Tech Meeting Minutes 20100527

From ScotGrid

Present: Graeme, Mike, Sam, Andrew, Wahid

Indico: http://indico.cern.ch/conferenceDisplay.py?confId=95724

Table of contents

Open Problems

  • LHCb upload problems at Glasgow. Dug has confirmed that the lcg-cr fix for dteam still works, so the NAT problem is "solved" (100 transfers, 98% success; 300 83% success, but likely due to RAL endpoint limitations). Can we try this with an LHCb proxy? ACTION: Ask Vladimir. Have applied for VO membership (Mike). LHCb don't seem to send pilots to sites which are full (for

us full of ATLAS). Maybe chase this later.

  • LHCb problems at Edinburgh. Ticket 57042 finally closed. Andy will still migrate s/w area to nfs (pending ACTION). Software should be wiped and reinstalled from scratch.
  • ATLAS production has restarted slowly at ECDF. ACTION: Andy to keep an eye on this.
  • Sam fixed an ILC authentication problem on the Durham SE, so cross-site login still works.

Infrastructure

  • squid: ECDF fails SAM tests through the Lancaster squid, so there is still a firewall issue. ACTION: Wahid to fix.
  • CREAM: Still waiting for extra servers from systems team. Decomissioning old UI is slower than expected.
  • Storm + GPFS: ACL problem fixed (no attribute caching mount option). Suggested now to move the ECDF SCRATCHDISK to storm to avoid test endpoints in ToA. To do this:
    • Set ANALY_ECDF offline (it's online right now).
    • Move datasets off current SCRATCHDISK to Glasgow SCRATCHDISK.
    • Delete ECDF SCRATCHDISK copies.
    • Change ToA.
    • Move user datasets back.
    • Set site 'brokeroff' to hammercloud it.
    • Set site 'online'.

glexec

  • Graeme sent email to Jose to test this.

Procurements

  • Glasgow is in cooling off period.

AOB

  • Graeme has a dead DSL router at home. He's trying to get to Ferney to get France Telecom to replace it. Home internet down for now!
  • Dug is leaving at the end of this month, so big thanks to him for all his work in the last 18 months. Thanks!