TMB Meeting

Wednesday 3 Sept 2008, 10:30 → 13:55 Europe/Zurich

600/R-002 (CERN)

Oliver Keeble (CERN)

A phone conference has been set up:
To dial in to the conference call:
Call: +41227676000 (Main)
Enter access code: 0170047.

To join the web conference, click here: https://audioconf.cern.ch/call/0170047.

TMB 03/09/08 Telephone: Jeff, Cal, Andrei, Stefano, Vangelis CERN Room: Oliver, Francesco, John, Nick Task list. ---------- (etc etc)... various tasks I report numbers only. OK: Task 6652 Nick will take care of for next time. OK: Task 6823: Multiplatform support... no feedback from last week. We assume that the plan as presented is implicitly agreed. Daniele had circulated to ROCs the list. No feedback. Considered accepted. NT: Agreed. (SA1). CL: What is the status of the milestone document associated? (Underway). OK: Task 6901. Storage Semantics. VF: Nothing to report. Other than that on the task. OK: Task 7106. Anything Nick? NT: Nothing... Steve T. on vacation until last week. OK: R-GMA. Skip until later. OK: Send use-cases to Grid Observatory. (Number?) (Tasks 7188 and 7187 JW) FG: No news on this? Have to check with Erwin if this is for JRA1 or not? OK: Erwin is going to be away for a long time, at least until next TCG meeting. NT: No update as well on Task 7187. OK: Task 7281. On the agenda today. Discussed later today OK: Task 7674? NT: If a site does not support the approved CAs then SAM fails and they go down "in availability" OK: Why was this brought up? CL: Came up because someone is rejecting another national CA. NT: Site will fail the SAM test. OK: Then it fails the EGEE policy. CL: Trouble is the software fails as the CA is there but invalid. The policy says SHOULD and not MUST. OK: Task 7675. Method for transition to CREAM... End result is the obsoletion of the LCG-CE. NT: We are still working on these prerequisites. OK: Task 7676. Patricia. (After EGEE08) JT: (General comments) What is the minimum a group could do to install a dashboard? OK: Task 7677. Erwin/Francesco. FG: No news. OK: Task 7678. JW: Ongoing. The SCG will contact the dCache developers to see if progress has been made. May take advantage of the EGEE08 conference. OK: Any other comments? (None... no tasks to be entered or assigned) Obsoletion of Services ---------------------- OK: We are "good at adding and bad at subtracting". What we need is a long-term roadmap to remove services. - The gLite 3.0 "everything" ie. All the services that are gLite 3.0. - How this is propagated and enforced. CL: From NA4 point of view. YThis needs to be announced and centrally explained. OK: From SA# point of view: For SLC3 to SLC4... if there is month of proper SLC4 operation then the SLC3 is obsolete. OK: For WMS... the meta-rpm would put some files on disk to explain the versioning etc. The gLite version command would explain the rest. Could be fed into an info system. NT: Being able to query the version of a service would be very interesting from ops point of view. OK: We need something that is more up to date than the milestone is somewhere on the "glite.org" website. JT: We are having this discussion because we don't have a proposal to push groups onto new services. OK: Is the RB retired? JT: Nope. Still have plenty around. OK: The 3.1 WMS has not yet shown the month of good service. Patch still on the way. JT: Has this type of migration been negotiated? CL: SLC3 to SLC4 is not such a problem. But moving from RB to WMS should need 6 months. OK: We need to discuss this at this (TCG) level for retiring services. JT: Also what about for interface change? OK: That's effectively a new service. FG: Sometimes you have to break the backwards-compatibility but it is not a new service. JT: Debatable point... OK: SA3 will put this into something more coherent. NT: Need a list of supported services and their versions. We need a clear list. CREAM ----- OK: Present status... we are about to release a version to Production. For sites ready and interested to install in parallel. Needs direct submission from the CLI as no WMS ready. There is an issue with the proxy renewal. Short executive summary. Proxyrenewal invokes BPR service that listens for incoming connections. Need to open ports on WNs for incoming connections. - What are the operatonal and deployment problems? JW: Ports 20000:25000 are the required range BUT these are only accessible from the site CE. JT: Let me get back to you guys on this... (Discussion on CREAM and renewal... how does CREAM make sure the correct proxy is used) FG: There is a check for grid job ID. CL: Who kills this process?? Needs to be an active clean-up. (Unknown) OK: So this is an implementation point... any objections from the sites side? (No other site reps present) NT: Any other site representative? JT: Ask the Lyon site, they usually have opinions. Ask Steve Traylen who also knows. Pilot Jobs ---------- OK: Back up solution for SCAS. Only other "game" is GUMS. LCMAPS plugin exists in a way. OK: Don't make investments yet in investigating GUMS. Next part: Task 7281... "schedule a full discussion on pilot job framework in TMB" Question: What is the project responsibility to supporting pilot job frameworks? JT: Documentation on best practices... could be the best idea. Multi-user (MU) frameworks need to be explained in simple terms. Users can be pointed to an approved framework. Is anyone going to come for a new MU pilot job framework? CL: Outside HEP, they mostly use the personal pilot job framework. OK: What about the "part of RESPECT program"? DIRAC is advertised as not LHC-b specific? I know of no other pilot job framework that claims to be generic. CL: Fine for RESPECT. It has to be generic and have more than one user. AT: This is a fair enough point for DIRAC. OK: As far as core gLite stuff is concerned, security system is the limit for MU pilot jobs. Personal pilot job systems are not really on the radar. AT: Dissemination of MU pilot jobs needs to be done. AOB --- Next Meeting: tasks to be reported ---------------------------------- Task 6652 (SA1) Task 7106 (SA1) Task 7188 and 7187 (Need further clarification?) (JRA1 and SA1)

10:30 → 10:50 Minutes of last meeting and action list

list of actions https://savannah.cern.ch/task/?group=tcg

10:50 → 11:10 Retiring services and components

In order to avoid large populations of old services we need to better manage 'end-of-life' for old versions of existing services, or even for entire services (such as the lcg-RB). This raises certain questions; * how do we identify the versions of services in production * how do we decide what needs to go (ie the policy) * how do we publish the decision * how do we enforce the decision

11:10 → 11:30 CREAM proxy renewal

Implications of the BPRserver on the WN and possible solutions

11:30 → 11:50 Pilot Jobs

* what are the backup plans in case of problems with SCAS. Discussion on GUMS. * we should look into generic solutions we could offer (maybe as part of the respect program) applications that don't have resources to build their own pilot job framework.

11:50 → 12:00 AOB