Running and testing GRID services with Puppet at GRIF- IRFU

21st International Conference on Computing in High Energy and Nuclear Physics

物理学;计算机科学

Ferry, S.^1 ; Schaer, F.^1 ; Meyer, J.P.^1

CEA Saclay, Gif-Sur-Yvette

91191, France^1

关键词: 10 Gbps;    Configuration management;    File systems;    Grid services;    Multi core;    Private networks;    Suspension system;   

Others  :  https://iopscience.iop.org/article/10.1088/1742-6596/664/5/052013/pdf
DOI  :  10.1088/1742-6596/664/5/052013

学科分类：计算机科学（综合）

来源: IOP

【 摘 要 】

GRIF is a distributed Tiers 2 centre, made of 6 different centres in the Paris region, and serving many VOs. The sub-sites are connected with 10 Gbps private network and share tools for central management. One of the sub-sites, GRIF-IRFU held and maintained in the CEA- Saclay centre, moved a year ago, to a configuration management using Puppet. Thanks to the versatility of Puppet/Foreman automation, the GRIF-IRFU site maintains usual grid services, with, among them: a CREAM-CE with a TORQUE+Maui (running a batch with more than 5000 jobs slots), a DPM storage of more than 2 PB, a Nagios monitoring essentially based on check-mk, as well as centralized services for the French NGI, like the accounting, or the argus central suspension system. We report on the actual functionalities of Puppet and present the last tests and evolutions including a monitoring with Graphite, a HT-condor multicore batch accessed with an ARC-CE and a CEPH storage file system.