Minutes from the first Linux certification coordination meeting on 17.10.02 --------------------------------------------------------------------------- Agenda: * introduction (Wolfgang v.Rüden) * formalities: - name (suggestions?) - chair person, secretary - meeting style (regular vs ad-hoc) - reporting * objectives and scope of the "Linux certification coordination body" * "hot topics" / next agenda Participants (affiliation): Bruce Barnett (ATLAS-online Eric Cano (CMS-online) Marco Cattaneo (LHCb) Günter Duckeck (ATLAS-offline) Benigno Gobbo (non-LHC experiments Jan Iven (IT-services catchall) Nicolas de Metz-Noblat (PS/CO) Bernd Panzer-Steindel (IT/ADC) Jarek Polok (Desktop) Fons Rademakers (ALICE) Wolfgang von Rüden (IT) Tim Smith (IT PLUS/BATCH) Stephan Wynhoff (CMS-offline) excused: Alastair Bland (SL/CO) The meeting started with an introduction from Wolfgang, in which he outlined the reasons for this new body to be created (perceived sluggishness of certifications, CLUG not a representative body). He pointed out the need to collaborate with CLUG (e.g. by attending CLUG meetings and have decision-taking meetings shortly after subjects could be discussed openly in CLUG), as well as the need for long-term commitments for collaboration. Formalities: Name: no conclusive result, left open. Proposals: * LCCC: Linux certification coordination committee: - too many 'CCC's already - easy to confuse with LCG * LXCERT: - CERT is Computer Emergency Response Team (computer security) * LXGO: - no suitable expansion to the acronym found Secretary: J.Polok Chair: J.Iven Meetings: to be announced in advance, preferably for the whole certification in question. For this round (7.3.1), the following dates were set: Thursday, 31.10.02, 10:00 (topic: compiler) Thursday, 14.11.02, 10:00 Reporting: FOCUS for start/status/end of certifications CLUG when possible DTF when subject overlap minutes to be sent to linux-certification@cern.ch Proposed scopes and objectives: * to monitor the certification process: accepted * to implement the process: Some discussion, with general agreement that CERN customization should be minimized in order to benefit from Red Hat certifications and to make live easier for outside institutes. Discussion drifted to compilers and multiple versions of software. Attendants agreed to map their dependencies. Still looking for a tool to help managing these dependencies. * to decide CERN-wide issues: (example: compiler. started discussion, but deferred to next meeting so that the relevant external experts can have a say. Clearly an important topic) Overall agreement on the proposed objectives was achieved. Some details of the discussions: * need for explicit dependencies was pointed out, should allow to certify for a given usage. * Release frequency: 4x per year is too often. Propose to have 1 major release every 12-18month (following vendor release), 1-2 minor per year. * Best time to certify: - before the end of the year (accelerator shutdown) - consensus not easy - certify != immediate implementation. As long as upgrade happens "relatively soon", this is acceptable. Support for old releases is non-negligible, mostly due to security problems. Shared resources (PLUS/BATCH) may need to be upgraded in stages and should be used to heighten pressure for migration (new machines should be running the new version, old version left on slow/few machines) * Announced cut-off dates for old versions are ok, they allow for planned upgrades