Tier1 and Tier-2 Weekly Report 2015-05-26

Tier1 and Tier-2 Weekly Report 2015-05-26

General News

ADC and ATLAS Computing News [201500526]
- ADC Weekly meeting
- Rucio Update
  - Ran stress tests May 22
    - Simulation of 1 day of data taking (66TB, T0 export to 3xT1, 1xT2)
    - T1 to T2 replication of mc15 AOD (170TB, 1xT2)
    - All went well, the transfers finished in 24h, except long tail to some T2s
  - Data replication policy for 2015 data decided: https://twiki.cern.ch/twiki/bin/view/Atlas/ReplicationPolicy2015
    - data15_13TeV: T2s are selected from the following criteria: more than 400TB capacity in DATADISK and more than 90% availability for analysis in the last 3 months
  - Rucio pre-production testbed is being set up
    - New releases will go there and take small percentage of load for one day
    - If all ok then deploy on production nodes
  - Lost files from FTS double submission
    - Investigation into the impact of the problem reported 2 months ago has finished
    - Only ~2000 files were lost as a result
- HLT Reprocessing for Run 2
  - Preparing for week 1 starting this week (see slides for prepartaion and testing details)
  - Normal HLT Reprocessing requirements:
    - Purpose: release validation, weekly cycle
    - HLT+reconstruction+merging
      - Looking for subtle differences between histograms so no failures
      - Large sample to find rare bugs/crashes
    - Data set: Enhanced Bias: e.g. data12_8TeV.00212967.physics_EnhancedBias 1773 files 1M events
      - Often add reprocessing of signal skims to check efficiency
    - HLT mem 3.7GB, reco mem much higher (still investigating if there is a bug)
    - Results on group space at CERN (CERN-PROD_TRIG-HLT)
    - Aim for 24h turnaroun
  - Start up time depends on general CERN and Grid status, data replicas and memory requirements
  - Highest priority other than exceptional tasks
  - Run where data is:
    - CERN for data just taken
    - Tier1 sites once they have replicas (we can make more for hot datasets)
    - Memory requirements
      - HLT only < 3700MB can run at any Tier1; 7-24 hours
      - HLT+Reco requires HI-MEM queues, limited slots can be a bottleneck; 36-60 hours
- CRC Report
  - Not much to report - fairly smooth shift
CA Cloud Squad Actions
- User DaTRI requests approved / processed (last week)
Open Issues and Ongoing Discussions
- Availability & Reliability Report [20150519]
  - April 2015 - All Sites Report ATLAS
- Sites should be providing the 2015 storage pledge.
  - REBUS Site Capacities
  - REBUS Federation Resource Pledges
- ATLAS Canada HammerCloud Tests
WLCG and General Computing News
- Security News
- gLite/EMI News
ROC_Canada News
- Current shift