[Brahms-dev-l] Fwd: RCF Processor Farm OS Upgrade

Hi All,


RCF is on another upgrade cycle. I propose, essentially since I have 0  
time for this stuff until mid-October that
we wait until then. Kris and I have successfully tested our brat,  
brag,  packages , and they work with almost no mods
once compile flags were set correctly.
Only thins not checked was the proof part - this remains.

I appreciate your comments.

Flemming






Flemming videbaek
videbaek @ bnl.gov
Brookhaven National Lab
Physics Department
Bldg 510D
Upton, NY 11973

phone: 631-344-4106
cell: 631-681-1596




Begin forwarded message:

> From: Chris Hollowell <hollowec_at_bnl.gov>
> Date: August 20, 2009 5:09:12 PM EDT
> To: videbaek_at_bnl.gov, steinber_at_bnl.gov, haggerty_at_bnl.gov, throwe_at_bnl.gov
> Cc: jeromel_at_bnl.gov, cmvale_at_bnl.gov, pinkenbu_at_bnl.gov, tony_at_bnl.gov
> Subject: RCF Processor Farm OS Upgrade
>
> Hi All,
>
> Following recent discussions with PHENIX and STAR at the RCF liaison  
> meetings,
> we're planning on transitioning their processor farm systems to 64-bit
> Scientific Linux (SL) 5.3 in October/November, before the next RHIC  
> run
> begins at the end of December.  The migration to SL5 is necessary to  
> support
> new servers, and the transition to a 64-bit kernel is necessary to  
> officialy
> support the use of over 16 GB of RAM in SL5 (see "Technology  
> Capabilities and
> limits" in this document: http://www.redhat.com/rhel/compare).  SL64  
> 5.3
> is already in use on the ATLAS processor farm hosts (upgraded in  
> July).
>
> Our general policy has been to keep all RCF (STAR, PHENIX, PHOBOS,  
> BRAHMS,
> LSST) farm hosts at the same OS revision.  While LSST is not a RHIC
> experiment, its machines exist on the RCF side of the RACF facility
> (they use the RHIC/RCF AFS Cell, Kerberos KDCs, Condor central  
> manager,
> etc.).  To allow RCF general queue operations to continue, and to ease
> administrative overhead, we would like to continue the RCF farm's
> homogenous OS image policy.  Therefore, we would like to upgrade the  
> PHOBOS,
> BRAHMS, and LSST hosts to SL64 5.3 at some point as well.  This does  
> not
> have to be done at the same time as STAR/PHENIX, since PHOBOS/BRAHMS/ 
> LSST
> are not constrained by RHIC machine scheduling.  The PHOBOS/BRAHMS/ 
> LSST
> farm upgrade could be completed earlier, or slightly later if  
> necessary.
>
> PHENIX and STAR are still running tests to determine if they will
> use gcc 4.1 or gcc 4.3 as the default SL64 5.3 compiler (the stock  
> default
> is 4.1).  There are currently two play machines available for SL64 5.3
> evaluation: please test the functionality of your software on these  
> hosts.
> I've attempted to make these systems as ABI compatible with 32-bit  
> SL4 as
> possible, so you should have little/no issues running SL32 4.4  
> binaries
> on these machines.  An SL4 compatibility compiler, gcc34, is also  
> available
> on these systems.
>
> rplay45.rcf.bnl.gov - SL64 5.3 gcc 4.1
> rplay39.rcf.bnl.gov - SL64 5.3 gcc 4.3
>
> Thank you.
>
>
> -----------------------------------------------------------------
> Chris Hollowell <hollowec_at_bnl.gov> - Advanced Technology Engineer
> RHIC/ATLAS Computing Facility [https://www.racf.bnl.gov]
> Physics Department, Brookhaven National Laboratory
> Bldg 510M, Upton, NY 11973
> (631)-344-8031

_______________________________________________
Brahms-dev-l mailing list
Brahms-dev-l_at_lists.bnl.gov
https://lists.bnl.gov/mailman/listinfo/brahms-dev-l