Good morning, all.

Since CERN dropped support for IA64 with SLC 5.4, there isn't a freely 
available, installable, EL5 rebuild for IA64.  In fact, the only free option 
that's reasonably up to date is Debian 6, and while that works fine, I'm pretty 
much standardized on either CentOS or SL here, and would rather not have to run 
Debian, simply for standardization.  Plus, SGI's tools are built either for 
SLES or RHEL; thus, to use any of SGI's stuff (like Performance CoPilot, a part 
of ProPack) an EL rebuild is required.  

I have access to three SGI Altix machines, and am planning on rebuilding, using 
SLC 5.4 as a base, up through 5.8, either using a CentOS source base or an SL 
source base, or a mix of both.  They aren't especially large machines, at least 
from the Altix standpoint, but they are useful for us.  One is a 3700 with 20 
CPU's, but it seems to have some issues of some sort, and I need to work it 
over with a fine-toothed comb.  Another is a 4 CPU 3200, but one node of the 
two in the single C-brick is broken so only 2 CPU's are working, but they are 
stable, completing a full kernel SRPM rebuild under SLC5.4, even though it took 
several hours.  And the third is a 30 CPU Altix 350, and it's also running well 
and seems to be stable on SLC5.4, also completing a kernel SRPM rebuild, but 
much much faster than the 2 CPU 3200.  Due to the obvious security issues with 
running a 5.4 release, these machines aren't externally visible.

I'm posting this to gauge the potential interest in such a project by the 
larger SL community.  I know the amount of interest in the CentOS community 
(essentially none from the user standpoint, and a small amount of interest from 
a couple of the CentOS developers, including Karanbir), and the SL CERN 
developers have seen basically no interest from their user community (which is 
why they dropped IA64 support), but I'd like to see if anyone here has an 
interest.

If so I'd like to collaborate; if not, well, I'm building up the infrastructure 
to do my own internal-only rebuild, which will probably have a mixed SL/CentOS 
base.  I don't plan on distributing my rebuild (otherwise I would need to do 
the same trademark stripping CentOS and SL do with the upstream sources), but I 
do plan on documenting the process and the tools, once I have it rebuilding and 
running.  If you are interested, but you're not able to collaborate but are 
willing to follow along on your own hardware, I'll be happy to post status 
reports and scripts, etc, to this list or the SL-developers list as the build 
progresses so you can follow along and get your own rebuild running.

I'm specifically looking into smock and/or mockchain to drive a mock rebuild, 
and, starting with the fully updated SLC5.4 base, rebuild 5.5, 5.6, 5.7, amd 
5.8 in order, again either from the CentOS, SL, or even the upstream source 
base.  And then, of course, rebuilding EPEL, RPMforge, and maybe even ATrpms 
will be on the plate.  There are a few other small repos and packages, too.

And my thanks to the Jaroslaw Polok and the team at CERN for doing the really 
hard work of getting SLC5 on IA64 bootstrapped in the first place; that is much 
more difficult of a process than what I'm getting ready to do.

And, if interest is there, an EL6 on IA64 could be on the plate, but I know 
that one will be a lot of work and may not even be possible.  But I reserve the 
right to be wrong....  at least with an EL6 rebuild, binary compatibility won't 
be a consideration, since upstream has no IA64 EL6 (unlike EL5, where IA64 is 
still supported by upstream).

I welcome your comments and suggestions.

Reply via email to