Go to the Indico Home Page Oldest event Older event Up to category Newer event Newest event Printable version
 
Focus on:  Hide Contributions
  • Compact style
  • Indico style
  • Indico style - inline minutes
  • Indico style - numbered
  • Indico style - numbered + minutes
  • Indico Weeks View
    • Switch to the management area of this event
    Login

    RAL Tier1 Experiments Liaison Meeting

    → Europe/London
    Access Grid (RAL R89)

    Access Grid

    RAL R89
      • 12:38 → 12:39
        Major Incidents Changes 1m
      • 12:39 → 12:40
        Summary of Operational Status and Issues 1m
        Speakers: Brian Davies (Lancaster University (GB)), Darren Moore (Science and Technology Facilities Council STFC (GB))
        RT1EL-13052020-2.docx
        RT1EL-13052020-2.pdf
        RT1EL-13052020-2.pdf
      • 12:40 → 12:41
        GGUS /RT Tickets 1m

        https://tinyurl.com/T1-GGUS-Open
        https://tinyurl.com/T1-GGUS-Closed

      • 12:41 → 12:42
        Site Availability 1m

        https://lcgwww.gridpp.rl.ac.uk/utils/availchart/

        https://cms-site-readiness.web.cern.ch/cms-site-readiness/SiteReadiness/HTML/SiteReadinessReport.html#T1_UK_RAL

        http://hammercloud.cern.ch/hc/app/atlas/siteoverview/?site=RAL-LCG2&startTime=2020-01-29&endTime=2020-02-06&templateType=isGolden

      • 12:42 → 12:43
        Experiment Operational Issues 1m
      • 12:44 → 12:45
        VO-Liaison ATLAS 1m
        Speakers: James William Walder (Science and Technology Facilities Council STFC (GB)), Dr Tim Adye (Science and Technology Facilities Council STFC (GB))

        * DirectIO tests:
        PR should fix Singularity problem: https://github.com/stfc/ral-htcondor-tools/pull/9
        How to get into production?

        * RAL-LCG2_MCORE queue set offline by Atlas (RW) (GU queue RAL remains):
        Internal discussion ongoing to a valid strategy on optimisation.
        I.e. User analysis is highest global share; RAL typically runs a lower fraction, and reassigns the jobs if not started quickly enough.

        * Hyper-threading:
        UK T2s; aside from Lancaster (and some old machines) HT is generally enabled at sites.

        * Hammercloud:
        Typically run 350 jobs/day; ~ 13% on new (SSD) WNs;
        Last week max of 2 failures a day; 4 overall.
        (not aware of other good "public" metrics).

        * CPU / Job Walltime 
         - Stage-in (-out) time can contribute significantly to overall wall time of jobs:
        (overall,  payload is ~ 80% of wall time),

        - CPU time is a well defined number (?), and 'correctly' reported by ATLAS and Internal tools (Although one needs to account for the number of cores used).
        - Walltime is a well defined number, and 'correctly' reported by ATLAS and Internal tools
        - CPU efficiency or Job-walltime ratio is not well defined:
            - Depends on hyper-threading,  node utilisation, (e.g. overcommit model).
        - Last piece of the calculation to determine is the "number of CPUs", (and differences between number of CPUs allocated vs used) which might be less trivial than initially appears.

        e.g. for COVID jobs payload is ~ 100% of wall time; but reports <job wall time ratio> ~ 60%

         

         

         

         

         

         



         

      • 12:46 → 12:47
        VO Liaison CMS 1m
        Speaker: Katy Ellis (Science and Technology Facilities Council STFC (GB))

        AAA - Throttling on ceph-gw11 was changed from 200 to 300 nearly 2 weeks ago. I have seen higher demand on the service in the last 24 hours, but the data rate has not changed much. It is still around 500 MB/s per gateway.

        Job input data has been rather high, in a sustained way this week. 

        Job efficiency in the CMS Job Monitoring was updated in the past week. The calculation now takes into account the job length, whereas before this information was lost. 

        Amount of data transfers is currently low. Not aware of any problem. 

        Have been discussing the number of cores used by CMS at RAL T1 with James A.

      • 12:48 → 12:49
        VO Liaison LHCb 1m
        Speaker: Raja Nandakumar (Science and Technology Facilities Council STFC (GB))

        Operations:

        LHCb
        1. ECHO access issue
          • LHCb test submission broke last Thursday. Back since Monday.
          • RAL WNs dedicated to LHCb testing still being fixed to take bulk of the jobs.
          • ECHO / CEPH nautilus upgrade this week. Wait to see if this affects anything (it is among the list of things to check / test)
        2. Job failures overnight
          • Change of "login script" by LHCb yesterday broke singularity for jobs needing sl5
          • Affected MC simulation for Run I
          • Mainly affected sites on singularity 2

        3. Upcoming LHCb VO-box updates to sl7

        DUNE 
        1. Low level of running jobs (currently 9 dunepro jobs on the grid)
          • Up to 400 running jobs at RAL last week.
        2. Following up on SAM tests at different sites
          • NIKHEF fixed earlier today to increase count of good sites by 1

        Question about Hyperthreading at LHCb Tier-1s

        1. Not possible to query DIRAC for information
        2. Individual site liaisons / representatives quite helpful
        3. All Tier-1s seem to use hyperthreading.
        4. Scale factors between 1.3xcores (IN2P3) to 1.5xcores (PIC, GridKa)  2xcores (RRC-KI)
      • 12:52 → 12:53
        VO Liaison Others 1m
      • 12:53 → 12:54
        Experiment Planning 1m
      • 12:54 → 12:55
        Dune/protoDune 1m
      • 12:55 → 12:56
        Euclid 1m
      • 12:56 → 12:57
        SKA 1m
      • 12:57 → 12:58
        AOB 1m
      • 12:58 → 12:59
        Any other Business 1m
        Speakers: Brian Davies (Lancaster University (GB)), Darren Moore (Science and Technology Facilities Council STFC (GB))