Main Page

From HPC documentation portal
Revision as of 14:00, 12 September 2017 by Alexander Oltu (Talk | contribs) (Current tentative schedule)

Jump to: navigation, search

HPC documentation portal

Welcome to the HPC documentation portal. Here you will find documentation for the high performance machines operated by the Scientific Computing group (Vitenskapelig Databehandling), part of the University of Bergen (UiB). This is also a hub to the machine status, recent events and activities. For the latest updates please visit syslog.hpc.uib.no or follow us on Twitter @UiB_VD.

Decomissioning of Hexagon

DISCLAIMER
From 01.07.2017 hexagon is running without hardware support, this could increase downtime in case of hardware failure.
All dates below are tentatives and further plan changes might come up and will be communicated on syslog.hpc.uib.no and also updated here.
Please don't hesitate to contact us at Support have you any question.

Summary

  • Hexagon will operate with the national, local and IMR CPU accounts until Fram will be fully operational and relevant Notur projects transferred to Fram.
  • There will be a window of 2 weeks after official Hexagon decommissioning date from Notur. This is to allow users to move their data from Hexagon. Please don't panic if something goes wrong during copy process, but just simply contact us at Support.
  • After this 2 weeks window Hexagon will be stopped and reconfigured for local and IMR access only.
  • Reconfiguring is strictly required to continue to run Hexagon without any maintenance agreement, it will also reduce power usage footprint of the system..
  • We expect that the reconfiguration will take approx. 2 weeks.
  • We will consolidate Fimm local workload and the file system with Hexagon. (Fimm will remain as a dedicated CERN ALICE Tier1 resource).
  • A few changes to name after the reconfiguration:
    • Hexagon will shrink by 8 cabinets, roughly by half, to 9968 cores.
    • /work and /work-common will be reformatted:
      • /work will shrink by 1/3 (86TB) and will be at about 172TB
      • /work-common will be reduced to 176TB size
      • /fimm file system will be reformatted to be a part of /work-common file system.
    • /home will be reformatted
    • reduced amount of software available through the modules (mostly old and rarely used software will be removed)

Data

  • All data needed from HOME, /work and /work-common must be moved or copied by user.
  • Please start moving or copying as soon as possible and we recommend you to move all unused data out from Hexagon already now. At least please plan this well in advance so that we avoid file system and network overload.
  • After ceasing job runs on Hexagon, /work and /work-common will be re-mounted read-only. Users will be given 14 days to move out remaining, newly generated data. According to current time schedule, all data laying on /work and /work-common will be deleted during reconfiguration (see above Summary section).

Current tentative schedule

  • 2017-10-01 - No jobs are allowed, all running jobs are terminated. Start of the "data move window period". Hexagon official decommission from Notur. (This date depends on Fram production start)
  • 2017-10-16 - End of "data move window period" and start of Hexagon reconfiguration.
  • DURING THIS PERIOD ALL DATA ON /HOME, /WORK, /WORK-COMMON and /FIMM WILL BE DELETED
  • 2017-10-30 - Hexagon is online for local and IMR users

Syslog recent entries

Failed to load RSS feed from http://syslog.hpc.uib.no/feed/: Error fetching URL: Maximum (0) redirects followed

view all entries

Resource monitor

hexagon.hpc.uib.no

CPU Parallelism
hex_cpu.png hex_para.png

fimm.hpc.uib.no

Graphs are temporarily not available.