EDG Sw quality meeting, Friday 14th June, 11:00, 513-1-27

Present: Gabriel Zaqunie, Jeff Templon, Dave Foster, Peter Kunszt, Markus Schulz, Emanuele Leonardi, Mario Reale, JJ Blaising, Stephen Burhke, Frank Harris, Maite Barroso, Bob Jones, Marian Bubak, Fabrizio Gagliardi, Paolo Capiluppi, Massimo Sgaravatto, Charles Loomis, Laura Perini, Federico Carminati, Les Robertson, Louis ?, Flavia Donno
Phone: Francesco Prelz, John Gordon, Mirco Mazzucato, Yannick Legre, Julian Linford


Bob chaired the meeting and suggested the following subjects for the agenda:
Subjects: 
* What quality measurement do we want
* How are we going to achieve it?
* What effects will it have on the sw release procedure?


The participants agreed with subject for the meeting.
Fab added that this meeting should be put in context with the next meeting on Monday and the PTB on July 3rd.


Federico presented slides (http://doc.cern.ch/archive/electronic/other/agenda/a02737/a02737s1t2/transparencies/WP8-to-PTB.2002.06.14.ppt) giving the WP8 point of view of the subject matter

WP1 sw:
96% success rate for JJ's job submission rates was achieved with WP1 job-resubmission facilities being disabled.
This success level is considered acceptable.
Most problems came from configuration issues for Res Broker machine.


WP3 sw:
Nordugrid bug fix for MDS problem looks promising but needs testing on the development testbed
It is important that these bug fixes are taken back into Globus and included in their future releases
Fab added that such a mechanism is in place with the help of Bill Allcock from Globus.
Unfortunately, no plans have been made by Globus for MDS development beyond MDS release 2.2 (June 2002)
Most important point is that the info system goes down when a site is removed and block the testbed.
Again, actual number of level of success can be discussed but means "x% LDAP search success rate."
For R-GMA introduction and use as default basis for info system must be as reliable/stable as MDS.

WP2 sw:
Insufficient testing done to date with GDMP and replication facilities.
Believe one of the problems is the GDMP developers are not testing it in a realistic, multi-VO environment
Understand the GDMP is a temporary solution but still need basic replication facilities now.
Not ready to give acceptance figures.

WP5 sw:
Insufficient testing performed for access to mass storage at CERN, Lyon & RAL
Not ready to give acceptance figures.
Use from non-priviledged accounts need to be tested. Documentation required.
WP5 will report on the work don to date at the next weekly WP mgrs meeting.

WP4 sw:
Information providers are very important and some recent fixes have been introduced.

WP7 sw:
No application dependencies hence OK.

A proposal for the deployment of the agreed release 1.2 on the production/application testbed was discussed.  The mware groups were reluctant to allow the development testbed evolve into the production testbed. The preferred solution is to allow a few more application group users on to the development tested and finalise the release 1.2, tag and freeze it so it can be installed correctly on the production testbed. Once this operation is completed the development testbed will be used for new development work. In the future (after release1.2) we can consider if it would b possible to release components independently.

Long-terms issues concerned how the integration is made and dependencies on too specific library versions.
Concerning testing, the projects still needs an integration testing phase between the application get involved. The ITeam cannot do all these tests themselves since they are already busy with building activities. Maite and Flavia requested they have scheduled time for testing the contributions on the development testbed before the Loose Cannons get involved. The idea of scheduling WP testing activities on the development testbed was agreed during which the WP can test themselves and their software in a realistic environment.

This lead to a discussion about how reorganise the release procedure to do it on a WP at-a-time basis and whether multiple development testbed are necessary.
The idea was to keep one project development testbed one which we sequentially add new developments from each WP. We don't move on to another WP until the testing is finished. Then move onto next WP. In parallel, another testbed is required for WP specific work. Markus, JJ, Flavia & Bob (+ others?) will make a proposal in this sense.


Federico's slides added that the ITeam lacks authority and suggestions for how to empower the group.

The meeting was curtailed at 13:00