Participants: Martin Savko, Mattias Ellert, Cristina del Cano Novales, Valery Tschopp, Andrea Ceccanti, Owen Synge, Christian Bernardt, Paolo Andreeto, Zdenek Sustr, Bernd Schuller, Vincenzo Ciaschini, Giuseppe Fiameni, Eva Cernakova, Jozef Cernak, Lorenzo Dini, Cristina Aiftimiei Cristina: - Service Reference Card – UNICORE provided SRCs for their components - ETICS-configurations: o Last week the import script was run on the production instance of ETICS, because it would have taken to much to do it on test-instance and then on the prod-one. Lorenzo: - no feed-back received - configurations have been copied/clonned Cristina: - one news – the etics-component naming convention has changed (from emi.component to emi.subsystem.component), because almost all PTs used the this old convention; the Guidelines were updated Paolo: - it’s not enough to change the component-names; also the properties published by that component should be changed; the JobManagement PT has now to update all their configurations. Owen: - are the dependencies suppose to be missing after the migration? Lorenzo: - yes: vdt, globus, some others that are not supposed to be used in EMI. Cristina: - if PT discover missing dependencies or other errors, this should be notified to be corrected - some components were not migrated because we didn’t know who should be supporting them, or if they should be assigned to certain PTs Lorenzo: - glite-info-templates, glite-info-generic – dependencies on this components were removed, because there is no maintainer and we didn’t know to whom assign them. Cristina: - on Paolo request – discussion about org.glite.service-discovery Paolo: - asked WMS developers if the dependencies on this components could be removed, but the answer was that they are needed, so it should be decided who will maintain this subsystem. If no other PT is requiring this components, the JobManagement PT will have to maintain them. Lorenzo: - looking at a glite build – this components are used by – FTS, FTM, data.utils-c Pedro: - in the past were maintained by DM people, but they are not more present. He will try to clarify who was maintaining them and if those persons are still available Action on Pedro - to get back with more news about this. Pedro: - glite-info-templates will not be provided by the InformatonSystem PT, not even glite-info-generic Cristina/Owen – a lot of components/all storage components are using them Cristina: - the glite-info-templates, glite-info-generic were added in the emi.misc subsystem with the last versions available in production. SIDE-NOTE1 (at the end of the minutes) Francesco: - this are run-time dependencies, we should concentrate on the build-time dependencies Owen: - how can the metapackages be identified ? AndreaC: - the names are different, they have only dependencies; it’s a way of packaging together, can have dependencies on components from other subsystems Owen: - they had all the metapakages under the same tree, that wasn’t migrated. Francesco: - the metapackages were moved under the subsystem Cristina: - it’s ready the EMI project-configuration, emi_0, where there are defined as Properties, all the external-dependencies - a test-build was done on Friday (22.Oct), using this configuration. - each PT should communicate the configurations to be used to build the project. There were added all the ones found, but there are still a lot of them using _HEAD. Action on PTs – communicate the configuration-names to be used - Work going on – to provide the properties missing, reported as “Unresolved” in the first build Francesco: - are there nightly-builds enabled? Lorenzo: – no, at the moment, it will be enabled (SIDE-NOTE2) – the list of rpm installed on the build-machine was updated, it contains now also the rpms-versions – Cristina has to check the versions present there and update the Properties of emi_0 to match them; if the Properties are not complete, the checkout step of the build is not completed and the build is not started. – All the STATIC dependencies were transformed in DYNAMIC dependencies, but some subsystems, like VOMS, were done manually, so they kept the dependencies on vdt & globus Vincenzo: - he kept the static dependencies, because at the moment when he did his components, there was’t available the project-configuration Lorenzo: - will do a script that goes through all the configurations attached to emi_0, and verify if there are properties that needs to be defined, like location, .DEFAULT - a report with the results this will be available Francesco: - packages names should not have changed! AndreaC: - VOMS package-names have to be changed to become EPEL-compliant Vincenzo: - VOMS package-names have changes, but the configurations names remained the same Cristina: - in general, the intention of the EMI project is not to change the packages names prefix, from “glite”, “arc” or else with “emi”, but to support the packages that are now in production, and improve them, making them compliant with EPEL/Fedors other OS guidelines, but not by changing a prefix in the rpm name. ACTION on PTs – to check their configuration and communicate their comments AOB: Owen: - they need documentation or change strategy for libraries once they get their libraries successfully build in etics. What’s the general opinion about this? What to do when there’s a need to upgrade for ex. ant or mysql? - They had problems for ex with the mysql in gLite, because there were 4 versions of mysql and no strategy on synchronizing them, so many different things were deployed with different versions of mysql. There are new versions of MySQL coming out, there are reasons why people want to stay with older versions. Libraries do change, they improve over time, how do we manage this form of change? Francesco: - distinguish from what comes from the OS and from other sources: what comes from the OS is frozen, we take what comes from there and with the versions available there. In the case of mysql, this is coming from the OS, so there’s nothing to discuss. What is not coming from the OS we have to agree, and to maintain it ourselves and distribute it Andrea: - what about when the OS provides more than one version? Francesco: - in this case this are different packages, they have a name that has the same prefix, but are different things so one can install both of them Owen: - a release of a software will be hold-back for a build-dependency that wasn’t in the OS? Francesco: - build-deps are a special case, can be treated slightly different than the runtime dependencies Owen: - need to discuss build-dependency on ant in the OS, within the next months - it is an important discussion that needs to be done in the EMT, with all the people present Agreement that a discussion about the dependency-update strategy will take place during the AHM in Prague, on the SA1 session, Tuesday, 24.Nov, 8:30-10:15 Andrea: - argus build for emi_0 – some of them fail, complaining on missing maven, so not all the WNs have maven – to be communicated to Lorenzo & AlbertoResco Cristina: - Monday 1.Nov, holiday in Italy, probably the meeting will be postponed; an e-mail with details wil be send on Friday, 29.Oct SIDE-NOTE1: - after the meeting was discovered that glite-info-generic and glite-info-templates are not required anymore (https://tomtools.cern.ch/confluence/display/IS/ResourceBDII), so this components will be removed from emi.misc. SIDE-NOTE2: - link to the nightly builds: http://etics-repository.cern.ch/repository/reports/name/emi_0/-/reports/index.html