Minutes of storage phone conf 13 Dec 06

Present:
	Glasgow: Graeme
	Lancaster: Brian, Matt

Apologies:
	Durham: Mark
	Edinburgh: Greig (stuck in traffic)

0. In the interest of time, we postpone the action review to the end
   of the meeting for once - we must close at 10:30 for the UKI ROC.

1. DONM

   Shall we have meetings -

	- 20 Dec ?	YES

	- 3 Jan ?	NO MEETING

	- 10 Jan ?	TBD, probably not

	- 17 Jan ?	NO MEETING (Virtualisation/dCache workshop @DESY)

	- 24 Jan ?	NO MEETING (CERN workshop)

	- 31 Jan ?	YES.

2. Using T2 helpdesks for chasi^W reminding admins

   This was a proposal made by Olivier yesterday.  On closer
   inspection, there is indeed a Tier 2 queue in the NGS helpdesk, but
   there were only a handful of tickets, mostly London.  Not clear
   whether T2 admins have access to the helpdesk.

   Nevertheless, submitting via GGUS is a possibility; it should
   eventually filter through to the site.

   Northgrid have a common mailing list.  ScotGrid have a helpdesk,
   but not used for cross site support in general.

3. Items for dCache workshop in January
   https://indico.desy.de/conferenceDisplay.py?confId=138

   There is also a virtualisation workshop two days before the dCache
   workshop.

   Greig will be attending both.  Anyone else?

4. VO pools

   Best practice is to have at least one pool dedicated to the site's
   main VO (for example, Glasgow and Lancaster, being Atlas sites,
   have a large dedicated pool for Atlas, and a smaller shared one).

   Some sites are thought to have all users swimming in the same
   pool.  This is not so good - may need to work with them to change
   this setup.

   How do we discover this?  For sites that haven't upgraded the DPM
   plugin, we can see the same available space being published for the
   VOs sharing a pool.

   Plus, of course, sites should upgrade their DPM plugins.

5. Deliverables revisited - status, todo.

   Jens mentioned the plan, but this item now mainly affects Greig and
   Jens.  It's planning for the rest of GridPP2, and GridPP3 (assuming
   it gets funded).  May need more feedback on middleware gaps from
   August deliverable - see agenda for meeting
   http://indico.cern.ch/conferenceDisplay.py?confId=9245

6. SE Robustness revisited

   Was also raised in dteam yesterday.  Failures are often due to
   other services falling over (like the BDII), but nevertheless it is
   useful to discuss how the robustness can be improved for the SE.

   The most straigtforward way ("low hanging fruit" - sorry :-) is to
   improve the internal monitoring, to provide early warning to the
   admin when something is about to go wrong.  Graeme hasn't had much
   time to work on the MonAMI plugin recently.

   Pete Gronbech's DPM failed - daemon failed or locked up.  Failure
   easy to detect; lockup less easy.

7. AOB

   Graeme mentioned RFIO testing for DPM from the WNs.  Greig is
   talking to Dave Newbold about the jobs DN used to test CASTOR RFIO
   at RAL.

   Analysis jobs will be using RFIO - but strange routing problems can
   arise for the jobs on the WNs, cf
   https://www.gridpp.ac.uk/wiki/DPM_Dual_Homing for GridFTP.  This is
   not required for RFIO but may improve robustness.

0. Usual review of actions (see below)


------------------------------------------------------------------------
ACTIONS

160	21/11/2006	Follow up with Brian & Frederic re Atlas T1->T2 xfer/tests	Jens	Open	

Ongoing.  Waiting for CASTOR resources for Atlas to become ready.

163	06/12/2006	Volunteer storage advice expert panel to assist sites purchasing	Greig/Jens	Open	

Currently this panel is Graeme and Greig - since they have recently
gone through purchases.  Next time it may be Duncan.

Tier 1 can perhaps also help.  Disk expert is Nick White.


--- no new actions this week.
------------------------------------------------------------------------