On Mon, 2011-04-04 at 18:03 +0100, Marko Vojinovic wrote:
> On Monday 04 April 2011 12:25:06 Mister IT Guru wrote:
> > The one thing I would love to be able to contribute my time to is
> > helping test new code, and get it out the door so guys on the street can
> > test it out.
>
> Before you get flamed-off by people who are already extremely pissed by previous
> infinity of discussions on this topic, let me try to summarize the answers to
> your questions, collected from all previous flames that were going on for the
> past three months. ;-)
>
> Hopefully, my answer could prevent yet another flame starting up... :-)
>
> Also, I am not a developer of CentOS (or of anything else) myself, but just an
> ordinary user. So I am just going to rehash and summarize what I have read
> from more knowledgeable people on this list.
>
> > Maybe it's my curiosity, but my brain tells me that Fedora is the
> > forerunner for RHEL. And the Fedora code is out there. CentOS is built
> > from the RHEL code, with all RHEL specific items removed. Ergo - If I
> > replicate the build environment on some of my machines,
>
> Herein lies the main problem: *there* *is* *no* *build* *environment* yet. In
> other words --- the Fedora environment is far too
> big/generic/unsuitable/whatever (am I right here?), and RedHat is not
> interested in giving details about their build environment.
I think I am beginning to understand. No build environment? *applauds
the CentOS devs* Wait does this mean that you use trial and improvement,
as you attempt to get to 100% binary compatibility? Awesome
If this is the case, surely that must take a lot of man power? Dev guys
- Ask the list for ten minions to do your bidding. I am prepared to
become a minion, it would really help if more people had your back when
it comes to RHEL doing updates etc. You won't even need to remind the
list, you got minions for that!
>
> So the main problem that CentOS team has to solve with each major release is
> to construct a build environment that will produce binaries that are bit-by-
> bit equivalent to official RHEL (up to trademarks, branding and some other
> stuff).
Okay - This makes sense. Do we have a flow chart somewhere online that
details this process? Where can assistance be provided? If the CentOS
devs can give me a spec on thier build environment, I'm sure I could
devise a way to allow others to duplicate the same environment in KVM
and help.
>
> >From my naive understanding, this boils down to the proper order in which
> packages are supposed to be built. There is more than one possible ordering,
> and only one will give binary equivalent set of packages.
A lot of coffee required here! Woah, serious dev guys, is the workload
to this degree? Hey Devs, we *OWE* you! we owe you BIG time, put us to
work dammit!
>
> I am probably oversimplifying things, but it roughly goes as follows:
>
> 1) start from some build environment
> 2) compile the whole distro
> 3) compare the result bit-by-bit with RHEL binaries
> 4) if it matches you're done; if it doesn't match, modify the build
> environment and go back to 1).
This is a major achievement for the CentOS devs. Can't we share our
spare cycles, and build some sort of bastardised deep blue? Crank
together our own grid! *maybe when we hit CentOS 9 or so we will be,
here's hoping!*
>
> AFAIU, the CentOS devs are currently in the above loop. Once they are done,
> testing will begin and CentOS 6 will probably be released shortly thereafter.
>
> However, nobody knows how much time is it going to take to finish the loop. Not
> even the devs can estimate that, so better don't ask them! ;-)
Time, time time! I don't care how long it takes, so long as it gets
done! I have enough faith in previous CentOS builds to be able to wait
until the next one is ready. Anyway, I *never* update my production
servers until my test rigs are rock solid, and there is at least talk of
another update :)
>
> I hope that this clears up some things.
>
> > (KVM and XEN
> > both running riot all over my systems, but not doing anything useful for
> > me! :( ), then surley I should be able to get some postive results, and
> > be able to contrib that back to the guys upstream.
> >
> > That's what my brain tells me. I don't mind running build environments,
> > or test environments or whatever - I guess what I'm saying is GIMME SOME
> > OF YOUR WORKLOAD!!
>
> As should be obvious from above, the problem is not in the workload. It's
> about reverse-engineering the build environment. More computing power (or
> manpower for that matter) will not help in a significant way.
Woah, what a way to crush my hopes of a grid of global CentOS systems
kicking IBM in the nuts. So to further my understanding, just so that we
can maintain binary compatibility with RHEL, the CentOS devs have to hit
on by chance a build environment that produces the same output as the
equivalent RHEL version.
>
> In general it could help, but the devs need to invest some serious time to
> train you to do that job, and they don't have the time for that now. It's a
> good idea to report back with the "gimme some of your workload" statement a
> couple of months *after* the CentOS 6 is finally out, ie. when the devs do have
> some free time on their hands to teach you what and how to do. Then you'll be
> able to help with CentOS 7, for example.
It's a start! I understand that at this point in time they will indeed
be busy. Well, in that case, I think I might have to jump on some CentOS
devs twitter feeds, so we can keep in touch *hint*
> The problem is that after the major release is out, users lose enthusiasm to
> provide help, and basically nobody wants to invest time to learn and train to
> build the distro. AFAIK, this is the experience from the time of building of
> CentOS 5. If I remember what Johnny said about this, out of a whole bunch of
> people who offered help during the C5 build, *only* *one* was interested to
> offer his help *after* the build. So people are not consistent in this.
If your a linux admin - it is REALLY good to have this stuff on your CV!
I understand that the timing of these questions on the mailing list
probably would always come up just as the devs are at their most busiest
working on a release. Well I for one hate(love really) the fact that
ubuntu gets spoken about all day long. Are you telling me, that of all
the full time employed people worked earning only one carried on
working, once everything was released? Sounds almost biblical.
Well, I guess the question is, do the CentOS Devs actually *want*
assistance in this area? if it would be more of a hindrance, then we
need to 'remove' that hindrance. If it's a lack of skills within the
CentOS community to actually be able to complete the work, then we need
to get some people in that will help. If a dev got hit by a bus, we're
screwed!
>
> My suggestion to you is to wait for C6 to be finished, and *then* offer your
> help and time to devs. I bet that they'll get you up to speed with everything
> you need to know, if you have proper skills to do the job. ;-) Then you could
> help for the build of CentOS 6.1 when it becomes relevant, or CentOS 7 later
> on.
I sure will do. I can admit right now, that no, I do not have the skills
to compile a distro, but give me the weblinks and a test rig, and I'll
write you a masterpiece! (eventually!) Okay, at least make a
contribution.
>
> > Or at least make it easy for other bored sysads to help you out. All
> > this spare processing power and capable guys and girls eager to support
> > our distro of choice to get the best bleeding edge stable code. It's
> > almost like following a football team! How DARE debian get ahead of us!
> > Gentoo!? Who the bleeding hell do you think you are!? Don't you know
> > CENTOS is in the HOUSE!?
> >
> > *calms down*
> > Excuse my excitement. I could edit this email before I hit send, but
> > then you guys wouldn't really know how I feel towards CentOS. How can
> > the average guy get involved with testing, can we build the same
> > environments as you guys? Do you have a standard way of operating that
> > maybe some of us could learn, and contribute? Is it out there already
> > out there and documented? How can we get our hands dirty?
>
> See above. Also, you can get your hands dirty in other ways (testing,
> maintaining the website, mailing lists, etc.). Ask on the devel list for TODO
> jobs that you can help out with.
Ah yes! The DEV list? *off to subscribe* I will lurke on here. And I
urge others who are even mildly interested in testing, and what Marko
said above, to subscribe too, but only talk if your helping! hehe!
>
> But don't ask to help with building C6, you're probably too late to offer help
> there. Instead offer your help in building C7, but early enough so that people
> can teach you what to do and how to do it.
I'm off to join the dev list to do just this.
>
> As a final comment, note that RHEL has put out versions 5.6 and 6 basically
> simultaneously, so the CentOS devs decided to first build 5.6, and then to go
> to 6. The logic behind this is that people on 5.5 do need the update to 5.6 as
> soon as reasonably possible, while noone needs 6 so promptly, so it can wait.
> That is also one of the reasons why C6 is lagging behind RHEL6 for such a long
> time (not that there is any sane definition of "long" in this case...).
The above is just common sense, no one should really complain about
that! How many times have we been asked by a manager to do two workloads
at once, and how much do we actually achieve? So "lay off the DEVS, they
have buckets of work to do as it is! And be happy with 5.5 dammit!!"
>
> I sincerely hope that this will answer your questions and prevent yet another
> escalation of flame on this list... :-)
>
> HTH, :-)
> Marko
Thank you Marko, and others that replied too. I'll only reply to this
response but I did read, and I appreciate the comments in this thread. I
just want to see CentOS everywhere, and be an expert in it :)
--
Mister IT Guru
Blog http://solosysad.blogspot.com/
Twitter @misteritguru
Follow me, I follow you - it's only fair
_______________________________________________
CentOS mailing list
CentOS@xxxxxxxxxx
http://lists.centos.org/mailman/listinfo/centos
