News, GPFS, data migration policy, Absorber P07, Albula, backup /gpfs/local

• GPFS:
  • Stolen ESS node number: we are the first to encounter the problem. Depends on the size of the compute cluster. Above 128 the problem is more likely to occur, DESY: 221. So far no fix by IBM.

  • 80 TB miracle: this space is fragmented, needs a de-fragmentation before it becomes available, will be done next week.

  • Full core file system full on Monday: caused by the fragmented space. Snapshots have been deleted. Currently 150 TB free space.

  • GPFS outage Monday, Nov. 6, 9:00 h. Expected downtime 15 minutes. Reason: 2 bug fixed.

  • Planned work on GPFS next week: fix the bugs, de-fragment the core file system, study GL4S.

• Data migration policy:
  • Aim: data not touched for one year can be migrated to dCache. 2 physical tape copies will be created.

  • So far data can be staged by an email to the request tracker. It is planned to automatize the procedure.

  • In case of a severe shortage on disk space, younger files will be migrated.

• Absorber P07: move widget gave an error, DONE

• Albula: solved since long by IT

   
  $ module load xray
  $ albula
  

• Request to backup /gpfs/local: consider to use /bl_documents instead.

  Confluence says:

  bl_documents:
      - Mounted read-write in the beamline space with 1TiB hard quota per beamline
      - only the beamline specific part is mounted
      - stays there independent of the beam times
      - has 2 snapshots daily, 28 (4 weeks) are kept.
      - For scripts and documentation
      - Is not visible from Maxwell and will remain so
  

• Distributed a new version of HasyUtils (SardanaMonitor, AIO, etc.): reason: bug in ssa, extended nxsreader.