Choose timezone
Your profile timezone:
IT Piquet Project meeting
→
Europe/Zurich
31/1-012 (CERN)
31/1-012
CERN
Description
More about Service dependencies, LCG MoU, Piquet Service organisation
Present:
Jamie S. , Ludwig P, Maria G., Nick T., Veronique L.
More about the LCG MoU
It is stated in the October version of the LCG MoU, that the numbers given in the tables define "minimum levels of service. They will be reviewed by the operational boards of the WLCG Collaboration".
It has been agreed by the working group that it is now a good moment for reviewing the numbers. One has to take into account reality and take the levels of service currently reached as a basis for defining more realistic target values. The revision of the numbers is a prerequisite for any decision to be taken regarding Piquet Service organisation.
Jamie has reminded that the recent data challenges have shown that a "catch up" scenario works: an interruption of up to 8h in file transfer may be acceptable.
Ludwig said that the current numbers in term of network availability is 98% in average, per month.
Maria said that she is in contact with the Machine people. It is important to know machine operation plans since the required IT support is different during and outside accelerator operation periods.
Maria reported that several persons have shown interest to the piquet project at the last C5 meeting. It has been agreed that the relevant persons will be invited the join the one after the next meeting we have. In particular, critical communication services such as mailservers and telephone, should be covered by a piquet service as well.
Service Review
Ludwig described the service dependency of the Network area. The conclusion is that no extra Piquet is needed, as they already have a 24/7 support line for hardware, and no need for more than that has been shown during the last year.
Maria reminded that DB services are already covered by 2 Piquet Services: 5 developpers on one side, and 6 DBA's on the other (incl. 1 fellow). Maria reminded that the service recovery procedure relies on the back up service to be available.
Nick has produced a detailed table of service dependencies. The FTS and LFC services are the ones for which a Piquet Service is needed. But the piquet covery can be 16h iso 24h a day as problems are usually quickly fixed and, as said before, a ~8h transfer interruption is acceptable.
Veronique said that on the FIO side, a CASTOR piquet is needed. Olof is already doing that on a best effort basis. Several Service Managers are willing to take part to the rota. AFS is also a very critical service, but has shown to be quite reliable on one side, and is lacking of manpower on the other side, so no Piquet coverage for AFS seems possible.
Action List:
--------------
- Jamie: review of LCG MoU table structure, wording and numbers, with input from Ludwig for the Network
- Jamie: Numbers and plots from recent data challenges, demonstrating that a "catch up" scenario is possible in case of transfer problems -> related piquet can be for 16h instead of 24h a day.
- Maria: circulate slides about Machine operation plans
- Maria: establish 2 or 3 scenario of Machine operations(but in fact we finally do not need that, do we ?)
- Nick to complete and circulate his tables
- Maria to complete and circulate her Service List etc
- Veronique to merge Service diagrams and tables
- Veronique to compute the minimal number of persons needed per piquet to satisfy the rule of max 9(12) piquet weeks a year per person, taking holidays and leave compensation into account.
There are minutes attached to this event.
Show them.