Minutes of dTeam meeting 25 Nov 2005.

Present:
	Glasgow: Graeme
	Imperial: Olivier
	RAL: Chris, Jeremy (chair), Stephen, Jens (mins)

Review of actions

"Follow up on questions about Brunel and submitting SFTs to sites not
in the RB used by Polish SFT submission" - new SFT questions with LeSC
joining (although jobs run successfully at LeSC).  May need to test
RB.  New Action Olivier to follow up with appropriate lcg-rollout.


"Reconfirm expectations for Tier-2 hardware" - Olivier announced at
IC-HEP will get new hw on time.

"Starting with Alessandra review document in Wiki" - it is known not
to have reached Graeme so must be either with Alessandra or Fraser.

"Investigate methods to remove old transfer files (for SC4)" - should
also be raised in storage group (the storage group decided that FTS is
also within its scope, but mainly because there are 'delete' issues
with some SRMs - see e.g. bug #10120).  In fact it has been raised but
nothing much has been done yet.

"Find out hs are to know which experiment software version to have
installeed" - Olivier reports that he talks to (Atlas) SGM, but that
doesn't scale.  Jeremy needs to raise issue at GDB.


Site name change proposals

Done at IC after shutdown; RHUL interested but need procedure (other
admins have also asked).  Suggested naming scheme UKI-* where * is
decided by Tier 2 coordinator and must be consistent.


Feedback from LHC CRRB


Graeme's talk well received by review board, also good feedback from
Les & Ian.  Q: about communication between T2 & experiments, A: needs
improving.  Board also asked about install tools.  SC3 throughput
tests failed and need redoing in January.  VOBox mentioned but no
comments.


Update on security challenges

Jeremy mentioned the jobs were running but delayed due to a cleanup at
RAL.


Progress with SC4 preparations

SRM deployment OK, sites that said "mid Nov" have started.  Only
concerns about ~2 sites now.

Graeme reports Matt Hodges involved in FTS, setting up channels,
documenting.  Performance problems (low xfer rate) to dCache needs
resolving, getting good performance between DPMs.  Testing between
Glasgow and Edinburgh because the latter has both.  Needs testing with
1.6.6 too - Edinburgh and RAL-Storage (not Tier 1) have 1.6.6, but
Tier 1 upgrade imminent (30 Nov).

Question about monitoring GridFTP rate for dCache - storage group had
this as bug #10092, but annoyingly it was closed without commenting
what the solution was.  I will try to locate the information, IIRC it
required increasing log verbosity in the old days but this may have
improved.  Data needs to be published via R-GMA.  ACTION Jens&Graeme:
how to do this.

How to get at-a-glance xfer rates between sites?  The status page is
already getting crowded.  May need planning on a different page from
the status?

LFC document: pretty much done, Durham tested it, worked fine modulo a
certificate problem.  Difficult to do more than a nameserver test.

ACTION coordinators: need >= 1 LFC in each Tier 2 by end Dec -
nominate a site.

Chris didn't need much interaction with Tier 1 to set up FTS client;
tuning may be more interesting.  Caution: There is more than one page
that claims to have the "latest" version.  There is only 1 server, at
RAL.

Network test: Graeme mentions to (1) let network people know you are
doing them (and when), (2) do them overnight.  Will try 1 TB
overnight.

Jens mentioned that he has asked Pete and Gareth to put SC4
storage/FTS preps on the hepsysman agenda, with speakers tbd.


UKI performance

Glasgow has large MPI jobs clogging up cluster, blocking out FTS.
QMUL had PBS problems.  Maybe need a queue for short jobs.


AOB

Jeremy says admins can now edit ROC report within a specific time
window, site admins can record reasons for failures.  See mail to
list.  Feedback to Jeremy.

Jens reports that he's interacting with ETF where ~3 sites are
installing gLite components, and a new gLite report was circulated to
ETF (but is not yet circulatable outside).  ETF evaluating DPM.  Also
asked to feedback on gLite install (as a non-LCG shop).

In general site admins advised to do same feedback.

------------------------------------------------------------------------
Summary of new actions

Olivier to follow up on LeSC/SFT/RB problem with LCG/lcg-rollout.

Jens&Graeme: (1) get dCache gridftp xfer performance out of dCache and
(2) publish via RGMA.

Coordinators: nominate a site in each T2 to have LFC installed by end
of Dec.  That is _installed_ by end Dec, not nominate by end Dec.