eulake technical coordination

Europe/Zurich

Participants: Daniele Cesini (CNAGF), Antonio Falabella (CNAF), Luca dell'Agnello (CNAF), Wybrand Lohman (SARA), Jean-Marie de Boer (SARA), Onno Zweers (SARA), Brian Davis (RAL), Enrico Bocchi (CERN), Xavier Espinal (CERN)

Checkpoint meetings scope

Discussion about the scope of this meeting. The intention is to keep the meeting short, if possible within 30 min. The proposed structure is: 

  1. Site 5min report (tour-de-table): running services, current resources, short term plans (pre-fill as minutes in advance if possible)
  2. Issues: benefit form join discussion of common problems/showstoppers in this early phase
  3. Planning: short term objectives and plans

This has been agreed by everyone.

Tour-de-table

CERN:

  • EULAKE (eulake.cern.ch) central services are ready and functioning:
    • Central node (MGM) with namespace running on KV-Store (Frontend in Wigner, Backend in Meyrin)
    • 3 Storage nodes (2 in Wigner, 1 in Meyrin) 500TB, 120 filesystems
    • 2x Gridftp doors (eulakeftp.cern.ch) ready and tested
    • Namespace has been pre-configured with directory settings for different RAIN layouts (4+2, 8+3, ...) and FILE replicas (1, 2, 3)
    • 2xDubna nodes has been connected to eulake.cern.ch and successfully tested
    • 3xNL-Sara nodes connected to eulake.cern.ch
    • Basic monitoring is ready:  grafana-eulake this is a generic monitoring and need to add relevant metrics/plots for the datalake (put/get timings, network latencies, etc.)
    • Set of automated basic tests ready but not yet running. The idea behind is to have a well known benchmark scripts to do some operations in a cycle permanently (put/get/remove, tar, untar, ping, ls,etc.)
  • Container based deployment of EOS FST is progressing well:
    • Successfully deployed on Dubna
    • Ongoing work at RAL (missing firewall openings and DNS reverse lookup workaround) 
  • RPM based (standard) deployment of EOS FST:
    • Successfully done at SARA by Jean-Marie and Wybrand.

NL-SARA:

  • Jean-Marie reported that three nodes has been registered to eulake, missing firewall opening to enable the filesystems and start firsts tests. 
  • Next week there is a plan to add a dcache pool, which open interesting paths on the prototype as we need to interface with a site specific storage system. It has been agreed to explore the two possible configurations: NFS mounted pool and xrootd native doors.
  • Onno expressed the interested in knowing the interface between EOS and tape storage at CERN.
    • CTA (CERN Tape Archive) is the new tape storage system which is being developed and prototyped at the moment. Can invite CTA expert in one of our meetings to get news on the project and expected deployment plans. 

IT-CNAF:

  • Luca reported that CNAF is evaluating new storage systems and few resources available at the moment for datalakes
  • Report from Daniele about XDC (eXtreme Data Clouds) project and possible synergies between XDC and datalakes 

UK-RAL:

  • The storage resources for the datalake will be based on a volume with a CEPH backend exposed through an Openstack VM
  • DNS reverse lookup not possible which is a priori needed by xrootd (and by the FST daemon by extension). Enrico reported EOS and xrootd developers are having a look at this limitation.
  • Working with the Openstack manager to get the firewall opening.

Next meeting will be on the 13 of March at 16:00

There are minutes attached to this event. Show them.
    • 16:00 16:10
      Checkpoint meetings scope 10m

      The idea is to keep the meeting short, if possible within 30 min.

      The proposed structure is: 

      1. Site 5min report (tour-de-table): running services, current resources, short term plans (pre-fill as minutes in advance if possible)
      2. Issues: benefit form join discussion of common problems/showstoppers in this early phase
      3. Planning: short term objectives and plans
    • 16:10 16:30
      Tour-de-table 20m

      Participants: Daniele Cesini (CNAGF), Antonio Falabella (CNAF), Luca dell'Agnello (CNAF), Wybrand Lohman (SARA), Jean-Marie de Boer (SARA), Onno Zweers (SARA), Brian Davis (RAL), Enrico Bocchi (CERN), Xavier Espinal (CERN)

      CERN:

      • EULAKE (eulake.cern.ch) central services are ready and functioning:
        • Central node (MGM) with namespace running on KV-Store (Frontend in Wigner, Backend in Meyrin)
        • 3 Storage nodes (2 in Wigner, 1 in Meyrin) 500TB, 120 filesystems
        • 2x Gridftp doors (eulakeftp.cern.ch) ready and tested
        • Namespace has been pre-configured with directory settings for different RAIN layouts (4+2, 8+3, ...) and FILE replicas (1, 2, 3)
        • 2xDubna nodes has been connected to eulake.cern.ch and successfully tested
        • 3xNL-Sara nodes connected to eulake.cern.ch
        • Basic monitoring is ready:  grafana-eulake this is a generic monitoring and need to add relevant metrics/plots for the datalake (put/get timings, network latencies, etc.)
        • Set of automated basic tests ready but not yet running. The idea behind is to have a well known benchmark scripts to do some operations in a cycle permanently (put/get/remove, tar, untar, ping, ls,etc.)
      • Container based deployment of EOS FST is progressing well:
        • Successfully deployed on Dubna
        • Ongoing work at RAL (missing firewall openings and DNS reverse lookup workaround) 
      • RPM based (standard) deployment of EOS FST:
        • Successfully done at SARA by Jean-Marie and Wybrand.

      NL-SARA:

      • Jean-Marie reported that three nodes has been registered to eulake, missing firewall opening to enable the filesystems and start firsts tests. 
      • Next week there is a plan to add a dcache pool, which open interesting paths on the prototype as we need to interface with a site specific storage system. It has been agreed to explore the two possible configurations: NFS mounted pool and xrootd native doors.
      • Onno expressed the interested in knowing the interface between EOS and tape storage at CERN.
        • CTA (CERN Tape Archive) is the new tape storage system which is being developed and prototyped at the moment. Can invite CTA expert in one of our meetings to get news on the project and expected deployment plans. 

      NL-NIKHEF:

      IT-CNAF:

      • Luca reported that CNAF is evaluating new storage systems and few resources available at the moment for datalakes
      • Report from Daniele about XDC (eXtreme Data Clouds) project and possible synergies between XDC and datalakes 

      UK-RAL:

      • The storage resources for the datalake will be based on a volume with a CEPH backend exposed through an Openstack VM
      • DNS reverse lookup not possible which is a priori needed by xrootd (and by the FST daemon by extension). Enrico reported EOS and xrootd developers are having a look at this limitation.
      • Working with the Openstack manager to get the firewall opening.

      RU-Dubna: not present

      RU-Kurchatov: not present