#383 Proposal: ppc64 should not be a primary architecture for EPEL 6
Closed None Opened 13 years ago by kkofler.

= Proposal topic =

Conformingly with FESCo's decision to make ppc/ppc64 secondary architectures starting from Fedora 13 (see ticket 144), the EPEL SIG should be required to build ppc64 packages for EPEL 6 either on secondary architecture infrastructure or not at all.

= Overview =

It has been decided that ppc and ppc64 should not be primary architectures anymore starting from Fedora 13. There is no reason why this decision should not apply to EPEL. Sure, RHEL 6 has ppc64 as a primary architecture, but it is still our choice what architectures we want to consider primary for EPEL, e.g. we do not support s390/s390x as primary architectures for EPEL.

= Problem space =

One of the benefits of making PPC a secondary architecture is to eventually allow removing PPC hardware from our primary Koji farm. Having EPEL perpetuate ppc(64) support ruins that, in addition of making the architectures the Fedora project supports inconsistent.

= Solution Overview =

Ban the EPEL SIG from using Fedora primary architecture infrastructure for EPEL 6 ppc64 builds.

= Active Ingredients =

EPEL SIG, PPC SIG

= Owners =

kkofler


This is really not a workable proposal. EPEL would need to be built using koji-shadow for non intel arches. though we have committed to building for EL-5 until it is EOL which is March 31, 2014 We just now purchased 2 new ppc blades to continue to support ppc for EPEL. there is little point in tippling the epel release work load. since thats what would happen. it would also force us to setup a secondary-epel tree or some such thing for ppc and s390x. we have been offered a s390x builder and have been weighing up building EPEL for all supported RHEL primary arches for EL-6

I would like to know why building EPEL for ppc64 is such a bothersome thing for Kevin. While some arches for Fedora are secondary it doesnt mean that we as fedora don't care or don't want to support them. continuing to keep 2 ppc builders for EPEL seems like a very small thing. as long term that is all we would keep around.

I feel that this ticket is wasteful and should just be closed as invalid.

though we have committed to building for EL-5 until it is EOL which is March 31, 2014

Which is why this proposal is only about EL-6+, to prevent us from making such a commitment for 7 more years. Instead, we'd no longer have to carry around PPC machines less than 4 years from now (which is still extremely long on Fedora's timescale, but at least an end would be in sight).

We just now purchased 2 new ppc blades to continue to support ppc for EPEL.

Those machines could be added to the secondary arch instance instead (either now, or at EL4 EOL, or at EL5 EOL), where they'd also benefit Fedora proper. (And likewise for s390x.)

I still dont see what the problem is that opened this ticket.

Im the person maintaining the machines.

Im the person doing the epel release engineering.

All this proposal does is increase my workload for zero gain.

again i say this should just be closed as invalid

Replying to [ticket:383 kkofler]:

It has been decided that ppc and ppc64 should not be primary architectures anymore starting from Fedora 13. There is no reason why this decision should not apply to EPEL. Sure, RHEL 6 has ppc64 as a primary architecture,

If RHEL6 has PPC64 as primary arch, this should IMO apply to EPEL6 as well. F13 is no argument, if you want Fedora as an argument, you need to take F12 cause this is what RHEL6 is based on. And F12 still had PPC64 as primary arch.

but it is still our choice what architectures we want to consider primary for EPEL,

Speaking of "our" choice: What is the position of the EPEL SIG? Is there a place to catch up with their discussion?

Seeing as I'm leaving fesco, just like kofler is.

closing->invalid.

Anybody can submit a ticket to FESCo, not just FESCo members. So I'm reopening the roguely closed ticket.

@ausil: Will you still be maintaining those machines 4+ years from now when the long-term impact of this decision will make itself visible?

I will be maintaining them for the next 4+ years like i have for the last 4+ years, same as i have the rest of the Fedora build infrastructure. same as i would maintain the s390x builder if we decide to add it also.

you have not at all answered why i should have alot more work on my plate for this. as such im closing it as invalid.

cwickert https://www.redhat.com/archives/epel-devel-list/2010-March/msg00087.html though there was little response.

Dennis

cwickert https://www.redhat.com/archives/epel-devel-list/2010-March/msg00087.html
though there was little response.

The only response relevant to PPC pointed out stats showing that there's extremely little interest in EPEL for PPC, which makes me wonder why to support it at all, primary or secondary.

Kevin you are still ignoring the question.

The answer: Because if anything happens to you in the next 4 years, somebody else will be stuck with maintaining PPC builders in Fedora's primary Koji instance for at least 3 additional years (more if the same strange decision is perpetuated for EL7). Also because having a proper secondary architecture setup for EPEL can only help the EPEL project (e.g. there's interest in ARM which isn't even supported by RHEL, it could only be made to work as a secondary arch based on CentOS). You're only planning for your short-term convenience without thinking of the potential long-term burden on the project. 7 years is a very long timeframe.

Oh, and will there even be any PPC machines suitable to run EPEL left in 7 years? Macs haven't been PPC for years now. The current PS3 versions don't support GNU/Linux anymore. The XBox 360 never did. What's left? There's already very little interest in PPC.

Ill add it another way. to keep the builders the same, consistent and running we need EPEL for ppc64, since I plan to move to EL-6 on the builders once its released. being as im the one doing all the work i get the say and i say we are doing ppc64

Arm will use fedora, just like SPARC does there is no CentOS option for either.

Its not my short term convenience. There is lots of ppc based servers being made still.

Basically i feel you are stepping way out of line and im done talking to you on the subject.

Replying to [comment:11 kkofler]:

Oh, and will there even be any PPC machines suitable to run EPEL left in 7 years? Macs haven't been PPC for years now. The current PS3 versions don't support GNU/Linux anymore. The XBox 360 never did. What's left? There's already very little interest in PPC.

RHEL doesn't support PS3 or Apple Macs anyway. Your argumentation is based on Fedora, which is mostly valid there, but it has little relevance to Enterprise distributions.

Your point about Dennis getting hit by a bus or something in the next 4+ years is also semi-valid, yet there is time to try and involve members of the PPC SIG as it grows for Fedora.

Replying to [comment:8 kkofler]:

The only response relevant to PPC pointed out stats showing that there's extremely little interest in EPEL for PPC, which makes me wonder why to support it at all, primary or secondary.

Are you referring to [https://www.redhat.com/archives/epel-devel-list/2010-March/msg00088.html]?

Your initial statement was:

It has been decided that ppc and ppc64 should not be primary architectures anymore starting from Fedora 13.

and I still wonder who decided this. Was it the EPEL SIG or was it you?

I am going to add the following as best I can while trying to remain "being excellent to others".

1) This 'problem' was not brought up in either meeting or mailing list in the last 3 months.

2) The argumentative style of the ticket and replies is one that is meant to stir up problems. More of a "if the Rules don't apply to A how can you make them apply to B (or me)."

On the case of 1, I can deal with. However with 2, I am both tired and frustrated by these antics.

Login to comment on this ticket.

Metadata