Dear all, please see below the notes from our last meeting, Friday, 15 Jan 2010. Best, Johannes Meeting on 2010-01-15, 16:00 (CET) Notes Participants: Emmanouil, Etienne, Steven, Morris, Luigi, Johannes (i) Debrief minutes of last PGI TelCon nail down validation steps in ppt have to write it down and agree on it (ii) Debrief PGI actions Action: Johannes resend the action list (iii) Discuss regular change of telcon timeslot in 2010 instead of Friday afternoon action for Johannes: set up a doodle (iv) Continue discussion animation ppt of PGI operation & state model Focus: Waiting for a job-id until all validation steps have been performed is unrealistic from implementation view (e.g. 100 jobs in submit array) first spec: start with scientific gateway, no job id, no state create JSDL call createActivity shpping JSDL first state submitted maybe we should have valaidation process before most implementation logic parsing of JSDL going through XML, compliance checks GLUE link, optional validation check JSDL seems to be fine -> get job id new state submitted validated ->SOAP response when is submitted hold happening Etienne: submitted outgoing does not mean responding but sending the job to compute facility Morris: maybe change Etienne: because standard name Morris: we have SOAP response PGI benefit> provide job id and staging info staging info might not be there certainly take some time job id gets crucial if we have really an array and submit job with 100 JSDLs not really realistic create activity = 1 Next slide: lot of points but not a full list crucial points questions? no implementation issue if lot of JSDLs three identified approaches A. two operations in port type we really give the different activities a different symantics gives us a job id may take a long time second operation typical array, but does not guarantee next iteration: query unvalidated job ids B. one operation in port type give back job id before validating C. one operation in port type specify as part of JSDL problem related to job directory problem session directory: could be provided but must not be provided query BES container and get a complete list Steven: combinations of approaches WSRM provides a way ensuring getting message from one ep to another Etienne: strongly opposed to C you cannot guarantee that a client is unique enforced at the server level Morris: we should drop Option C. Etienne: yes Steven: client could provide an alias server could create a string which is unique for the server Etienne: server cannot reasonably believe that id is unique Morris: one server could be in workflow should not be allowed to create uuid -> drop C Etienne: B is not acceptable by ARC they want early staging poll for the job status and get session directory afterwards Morris: UNICORE same feature request Etienne: Strongly in favor of option A Morris: Approach A. Etienne: must have slightly differnet name Morris: createActivity createValidatedActivity Steven: many times this was discussed wanted to do we returned job ids without any expectation of validation createValidatedActivity client takes the risk of validating ppt edited - approach A Morris: anything missed? no metting close and need ARC guys for (v) Debrief Steven Newhouse comments on the list (if time) (vi) Debrief Bernd Schuller comments on the list (if time) (vii) Interop demonstration Setups for OGF28 (viii) Next meeting (ix) AOB Etienne: developing a BES client to be able to interoperate with ARC will ask where we are on this subject Emmanouil: monitoring service DMON Ilya Saverchenko could present this Action for Emanouel: GIN list et of DEMOS cll for more demos Action for Emanouel: get in touch with Ilia Next week - if not agreed earlier on the doodle - Friday -- _ _ _ _ _ _ Johannes Watzl |\/| |\ | |\/| Institut für Informatik / Dept. of CS | | | \| | | Ludwig-Maximilians-Universität München ======= TEAM ======= Oettingenstr. 67, 80538 Munich, Germany Room E 005, Phone +49-89-2180-9162 Munich Network Management Team Email: watzl@nm.ifi.lmu.de Münchner Netz-Management Team http://www.nm.ifi.lmu.de/~watzl