014 at 10:35 AM, James <wireless@tampabay.rr.com> wrote:
>
> I do not see the "/usr" types of systems (like a current gentoo workstation
> or server) going away any time soon. What I hope WE can pull off at Gentoo
> is integration of the best of the CoreOS ideas into "Gentoo proper".

I'm not suggesting that "/usr types of systems" are going away.  I'm
just pointing out that they're not really the focus of CoreOS (hosting
them inside containers is, but not running these kinds of applications
in the host itself).

You seem to be wanting a minimalist profile of Gentoo, not CoreOS.  I
think many of us would love to see that, and I've been an advocate of
paring down @system for just this reason.  I just wouldn't use the
term "CoreOS" with that as this is going to lead to confusion.  CoreOS
is a specialized distro intended to host containers, no more, no less.
It isn't intended as a starting point for embedded projects or such.
Sure, maybe you could make it work, but sooner or later CoreOS will
make some change that will make you very unhappy because they aren't
making it for you.

But, again, I'm all for a more lightweight Gentoo profile that doesn't
bundle stuff like openssh, or even an init implementation (since we
have several to choose from now).

--
Rich

[gentoo-user] Re: Custom ebuilds for CoreOS

[James]

Rich Freeman <rich0 <at> gentoo.org> writes:



> > is integration of the best of the CoreOS ideas into "Gentoo proper".
> 
> I'm not suggesting that "/usr types of systems" are going away.  I'm
> just pointing out that they're not really the focus of CoreOS (hosting
> them inside containers is, but not running these kinds of applications
> in the host itself).

I do not intend to follow the CoreOS commercial path. It intend to mod
gentoo to achieve those attractive attributes back into my "gentoo proper".
tftp, pxe, dhcp, uefi and many other tools give us a path to
running the least (embedded) to the most (complex traditional server)
as an extension (compliment) to the cluster. So as was pointed out,
I'm merely "lifting" form CoreOS what they lifted from their predicessors;
no more no less. I see the gentoo admins being able to move hardrware
in and out of the cluster, dynamically and being able to run many
sorts of gentoo systems (embedded to fulls server) on a myriad of
hardware they own and control.


> You seem to be wanting a minimalist profile of Gentoo, not CoreOS.  

YES!, I want Gentoo to "CRUSH" CoreOS because we can and our goal is not
to deceptively move users to a "rent the binary" jail. OK?


> < think many of us would love to see that, and I've been an advocate of
> paring down  <at> system for just this reason.  I just wouldn't use the
> term "CoreOS" with that as this is going to lead to confusion.  CoreOS
> is a specialized distro intended to host containers, no more, no less.

OK, we see CoreOS differently. For me it was an Epiphany moment of
where I'm been trying to end up, with the aforementioned Gentoo twists.

> It isn't intended as a starting point for embedded projects or such.
> Sure, maybe you could make it work, but sooner or later CoreOS will
> make some change that will make you very unhappy because they aren't
> making it for you.

CoreOS will never be in my critical path. Large corporations will turn
computer scientist and hackers into WalMart type-employees. Conglomerates
are the enemy, imho. I fear Conglomerates much more than any group
of government idiots. ymmv.

(warning digression)
 Just look at the entire "net neutrality"
turf struggle. That sort of "corner the market" monopolistic behavior
would not be possible, if we had just maintained the "MAE" precedence
for network peering.  Obama had little choice; but, putting networks
under SS7 style telecom regulations is a deceptive and horrible idea.
Conglomerates lobby congress and get very bad ideas written into law.
All we needed is regulation to allow (force) all networks to peer with
other networks. The entire concept of "private peering" is horseshit
and it should be ended immediately. CoreOS and the "Cloud" lobbyist can
easily get regulations passed to put an end to this linux experiment, imho.
Differnt subject I know, but the tactics of conglomerates are always the
same. Roll up competition and eliminate it, oh all in the name of better
security and portecting our 1st amendment rights  and our conglomerates.
(sorry of the digression).



> But, again, I'm all for a more lightweight Gentoo profile that doesn't
> bundle stuff like openssh, or even an init implementation (since we
> have several to choose from now).

Funny, ssh is one of a few things I would put into  drastically reduce
@system. ymmv, unless you are going to add something like netconsole.c
back into the bundle.

I do not see my vision of the cluster (CoreOS insprired) to be limiting
to anyone at Gentoo. Not the embedded folks, not the mimalist, not
any init-camp, not the devs, hackers, or wannabees. And certainly
not the users. Is this a large undertaking? Certainly. Are the pieces
mostly already in existence, just scattered about and transversing time?
(methinks YES).


It all depends on how your vision works. Being older, I see a return to
massive diskless nodes being what CoreOS and the entire "Cloud Vendor"
conglomerates want. Conversely, I see those cheap microP now accompanied by
enormous amount of ram and SSD that is dirt cheap forming the building
blocks for the Gentoo cluster paradigm shift. I see Gentoo "smashing" that
"Cloud-vendor CoreOS" paradigm by provide what they offer and so much more
(full /usr systems) out of the same core codebase. I see Gentoo keeping the
rank and file computer scientists and hackers, gamefully employed.   I see
the CoreOS folks migrating computer scientists and hackers to the Walmart
model of underemployment at a few conglomerates.

Gentoo provides an excellent set of choices  and a very bright future for me
(cluster). Other can pick their own poison....


peace,
&& thanks

James

Re: [gentoo-user] Re: Custom ebuilds for CoreOS

[Mark David Dumlao]

[-- Attachment #1: Type: text/plain, Size: 5085 bytes --]

Why do I get the feeling that this is another episode of the "i hate
LennartSoft(tm) too" circlejerk on the gentoo mailing list?

this mailing list used to be about gentoo.
On Dec 3, 2014 1:38 AM, "James" <wireless@tampabay.rr.com> wrote:

> Rich Freeman <rich0 <at> gentoo.org> writes:
>
>
>
> > > is integration of the best of the CoreOS ideas into "Gentoo proper".
> >
> > I'm not suggesting that "/usr types of systems" are going away.  I'm
> > just pointing out that they're not really the focus of CoreOS (hosting
> > them inside containers is, but not running these kinds of applications
> > in the host itself).
>
> I do not intend to follow the CoreOS commercial path. It intend to mod
> gentoo to achieve those attractive attributes back into my "gentoo proper".
> tftp, pxe, dhcp, uefi and many other tools give us a path to
> running the least (embedded) to the most (complex traditional server)
> as an extension (compliment) to the cluster. So as was pointed out,
> I'm merely "lifting" form CoreOS what they lifted from their predicessors;
> no more no less. I see the gentoo admins being able to move hardrware
> in and out of the cluster, dynamically and being able to run many
> sorts of gentoo systems (embedded to fulls server) on a myriad of
> hardware they own and control.
>
>
> > You seem to be wanting a minimalist profile of Gentoo, not CoreOS.
>
> YES!, I want Gentoo to "CRUSH" CoreOS because we can and our goal is not
> to deceptively move users to a "rent the binary" jail. OK?
>
>
> > < think many of us would love to see that, and I've been an advocate of
> > paring down  <at> system for just this reason.  I just wouldn't use the
> > term "CoreOS" with that as this is going to lead to confusion.  CoreOS
> > is a specialized distro intended to host containers, no more, no less.
>
> OK, we see CoreOS differently. For me it was an Epiphany moment of
> where I'm been trying to end up, with the aforementioned Gentoo twists.
>
> > It isn't intended as a starting point for embedded projects or such.
> > Sure, maybe you could make it work, but sooner or later CoreOS will
> > make some change that will make you very unhappy because they aren't
> > making it for you.
>
> CoreOS will never be in my critical path. Large corporations will turn
> computer scientist and hackers into WalMart type-employees. Conglomerates
> are the enemy, imho. I fear Conglomerates much more than any group
> of government idiots. ymmv.
>
> (warning digression)
>  Just look at the entire "net neutrality"
> turf struggle. That sort of "corner the market" monopolistic behavior
> would not be possible, if we had just maintained the "MAE" precedence
> for network peering.  Obama had little choice; but, putting networks
> under SS7 style telecom regulations is a deceptive and horrible idea.
> Conglomerates lobby congress and get very bad ideas written into law.
> All we needed is regulation to allow (force) all networks to peer with
> other networks. The entire concept of "private peering" is horseshit
> and it should be ended immediately. CoreOS and the "Cloud" lobbyist can
> easily get regulations passed to put an end to this linux experiment, imho.
> Differnt subject I know, but the tactics of conglomerates are always the
> same. Roll up competition and eliminate it, oh all in the name of better
> security and portecting our 1st amendment rights  and our conglomerates.
> (sorry of the digression).
>
>
>
> > But, again, I'm all for a more lightweight Gentoo profile that doesn't
> > bundle stuff like openssh, or even an init implementation (since we
> > have several to choose from now).
>
> Funny, ssh is one of a few things I would put into  drastically reduce
> @system. ymmv, unless you are going to add something like netconsole.c
> back into the bundle.
>
> I do not see my vision of the cluster (CoreOS insprired) to be limiting
> to anyone at Gentoo. Not the embedded folks, not the mimalist, not
> any init-camp, not the devs, hackers, or wannabees. And certainly
> not the users. Is this a large undertaking? Certainly. Are the pieces
> mostly already in existence, just scattered about and transversing time?
> (methinks YES).
>
>
> It all depends on how your vision works. Being older, I see a return to
> massive diskless nodes being what CoreOS and the entire "Cloud Vendor"
> conglomerates want. Conversely, I see those cheap microP now accompanied by
> enormous amount of ram and SSD that is dirt cheap forming the building
> blocks for the Gentoo cluster paradigm shift. I see Gentoo "smashing" that
> "Cloud-vendor CoreOS" paradigm by provide what they offer and so much more
> (full /usr systems) out of the same core codebase. I see Gentoo keeping the
> rank and file computer scientists and hackers, gamefully employed.   I see
> the CoreOS folks migrating computer scientists and hackers to the Walmart
> model of underemployment at a few conglomerates.
>
> Gentoo provides an excellent set of choices  and a very bright future for
> me
> (cluster). Other can pick their own poison....
>
>
> peace,
> && thanks
>
> James
>
>
>
>

[-- Attachment #2: Type: text/html, Size: 6032 bytes --]

Re: [gentoo-user] Re: Custom ebuilds for CoreOS

[Rich Freeman]

On Tue, Dec 2, 2014 at 12:37 PM, James <wireless@tampabay.rr.com> wrote:
> Rich Freeman <rich0 <at> gentoo.org> writes:
>
>> You seem to be wanting a minimalist profile of Gentoo, not CoreOS.
>
> YES!, I want Gentoo to "CRUSH" CoreOS because we can and our goal is not
> to deceptively move users to a "rent the binary" jail. OK?
>

Gentoo and CoreOS really target different uses.  I certainly could see
one being installed more than the other just as there are no doubt
more tubes of toothpaste sold in a year than there are iPhones sold in
a year (or, at least I hope there are).  That doesn't mean that
toothpaste is "crushing" the iPhone.

This isn't unlike Gentoo vs ChromeOS.  You're comparing a
general-purpose distro (and one that is even more
general-purpose/customizable than a typical one) to a tool made to do
exactly one job well.

CoreOS is just about hosting containers.  Sure, some of those
containers might be "rent the binary jails" - but you could run Gentoo
in one of those containers just as easily.  CoreOS really competes
with the likes of VMWare/KVM, or even OpenStack.  If you don't want to
run a bazillion containers, then sure it isn't something you're going
to be interested in.

>
>> It isn't intended as a starting point for embedded projects or such.
>> Sure, maybe you could make it work, but sooner or later CoreOS will
>> make some change that will make you very unhappy because they aren't
>> making it for you.
>
> CoreOS will never be in my critical path. Large corporations will turn
> computer scientist and hackers into WalMart type-employees. Conglomerates
> are the enemy, imho. I fear Conglomerates much more than any group
> of government idiots. ymmv.

Well, then don't run it!  Large corporations are actually the
least-progressive when it comes to adopting these kinds of
technologies.  I actually see thing being embraced by mid-sized
companies first.  The "new way" of doing these things lets you quickly
scale up from development to production without a lot of manual
configuration of individual hosts.  I work for a big company and
they're still doing lots of manual installation scripts that get
signed and dated like it is still the 80s.  It isn't Walmart-type work
primarily because it is so error-prone we always need people to fix
all the stuff that breaks.  My LUG meets at a mid-sized VoIP company
that uses the likes of Puppet/Chef for everything and I'm sure Docker
is on their radar as something to think about next - they're hardly
robots but they realize that they'd rather have their bright employees
doing something other than dealing with botched updates on hosts that
bring down 47 VMs at a time.  Their customers like that they can just
pay them for a VoIP account and get full service for a low cost,
versus paying the kid next door to figure out how to custom-rig a PBX
for them.  And, yes, they use Asterisk.

--
Rich

Re: [gentoo-user] Custom ebuilds for CoreOS

[Saifi Khan]

On Mon, 1 Dec 2014, Neil Bothwick wrote:

> On Mon, 1 Dec 2014 20:46:54 +0000 (UTC), James wrote:
>
> Given that CoreOS have sponsored some systemd development
> (systemd-networkd), I think it is reasonable to assume they plan to stick
> with systemd for the foreseeable future.
>

CoreOS a gentoo derived distro, had made a very elegant use of 
'systemd'.

systemd consists of 'unit' and 'target'.

'unit' is config file containing 'docker run' command.

'target' is the grouping mechanism (equiv to fig.yml and 
upcoming 'docker group' command)

systemtd is exclusively used to manage the lifecycle of 'docker' 
containers !


thanks
Saifi.

Re: [gentoo-user] Custom ebuilds for CoreOS

[Saifi Khan]

On Tue, 2 Dec 2014, Andreas K. Huettel wrote:

> Am Montag 01 Dezember 2014, 20:46:54 schrieb James:
>> Anyone know anything about coreos?
>>
>> Lookie lookie, they have "ebuilds"?
>>
>
> According to wikipedia, CoreOS is a fork of ChromeOS [1].
>
> ChromeOS is most definitely a Gentoo derivative [2,3,4], even though that fact
> is not really well known (and not really publicised).
>

i suppose, CoreOS uses the 'update mechanism' from ChromeOS to 
provide autoupdate service (a/b).

for all practical purposes, CoreOS is a Gentoo derivative.


thanks
Saifi.

Re: [gentoo-user] Re: Custom ebuilds for CoreOS

[J. Roeleveld]

On Wednesday, December 03, 2014 02:39:53 AM Mark David Dumlao wrote:
> Why do I get the feeling that this is another episode of the "i hate
> LennartSoft(tm) too" circlejerk on the gentoo mailing list?

Why do I get the feeling you just want another flamewar?
I don't see any mention of systemd or anything else written by Lennart, apart 
from your comment.

> this mailing list used to be about gentoo.

It still is.

> On Dec 3, 2014 1:38 AM, "James" <wireless@tampabay.rr.com> wrote:
> > Rich Freeman <rich0 <at> gentoo.org> writes:
> > > > is integration of the best of the CoreOS ideas into "Gentoo proper".
> > > 
> > > I'm not suggesting that "/usr types of systems" are going away.  I'm
> > > just pointing out that they're not really the focus of CoreOS (hosting
> > > them inside containers is, but not running these kinds of applications
> > > in the host itself).
> > 
> > I do not intend to follow the CoreOS commercial path. It intend to mod
> > gentoo to achieve those attractive attributes back into my "gentoo
> > proper".
> > tftp, pxe, dhcp, uefi and many other tools give us a path to
> > running the least (embedded) to the most (complex traditional server)
> > as an extension (compliment) to the cluster. So as was pointed out,
> > I'm merely "lifting" form CoreOS what they lifted from their predicessors;
> > no more no less. I see the gentoo admins being able to move hardrware
> > in and out of the cluster, dynamically and being able to run many
> > sorts of gentoo systems (embedded to fulls server) on a myriad of
> > hardware they own and control.
> > 
> > > You seem to be wanting a minimalist profile of Gentoo, not CoreOS.
> > 
> > YES!, I want Gentoo to "CRUSH" CoreOS because we can and our goal is not
> > to deceptively move users to a "rent the binary" jail. OK?
> > 
> > > < think many of us would love to see that, and I've been an advocate of
> > > paring down  <at> system for just this reason.  I just wouldn't use the
> > > term "CoreOS" with that as this is going to lead to confusion.  CoreOS
> > > is a specialized distro intended to host containers, no more, no less.
> > 
> > OK, we see CoreOS differently. For me it was an Epiphany moment of
> > where I'm been trying to end up, with the aforementioned Gentoo twists.
> > 
> > > It isn't intended as a starting point for embedded projects or such.
> > > Sure, maybe you could make it work, but sooner or later CoreOS will
> > > make some change that will make you very unhappy because they aren't
> > > making it for you.
> > 
> > CoreOS will never be in my critical path. Large corporations will turn
> > computer scientist and hackers into WalMart type-employees. Conglomerates
> > are the enemy, imho. I fear Conglomerates much more than any group
> > of government idiots. ymmv.
> > 
> > (warning digression)
> > 
> >  Just look at the entire "net neutrality"
> > 
> > turf struggle. That sort of "corner the market" monopolistic behavior
> > would not be possible, if we had just maintained the "MAE" precedence
> > for network peering.  Obama had little choice; but, putting networks
> > under SS7 style telecom regulations is a deceptive and horrible idea.
> > Conglomerates lobby congress and get very bad ideas written into law.
> > All we needed is regulation to allow (force) all networks to peer with
> > other networks. The entire concept of "private peering" is horseshit
> > and it should be ended immediately. CoreOS and the "Cloud" lobbyist can
> > easily get regulations passed to put an end to this linux experiment,
> > imho.
> > Differnt subject I know, but the tactics of conglomerates are always the
> > same. Roll up competition and eliminate it, oh all in the name of better
> > security and portecting our 1st amendment rights  and our conglomerates.
> > (sorry of the digression).
> > 
> > > But, again, I'm all for a more lightweight Gentoo profile that doesn't
> > > bundle stuff like openssh, or even an init implementation (since we
> > > have several to choose from now).
> > 
> > Funny, ssh is one of a few things I would put into  drastically reduce
> > @system. ymmv, unless you are going to add something like netconsole.c
> > back into the bundle.
> > 
> > I do not see my vision of the cluster (CoreOS insprired) to be limiting
> > to anyone at Gentoo. Not the embedded folks, not the mimalist, not
> > any init-camp, not the devs, hackers, or wannabees. And certainly
> > not the users. Is this a large undertaking? Certainly. Are the pieces
> > mostly already in existence, just scattered about and transversing time?
> > (methinks YES).
> > 
> > 
> > It all depends on how your vision works. Being older, I see a return to
> > massive diskless nodes being what CoreOS and the entire "Cloud Vendor"
> > conglomerates want. Conversely, I see those cheap microP now accompanied
> > by
> > enormous amount of ram and SSD that is dirt cheap forming the building
> > blocks for the Gentoo cluster paradigm shift. I see Gentoo "smashing" that
> > "Cloud-vendor CoreOS" paradigm by provide what they offer and so much more
> > (full /usr systems) out of the same core codebase. I see Gentoo keeping
> > the
> > rank and file computer scientists and hackers, gamefully employed.   I see
> > the CoreOS folks migrating computer scientists and hackers to the Walmart
> > model of underemployment at a few conglomerates.
> > 
> > Gentoo provides an excellent set of choices  and a very bright future for
> > me
> > (cluster). Other can pick their own poison....
> > 
> > 
> > peace,
> > && thanks
> > 
> > James

Re: [gentoo-user] Re: Custom ebuilds for CoreOS

[Mark David Dumlao]

[-- Attachment #1: Type: text/plain, Size: 5803 bytes --]

Look up. the very first post contrastd coreos' systemd as opposed to
openrc, bringing words like "evil"ution into the park.

later on we hear that coreos is "stealing" gentoo's ideas and hope that it
is CRUSHED.

but why? its its own frigging distro now. not gentoo by a long shot.
On Wednesday, December 03, 2014 02:39:53 AM Mark David Dumlao wrote:
> Why do I get the feeling that this is another episode of the "i hate
> LennartSoft(tm) too" circlejerk on the gentoo mailing list?

Why do I get the feeling you just want another flamewar?
I don't see any mention of systemd or anything else written by Lennart,
apart
from your comment.

> this mailing list used to be about gentoo.

It still is.

> On Dec 3, 2014 1:38 AM, "James" <wireless@tampabay.rr.com> wrote:
> > Rich Freeman <rich0 <at> gentoo.org> writes:
> > > > is integration of the best of the CoreOS ideas into "Gentoo proper".
> > >
> > > I'm not suggesting that "/usr types of systems" are going away.  I'm
> > > just pointing out that they're not really the focus of CoreOS (hosting
> > > them inside containers is, but not running these kinds of applications
> > > in the host itself).
> >
> > I do not intend to follow the CoreOS commercial path. It intend to mod
> > gentoo to achieve those attractive attributes back into my "gentoo
> > proper".
> > tftp, pxe, dhcp, uefi and many other tools give us a path to
> > running the least (embedded) to the most (complex traditional server)
> > as an extension (compliment) to the cluster. So as was pointed out,
> > I'm merely "lifting" form CoreOS what they lifted from their
predicessors;
> > no more no less. I see the gentoo admins being able to move hardrware
> > in and out of the cluster, dynamically and being able to run many
> > sorts of gentoo systems (embedded to fulls server) on a myriad of
> > hardware they own and control.
> >
> > > You seem to be wanting a minimalist profile of Gentoo, not CoreOS.
> >
> > YES!, I want Gentoo to "CRUSH" CoreOS because we can and our goal is not
> > to deceptively move users to a "rent the binary" jail. OK?
> >
> > > < think many of us would love to see that, and I've been an advocate
of
> > > paring down  <at> system for just this reason.  I just wouldn't use
the
> > > term "CoreOS" with that as this is going to lead to confusion.  CoreOS
> > > is a specialized distro intended to host containers, no more, no less.
> >
> > OK, we see CoreOS differently. For me it was an Epiphany moment of
> > where I'm been trying to end up, with the aforementioned Gentoo twists.
> >
> > > It isn't intended as a starting point for embedded projects or such.
> > > Sure, maybe you could make it work, but sooner or later CoreOS will
> > > make some change that will make you very unhappy because they aren't
> > > making it for you.
> >
> > CoreOS will never be in my critical path. Large corporations will turn
> > computer scientist and hackers into WalMart type-employees.
Conglomerates
> > are the enemy, imho. I fear Conglomerates much more than any group
> > of government idiots. ymmv.
> >
> > (warning digression)
> >
> >  Just look at the entire "net neutrality"
> >
> > turf struggle. That sort of "corner the market" monopolistic behavior
> > would not be possible, if we had just maintained the "MAE" precedence
> > for network peering.  Obama had little choice; but, putting networks
> > under SS7 style telecom regulations is a deceptive and horrible idea.
> > Conglomerates lobby congress and get very bad ideas written into law.
> > All we needed is regulation to allow (force) all networks to peer with
> > other networks. The entire concept of "private peering" is horseshit
> > and it should be ended immediately. CoreOS and the "Cloud" lobbyist can
> > easily get regulations passed to put an end to this linux experiment,
> > imho.
> > Differnt subject I know, but the tactics of conglomerates are always the
> > same. Roll up competition and eliminate it, oh all in the name of better
> > security and portecting our 1st amendment rights  and our conglomerates.
> > (sorry of the digression).
> >
> > > But, again, I'm all for a more lightweight Gentoo profile that doesn't
> > > bundle stuff like openssh, or even an init implementation (since we
> > > have several to choose from now).
> >
> > Funny, ssh is one of a few things I would put into  drastically reduce
> > @system. ymmv, unless you are going to add something like netconsole.c
> > back into the bundle.
> >
> > I do not see my vision of the cluster (CoreOS insprired) to be limiting
> > to anyone at Gentoo. Not the embedded folks, not the mimalist, not
> > any init-camp, not the devs, hackers, or wannabees. And certainly
> > not the users. Is this a large undertaking? Certainly. Are the pieces
> > mostly already in existence, just scattered about and transversing time?
> > (methinks YES).
> >
> >
> > It all depends on how your vision works. Being older, I see a return to
> > massive diskless nodes being what CoreOS and the entire "Cloud Vendor"
> > conglomerates want. Conversely, I see those cheap microP now accompanied
> > by
> > enormous amount of ram and SSD that is dirt cheap forming the building
> > blocks for the Gentoo cluster paradigm shift. I see Gentoo "smashing"
that
> > "Cloud-vendor CoreOS" paradigm by provide what they offer and so much
more
> > (full /usr systems) out of the same core codebase. I see Gentoo keeping
> > the
> > rank and file computer scientists and hackers, gamefully employed.   I
see
> > the CoreOS folks migrating computer scientists and hackers to the
Walmart
> > model of underemployment at a few conglomerates.
> >
> > Gentoo provides an excellent set of choices  and a very bright future
for
> > me
> > (cluster). Other can pick their own poison....
> >
> >
> > peace,
> > && thanks
> >
> > James

[-- Attachment #2: Type: text/html, Size: 7610 bytes --]

Re: [gentoo-user] Re: Custom ebuilds for CoreOS

[Rich Freeman]

On Wed, Dec 3, 2014 at 9:28 AM, Mark David Dumlao <madumlao@gmail.com> wrote:
>
> but why? its its own frigging distro now. not gentoo by a long shot.
>

I think it is actually a compliment to the flexibility of Gentoo that
these derivatives are so different.  Gentoo is a somewhat-generic
linux distro overall - in its default install it isn't too different
from Debian/Ubuntu/Fedora/Arch on the surface and in terms of typical
package selection.  However, ChromeOS and CoreOS are very
non-traditional linux "distros."

When people ask me what Gentoo is "good for" I of course talk about
enthusiasts who care about both understanding their systems and having
a high degree of control, but I also talk about projects where you're
trying to blaze new trails and departing significantly from the
typical "linux desktop" or LAMP box.  If all you want is a stable LAMP
box then honestly you're probably better off with the likes of
Debian/CentOS/etc.  However, if you're doing something embedded, or
trying to change the world, then starting with Gentoo gives you a lot
more flexibility to blaze new ground while not having to build
EVERYTHING from scratch.

So, when people use Gentoo to do things that we personally don't find
useful, I think it is just a testimony to the fact that we've actually
accomplished one of our core missions: empowering our users to make
their own choices.

--
Rich

[gentoo-user] Re: Custom ebuilds for CoreOS

[James]

Mark David Dumlao <madumlao <at> gmail.com> writes:


> Look up. the very first post contrastd coreos' systemd as opposed 
> to openrc, bringing words like "evil"ution into the park.

That refers to the concept of conglomerates vs the people.
Systemd is only mentioned in passing. If it offends you, ignore it, OK?
I did not see any of the openrc camp chime in. Besides, as was pointed
out 


> later on we hear that coreos is "stealing" gentoo's ideas and hope 
> that it is CRUSHED.

That references my long history with large corporations, like the MAE
system that worked fine until the US congress gave the (US) internet to 
the conglomerate Telcos; and I issued a warning about that rant. It
was only to substantiate what conglomerates do to otherwise wonderful
open source projects, imho.


> but why? its its own frigging distro now. not gentoo by a long shot.

Many have stated that CoreOS is a gentoo (certainly inspired) derivative.
The focus of "MY THREAD" is the ideas and technologies that CoreOS
has lifted from Gentoo and my search for a robust "Clustering" paradigm
that is gentoo centric and thusly landed squarely where CoreOS is. I have 
found many legacy codes that did the same thing as what CoreOS is doing,
but for one reason or another they were abondoned.


> On Wednesday, December 03, 2014 02:39:53 AM Mark David Dumlao wrote:
> > Why do I get the feeling that this is another episode of the "i hate
> > LennartSoft(tm) too" circlejerk on the gentoo mailing list?
> Why do I get the feeling you just want another flamewar?
> I don't see any mention of systemd or anything else written by Lennart,  >
apart from your comment.

This is correct, but, very sad. Since Gentoo is openly supporting OpenRC,
I'm staying with Gentoo. If I want a thread on Systemd, I'll be sure to
put it in the title. If systemd is casually mentioned, please don't
get your "panties in a bunch", EVERYONE, as systemd is going to fine
and the other init centric folks will be fine too.


> > this mailing list used to be about gentoo.
> It still is.

AGREED.

> > > I do not intend to follow the CoreOS commercial path. I intend to mod
> > > gentoo to achieve those attractive attributes back into my "gentoo
> > > proper".

Boy, this and many other theme sentences pretty much spell out my
interest in this thread. If anyone researches gentoo's history
there was a rich environment on HPC, distributed, and clusters; somehow
it all was allowed to atrophy and I do not find any valid reasons.
My science/math needs dictate to me a need for a robust cluster based
on Gentoo. My embedded needs dictate a need for a gentoo cluster. The
deprecation of Tinderbox at Gentoo strongly suggests a need for a gentoo
cluster. My routine admin needs dictate a need for a Gentoo Cluster.

My girlfriend likes the idea of a Gentoo cluster.

CoreOS is nothing more than something where I can robb original gentoo
thunder from, for my gentoo cluster. Other than that, I do see CoreOS
and it's primary sponsers, as *EVIL* OK? ymmv.


And finally, I think that alll init systems are going to become very
irrelevant in the next few years, as what they provide, can be passed
from a *personal cluster* to any and all hardware, dymanically. That's
what the cell phones (smart phones) do now. That is what the NSA
has been doing for over a decade now. In fact that is what most all
major nation states have been doing for a very long time. It's been
"game set match" at the transistor level and with numerous back doors
in the Rf domain, hidden deeply in the "Rf noise domain" for decades.
Historically it was called "signal intercept". Do your research or 
find an accomplished EE with a few decades of experience in Rf and
listen to them. It's old hat. 

Get real. Systemd is a piss_ant and is irrelevant, IMO!
Openrc is not in my critical path either, although I have a very,
very strong affection to it. It's called loyalty and much
of the symbiotic relational world is build upon loyalty. Some
do not understand this, and I cannot help those folks that do
not understand loyalty.



So, let's focus on modernizing Gentoo, shall we?

OK? (focus dude, focus).

hth,
James

Re: [gentoo-user] Re: Custom ebuilds for CoreOS

[Saifi Khan]

On Wed, 3 Dec 2014, Rich Freeman wrote:

> On Wed, Dec 3, 2014 at 9:28 AM, Mark David Dumlao <madumlao@gmail.com> wrote:
>>
>> but why? its its own frigging distro now. not gentoo by a long shot.
>>
>
> I think it is actually a compliment to the flexibility of Gentoo that
> these derivatives are so different.  Gentoo is a somewhat-generic
> linux distro overall - in its default install it isn't too different
> from Debian/Ubuntu/Fedora/Arch on the surface and in terms of typical
> package selection.  However, ChromeOS and CoreOS are very
> non-traditional linux "distros."
>
> When people ask me what Gentoo is "good for" I of course talk about
> enthusiasts who care about both understanding their systems and having
> a high degree of control, but I also talk about projects where you're
> trying to blaze new trails and departing significantly from the
> typical "linux desktop" or LAMP box.  If all you want is a stable LAMP
> box then honestly you're probably better off with the likes of
> Debian/CentOS/etc.  However, if you're doing something embedded, or
> trying to change the world, then starting with Gentoo gives you a lot
> more flexibility to blaze new ground while not having to build
> EVERYTHING from scratch.
>
> So, when people use Gentoo to do things that we personally don't find
> useful, I think it is just a testimony to the fact that we've actually
> accomplished one of our core missions: empowering our users to make
> their own choices.
>

+1

more power to you Rich.


thanks
Saifi.

Re: [gentoo-user] Re: Custom ebuilds for CoreOS

[Tom H]

On Wed, Dec 3, 2014 at 10:17 AM, James <wireless@tampabay.rr.com> wrote:
>
> And finally, I think that alll init systems are going to become very
> irrelevant in the next few years, as what they provide, can be passed
> from a *personal cluster* to any and all hardware, dymanically. That's
> what the cell phones (smart phones) do now.

What do you mean? (Android has its own init and iOS has launchd.)