PPS all-sites meeting minutes

Istanbul, Harbiye Askeri Museum (c/o EGEE08)

14-Sep-08

Attendance

  • CERN_PPS: Antonio Retico (PPS Activity Coordinator, PPS Site Admin, chair)
  • CESGA-PPS: Javier Lopez (PPS-MQS Coordinator); Sergio Diaz (PPS Accounting)
  • INFN-CNAF: Daniele Cesini (PPS site admin)
  • FZK-PPS : Angela Poschlad (PPS site admin)
  • FHG-SCAI: Horst Schwichtenberg, Klare Cassirer (PPS site admins)
  • PIC: Christian Neissner (PPS site admin); Marc Rodriguez (Deployment test)
  • DESY: Andreas Gellrich(representing DESY-PPS)
  • ITWM: Christian Peter (Site representative)
  • STFC-RAL: Ned Geddes (Observer)
  • LIP: Mario David, Hugo Gomes (PPS Site Admin)
  • WCSS: Pawel Dziekonski (PPS Site Admin)
  • UKI-LT2-IC-HEP-PPS
  • CSIC-IFIC: Alvaro Fernandez (PPS Site Admin)
  • CE region, CYFRONET: MArcin Radecki (PPS Coordination in CE ROC)
  • SEE region, Kostas Koumantaros: (SEE ROC Technical Manager)

Introduction and Agenda (by Antonio)

Antonio welcomes the attendants and thanks them for their work during the past year

Clarification on the Agenda (point raised by Ned): Risk of confusion between PPS pilots and pilot jobs for non PPS aware audience

Recent developments: is PPS now used? (by Angela, Antonio)

Antonio goes rapidly through the changes in mandate, use cases, organisation structure and procedures then he describes quickly the status of the glexec activity at CESGA and PPS

Angela talks about her experience within the Cream pilot, which she rates in general very positive (asks for more :-))


External Note: The work done by Angela at FZK within the pilot and the numbers produced where positively remarked in other presentations given at EGEE08 by users (Alice) and middleware developers (Cream)

Some discussions were triggered by Christian N's request to Angela to clarify the nature of the interventions she had to do in production. Basically the only intervention needed was to configure a queue node.

Antonio stressed the fact that the boundaries between pre-production and production, in the new area of the MPS, should be less rigid. The layout of pilots in production is designed according to the test cases in the schedule and the requirements from the users. So the institutes should not fear to dedicate production resources to these activities when needed. If this conflicts with local site policies and the distinction between PPS sites and production at the centers is rigidly imposed, there is always the opportunity for the PPS site to contribute in pre-deployment activities, but PPS site admins and their ROCs should share the idea of a "logical" PPS that may require actions to be done both at production and pre-production levels of the infrastructure.

Activity management: What for? (by Antonio)

Antonio gave a presentation in the form of answer to FAQs about the newly developed activity management tool.

Some comments by people involved in the Cream pilot, the first ones to have received Savannah tasks: filling the Savannah task reports did not introduce a significant overhead for them (actually some said "I liked them"), and the pre-defined effort seemed well balanced.

Antonio pointed out that the pre-defined effort for tasks related to pilot services was defined with the help of Daniele, who has done a lot of administration activity in the framework of the experimental services. He stressed however the opportunity for the owners of the task to correct them and put the actual effort on them. That would help to make more correct report and, more important, better estimates and plans for future pilots.

Deployment test: present and future? (by Javier)

Javier described the current process and the plan for the immediate future. A particular stress was put on the need to extend the list of supported scenarios, where we have some significant holes to cover. We need also to prioritise this list, as the combinatorial approach is not feasible (nor economic) CESGA proposes the creation of a list of "orphan test scenarios". Sites interested in a particular scenario "vote" for it and help to maintain the priorities up-to-date

Also redundancy is important because we cannot delay the tests because of vacation. It would be good to have back-ups for the services

Kostas pointed out that we should study the scenarios and the test we do taking into account also the indications coming from the production infrastructure. He suggests to try and get involved in the tests sites directly interested in a peculiar scenario.

Antonio: In addition to that, we should consider covering, in our scenarios, the case of services which are kept at older versions at sites for specific reason. If a service is not upgraded on specific request of users, a failure in production for that scenario would likely be very visible.

Kostas suggests a policy to cover less frequent scenarios (which most sites have no interest to test). We could assign to sites scenarios to test in couples. One service that the sites can choose and a second mandatory service as back-up, decided by the MQS coordinator.

Finally, we went though the list of services currently not covered by pre-deployment testing. We registered the following commitments:

  • SEE (sites to be defined) for LB and WMS 32bit
  • CERN-PPS for voms and amga (32 bit, oracle versions)
  • DESY for the complex of dcache packages
  • SCAI for AMGA (postgres and mysql)

the list for receiving candidates is pps-support@cernNOSPAMPLEASE.ch

AOB

Antonio thanks Mario David for his contribution in the set-up and operation of the deployment test procedures and announces the appointment of Esteban Freire and Alvaro Fernandez from CESGA in the role of MQS coordinators, with the mandate to extend the deployment test activity and to review, develop and maintain processes and procedures inherent to the service area
Edit | Attach | Watch | Print version | History: r4 < r3 < r2 < r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r4 - 2008-10-06 - AntonioRetico
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    LCG All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback