On 01/09/2014 05:38 PM, Karanbir Singh wrote:

On 01/09/2014 05:37 PM, Andreas Thienemann wrote:

...

Is anyone else interested in a i[36]86 build of CentOS7 and would be willing and able to contribute to it? Or is x86_64 the only release really needed?

If there's some interest, I am sure this could be turned into a nice project.

Thanks Andreas,

there is one rather large giant in the room : almost a third of the CentOS userbase hits the i686 tree. Ofcourse, just like any other statistic that could mean quite a lot, or nothing at all.

Having said that, there is a large multilib component in the x86_64 distro - while I havent looked as yet for specific numbers and packages, it seems we might need to build large chunks of i686 to satisfy those builds anyway ( stress on not-been-quantified-yet ).

ofcourse making things interesting are things like redhat-rpm-config with a hardwired -mx86_64 :)

Over the coming days, i will try and make this effort easier for everyone and also setup something that allows us all to pool resources in. Stay tuned :)

btw. while everyone stays 'tuned' - it might be worth pondering how much i686 patch coverage we might expect from upstream. We might be able to build this once, will we be able to keep it going trivially, painfully, not-at-all.

Its the same question and scope ( but lesser depth ) for the powerpc platform.

• KB

The weird bit may be which packages are missing on 32bit. I know the kernel source does not build a kernel for i686.

Quick review of spec files (not filtered for "obviously don't run this on 32bit") shows the ExclusiveArch x86_64 and not any 32bit for: glusterfs.spec:ExclusiveArch: x86_64 aarch64 gnome-boxes.spec:ExclusiveArch: x86_64 gnu-efi.spec:ExclusiveArch: x86_64 hyperv-daemons.spec:ExclusiveArch: x86_64 infinipath-psm.spec:ExclusiveArch: x86_64 libguestfs.spec:ExclusiveArch: x86_64 libipathverbs.spec:ExclusiveArch: x86_64 pesign.spec:ExclusiveArch: x86_64 qemu-kvm.spec:ExclusiveArch: x86_64 ras-utils.spec:ExclusiveArch: x86_64 sanlock.spec:ExclusiveArch: x86_64 seabios.spec:ExclusiveArch: x86_64 sgabios.spec:ExclusiveArch: x86_64 shim-signed.spec:ExclusiveArch: x86_64 shim.spec:ExclusiveArch: x86_64 spice.spec:ExclusiveArch: x86_64 supermin.spec:ExclusiveArch: x86_64

I have just upgraded my laptop from 8GB to 12GB RAM. Previously I was using F19 i686 with PAE, as I had no applications that wanted more than 4GB, especially since there was only 8GB total. The main reason for this is that I still get to use all 8GB and my 32bit apps like skype work more seamlessly.

Now that I have 12GB RAM I installed F20 x86_64 this time.

Although my experience with PAE in Linux was good, I cannot say the same where I tried it on Windows.

dave.

On Fri, Jan 10, 2014 at 12:44 PM, Sharuzzaman Ahmat Raslan < sharuzzaman@gmail.com> wrote:

Hi Andreas,

I personally interested in this effort.

My customer is mostly running their system on quite old system, but it does the job.

And I'm always provide my customer with CentOS, in anticipation if they want to move to a supported OS later, they can just install Red Hat, and very high probability that their application will just work.

But, if Red Hat, thus CentOS stops i686 release, I have no other option than moving my customer to other distribution that support x86 machine.

I also believe a lot of 3rd world country still running a lot of x86 machine, and this is actually a big market that Red Hat is abandoning.

Maybe Karanbir can point that out to Red Hat when you have a meeting with them.

Thanks.

On Fri, Jan 10, 2014 at 1:37 AM, Andreas Thienemann andreas@bawue.netwrote:

...

Hi,

I'm in the process of building myself a 32bit tree of the RHEL7 beta as I do have a few x86 machines I am not ready to retire yet.

Now that CentOS is all open and community etc. I was wondering if there'd be some interest for me to rebuild the results within the CentOS project and turn it into an official release long term. My unfinished work so far is based on an older RHEL5 buildsystem I still had lying around but I am happy to switch this to the CentOS toolstack if there's interest. This would mean filing off the serials, removing trademarks and all these things which I hadn't planned on doing initially but on the other hand, it should be very easy doing that benefitting from the regular CentOS work on these topics.

Is anyone else interested in a i[36]86 build of CentOS7 and would be willing and able to contribute to it? Or is x86_64 the only release really needed?

If there's some interest, I am sure this could be turned into a nice project.

cheers, Andreas

---

CentOS-devel mailing list CentOS-devel@centos.org http://lists.centos.org/mailman/listinfo/centos-devel

-- Sharuzzaman Ahmat Raslan

---

CentOS-devel mailing list CentOS-devel@centos.org http://lists.centos.org/mailman/listinfo/centos-devel

Hello Ryan,

Am 10.1.2014 15:28, schrieb Ryan Wagoner:

...

I do have a few x86 machines I am not ready to retire yet.

Have you considered converting these machines over to VMs. At one client I purchased one new server and migrated 10 legacy servers over to VMware ESXi. When I did the math the power savings covered the server in 2 years.

Glad to hear that this was a viable solution and worked out well for you.

Dan was describing however 32bit embedded machines used as routers. I have similar use cases in mind where virtualizing is not the right solution and a native i686 build is needed.

Considering the numbers Karanbir gave which amounted to about 30% of the users hitting the i686 tree

cheers, andreas

On Fri, Jan 10, 2014 at 11:51 AM, Ljubomir Ljubojevic centos@plnet.rs wrote:

If you have a company with 40-50 workstations with 32-bit CPU and 1GB RAM, how would you provide all 40-50 via VM's?

Depending on what they are doing, there's a good chance they would be better off running one or two modern servers with some sort of thin client access. The existing workstations would probably work well enough running NX, x2go, etc. although you don't get any power savings that way.. Depending on the usage, the host side could either be simple multiuser access or be carved into VMs per user, group, or function.

Developing or poor countries use what they can, some even import old PC's from developed countries. Even in my country there are many people with 32-bit CPU's and 512-1024MB RAM. Some even only have 256MB. It works for what they need so they do not want to spend their money if they do not have to.

Actually, I'm with you on this. Years ago I used a re-spin called K12LTSP that was maintained up through CentOS5. You'd install it on a host with 2 NICs and it would come up working to PXE-boot thin clients on one interface, providing LAN/internet access on the other interface. It wanted to boot the same kernel on the clients as the server, so it was problematic for older client hardware when RHEL dropped 32-bit support in 6.x. There was some effort to package LTSP5 for 6.x as simple RPMs but I don't think it has been as popular - or as usable, partly because it is more complicated to set up and partly because it didn't work for the old hardware typically used as clients. (One of its other features was that it came with working java, flash, and a few other packages pointed at working update repositories back when that wasn't easy...).