Thanks Konstantin for all of this great information.
I'll second the recommendation to watch Troy's video.
It makes me feel a bit more comfortable that Alma/Rocky will be able to
deliver the 10 years of updates.
I still have a sour taste for this new "module" packaging scheme. It
seems to make rebuilding installation media with a custom package set a
headache. I've managed to hack something together, but it isn't
pretty. Has anyone else successfully tackled this?
-Mark
On 5/4/21 12:41 PM, Konstantin Olchanski wrote:
On Mon, May 03, 2021 at 09:14:11PM +0000, Dave Dykstra wrote:
Here's a presentation at HEPiX'21 from CERN that's publicly available:
https://indico.cern.ch/event/995485/contributions/4256466/
Thank you for the link, I was not aware of this presentation.
For further reading, it contains a link to the March meeting of the linux
future committee: https://indico.cern.ch/event/1019875/
All materials are publically available, the zoom chat transcript requires
a CERN login (which I have), I am reading through the stuff now.
First impressions from the HEPIX presentation:
- there is no explanation how CentOS Stream is a technically acceptable
replacement
for CentOS Linux. There is many questions here, none answered.
- there is no suggestion of reviving the CERN+FermiLab collaboration
(named SL or otherwise). This is surprising, if Princeton can roll out
a "new centos", CERN+Fermilab have even more resources to do same.
- there is no suggestion of CERN providing extended support for CentOS-8. To
me, this would be the obvious path forward.
- first slide of "distribution landscape" is nonsense, with everybody stuck
with el7 for another 3 years and bye, bye, c++14, c++17, c++20.
- second slide of "distribution landscape" has LHC experiments changing horses
in the middle of Run 3. Good luck with that.
- no discussion of "red hat takedown/takeover of centos, 2.0" scenario for Rocky
Linux & co.
First impressions from the "linux future" meeting in March:
- 30 minutes of our Troy Dawson leading Red Hat's video presentation:
https://videos.cern.ch/record/2756480 (where they say "rail", they mean RHEL,
took me a while to catch this)
- "CC7 is not an option to use for RUN3 at CMS", page 4,
https://indico.cern.ch/event/1019875/attachments/2214410/3751654/lfc002.pdf
- support for 64-bit ARM is a requirement (good, but what about existing 32-bit
ARM and 32-bit x86 machines?)
- page 5, key take away, "Most likely moving to CS8 is your best option, as this
allows for a trivial upgrade path to EL8 at a later date".
(Is this true? Can somebody point me to the instructions for upgrading Centos
Stream 8 to RHEL 8? A quick google search finds nothing)
- Linux BOF slides
https://indico.cern.ch/event/995485/sessions/386343/attachments/2209929/3739897/hepixLinuxBoFMarch16th2021.pdf
- Linux BOF 82 min video https://videos.cern.ch/record/2756412
- Ben Maurice slides https://codimd.web.cern.ch/p/Zcc5CqncC#/1
- zoom chat (requires cern login) - all questions raised are reflected in Ben
Maurice slides.
Bottom line.
All arrows point at CentOS Stream, but there is no review/evaluation of it's
technical suitability
for HEP specific needs. Good/bad/etc compared to el7 and compared to non-RPM
alternatives (ubuntu,
etc). And there is many questions, none answered.
It looks like everybody will have to do their own evaluation and decide for
themselves.
P.S.
Do watch Troy's video!
