Swiss WLCG Operations Meeting on 2011-09-13

• Date and time: Tuesday at 9:30
• Place: EVO, password: chipp
• External link / EVO: http://evo.caltech.edu/evoNext/koala.jnlp?meeting=MsMiMI282BDBD29892Dt9s

Agenda

Status

• CSCS:
  • Update of the recent issues with Lustre and Cream.
  • ARC Accounting and BDII.
  • Status of EMI1 upgrade
  • Migration to GPFS
• PSI (reports Fabio):
  • We suffer almost every week of 1TB failed disks, it takes time to replace the disk / check the ZFS rebuild / apply the Oracle/DHL process and bureaucracy for the new disk.
  • We're investigating a new 200TB raw storage scalable in 600TB raw in 3 years, we've invited several vendors to propose an offer, so far I've discovered these products, subdivided for RAID technology:
    • HW RAID + Scientific Linux + ext4/xfs
      • Nexsan 60 SATA disks 4u. Invited to present on next Tuesday.
      • NetApp E5400, 60 SATA disks 4u. Offer on next Monday.
      • HP, they propose P4500 G2, but by using 2TB disks, I want 3TB.
    • ZFS RAID-Z2 + Solaris
      • SuperMicro, 45 * Constellation ES.2, 3TB SAS2 disks * 4u, ~ 13k CHF from Happyware, but backplane is a single point of failure + not hot swappable SAS IO module, so it's cheap but present potential risks, Users has to accept them.
      • DataOn, 60 * Constellation ES.2, 3TB SAS2 disks * 4u: this one is interesting because can be installed with its server head with OpenIndiana + Multipath MPxIO driver + 2 independent dual SAS HBA LSI 9200-8e + ZFS. Attached by SAS to the storage we can use the DataOn server head. The JBOD full of 3TB disks is ~ 33k CHF from ZSTOR.
  • So far I've a preference for the DataOn solution but there are potential high risk because of the usage of OpenIndiana, I don't have operational experiences to report: conversely the DataOn product itself seems well accepted from the community. Officially DataOn comes with NexentaStor.
• UNIBE (reports Gianfranco):
  • Stable operations, ATLAS production and analysis. Also some SMSCG activity
  • Running stably with CVMFS since a few months now
  • Had to recover DPM install because of EPEL conflict (operations not affected)
  • Upgrade to ARC1 postponed due to Lustre bugs
  • Starting now with KVM
• UNIGE (reports Szymon):
  • Daily operations very stable and quiet, nothing to report.
  • A major upgrade of storage going ahead:
    • IBM x3630 M3 with 12 x 3 TB of disk for data + two small disks for the OS
    • hardware RAID, we will do RAID 6 with cold spares, 27.5 TB net/machine
    • we are getting 10 machines, only 8 can be powered up soon
    • to stay in fixed kW budget we will retire 12 login machines, in operation since 2005
    • most of the new disk servers will be added to our Storage Element

• CernVMFS updates (the requirements meeting is a different meeting)
• Convenience of the actual Date/Time for this meetings

Attendants

• CSCS: Pablo, Miguel
• CMS: Fabio
• ATLAS: Gianfranco
• LHCb:Roland
• EGI:

Action items

• Item1