On 27/04/17 16:09, David Colling wrote:
> Dear All,
>
> We are due our regular technical meeting tomorrow. So I am looking for
> suggestions for discussion topics. Please reply to the list as your
> reply might generate others.
>
> Best,
> david
One thing to mention about "Glite plain vanilla" (for want of a better
name) is that our (i.e. at Liverpool) Centos7 ARC/HTCondor worker node
cluster has run some tests from ATLAS, LHCb and ILC and so far it seems
to work OK. Since many sites and VOs will move to Centos7 sometime, it
might be right to drum up some more support and (hopefully) lay out a
firmer time-line/implementation plan for the switch, e.g. UMD software
release plans and ideas for "Yaimless configuration", and switch to
exclusive GFAL file copies. I hope to prepare some documentation for
some of this in due course, but I'm presently working on moving Service
Node to Centos7, so I'm short of time.
On a related note, various "modern and funky" options for deployment are
being openly discussed, trailed and even operated (at RAL). They combine
ARC, HTCondor and Singularity/Docker for an alternative way of
separating jobs. These options may become "Mainstream". It would be good
to know the chances of that; we have only so much effort to put into
config standards, and we should back the best technology candidates for
providing services. I'd like to hear some views on that.
Also I would be interested in news about the BDII. One problem relates
to VAC. VAC has no out-of-band publishing (aka BDII), and to transmit
capacity information it is possible to munge it into the values given by
a site's alternative platform, e.g. ARC. This is hardly satisfactory, so
we need to investigate how to deal with this. Munging is an arithmetic
chore, and it is not suitable everywhere. In my view, it's a desperate
solution, and we need to fix it.
On VAC, I'd like to hear some news/explanation etc. on multi-core
possibilities. We're running version 2, but I need to refresh my mind on
(a) how mcore on VAC works, and (b) when it might be possible to try it
etc. Also would like to hear if other VOs are eager try it out etc. etc.
That's all I can think of for now...
Cheers,
Ste
--
Steve Jones [log in to unmask]
Grid System Administrator office: 220
High Energy Physics Division tel (int): 43396
Oliver Lodge Laboratory tel (ext): +44 (0)151 794 3396
University of Liverpool http://www.liv.ac.uk/physics/hep/
|