Minutes of storage EVO workshop

Present: Chris, John, Matt, Brian, Alessandra, Sam, David, Govind, Stephen, Jens

WLCG: increase use of xroot.

OGF: included review of EMI and UMD.

New config requires downtime, but EMI insist on moving configuration.
Support for YAIM to cease.  CREAM CE breaks on YAIM anyway.

So we need to do our own configuration management, eg with puppet or
cfengine (or Quattor).  Would help if services were documented.
RPM should also create sensible config files for initial install.
(But the services may need to know which VOs are supported)

Jens then gave a summary of storage-related stuff at OGF, including
GFFS and SRM, and cross site Hadoop.  And enterprise SSDs (Gordon).

StoRM - fast data transfer

FTS replacement - FTS 3, services being prioritised.
Monitoring of FTS - now VO independent?

Discussion about LHCONE - our very own Mark Mitchell involved.

LHCb now talking about reprocessing at T2s, or at least some of them,
with transfers from T1 (presumably RAL for the UK).

10% of jobs failing - of which 1/4 fail due to not being able to
write output data - which is obviously bad since the job has already
consumed the CPU cycles.  Could write to second SE, possibly.  Also
bad from user experience, if jobs fail.

CMS moving to xroot, relying on its redirects if a file is missing.
Tape will increasingly be backup only at T1s.



[10:09:19] John Bland joined
[10:09:19] Matthew Doidge joined
[10:09:21] Brian Davies joined
[10:09:22] Alessandra Forti joined
[10:09:23] Sam Skipsey joined
[10:09:24] David Crooks joined
[10:09:25] Queen Mary, U London London, U.K. joined
[10:03:43] Alessandra Forti yes
[10:09:27] Govind Songara joined
[10:09:27] Stephen Jones joined
[10:19:37] Stephen Jones Sounds like we are each going to have to write our own version of yaim!
[10:23:03] Matthew Doidge Today's xkcd seems quite relevant http://xkcd.com/927/
[10:30:03] John Bland left
[10:36:58] Govind Songara left
[10:37:03] Govind Songara joined
[10:37:28] Jens Jensen  
[10:45:25] Alessandra Forti http://dashb-atlas-job.cern.ch/dashboard/request.py/failedjobsstatus_individual?sites=UK&sitesSort=8&start=null&end=null&timeRange=lastMonth&sortBy=0&granularity=Daily&generic=0&type=aadp
[10:45:40] Alessandra Forti last month summary of uk errors
[10:45:51] Alessandra Forti it can be tuned to different times
[10:54:11] Sam Skipsey Chris - NFS4.1 supports "multi-server namespaces" which probably allows redirection like xrootd does.
[10:54:19] Alessandra Forti left
[10:54:48] David Crooks left
[10:54:49] Govind Songara left
[10:54:54] Sam Skipsey left
[10:55:02] Matthew Doidge left