Daily Bulletin Archive

Oct. 16, 2017

No downtime: Yellowstone, Cheyenne, Geyser_Caldera, HPSS, GLADE

Oct. 14, 2017

Beginning Saturday, October 14th @ 06:00, the NCAR Colocation Data Center will be shut down to facilitate standard cooling plant maintenance. Operations begin @ 06:00 and are scheduled to end no later than 18:00.

The majority of computational operations in the Mesa Lab data center will be impacted for the duration of this Saturday outage. There are some special projects with regard to the fiscal year end process that will be allowed to stay online during this event and arrangements have been made to maintain infrastructure for these systems. Program system administration teams have been provided details on the work taking place. If you have questions regarding impacts to your program, check with your system administration team.

If you require additional information send your questions to: cislhelp@ucar.edu.

CISL Service Outage Information:

  • NETS MLDC: NETS will be moving fiber infrastructure in preparation for the MLDC project.  This will affect ml-mr-c2-gs, ml-mr-j5-es, ml-mr-c10-gs, and ml-4e-c1-gs and other fiber connections in these areas.  We will also walk redundant uplink connections for ml-mr-j2-es and ml-mr-j3-es.
    NETS will also be working with TAC to troubleshoot issues with ml-mr-j5-es.

  • Guest wireless authentication will be unavailable.

  • UCAS Password Changes via web: Changing UCAS passwords will be unavailable via the web at https://kpasswd.ucar.edu/ but will remain available on Kerberized workstations. UCAS password authentication, Role Principal initiation, and Role Principal authentication will remain in service.

  • FTP: The UCAR FTP server, ftp.ucar.edu, will be unavailable.

  • UCAR web services: jira, drupal sites and wiki.ucar.edu will be down.

  • ExtraView (CISL helpdesk ticketing system) will be down. - Please contact CASG at x4300 or 307-996-4300.

  • Sentry Power Manager (SPM): Will remain online.

Oct. 14, 2017

The Cheyenne cluster will be unavailable from 8 a.m. to 6 p.m. MDT tomorrow, Saturday, October 14, to allow CISL staff to perform important system tests.

Users will be unable to log in during the testing period. A system reservation will be put in place on batch jobs this evening. After the system reservation is in place, jobs submitted with specified job times that overlap the reservation period will remain on hold until the system is restored to service. Running jobs that have not finished when maintenance begins will be killed and will need to be resubmitted after the maintenance period.

We apologize for any inconvenience this might cause. Users will be informed via the CISL Notifier service when the systems are returned to production.

Oct. 12, 2017

Cheyenne’s economy queue has been enabled for a limited time to encourage greater use of the system and to make users’ core-hour allocations go farther. Jobs using the economy queue are charged a discounted rate of 30% compared to the regular queue.

Jobs submitted to the regular and economy queues have the same priority due to the current PBS Pro configuration, so all are encouraged to use the economy queue while it is available. Its availability will be re-evaluated as Cheyenne’s usage increases.

Check your Systems Accounting Manager (SAM )reports regularly to keep an eye on core-hours charges.

Oct. 12, 2017

NCAR postdoctoral researcher Alessandro Fanfarillo will present “Myths and reality of communication/computation overlap in MPI applications” at 3 p.m. MDT on Thursday, October 12, in the Mesa Lab Main Seminar Room.

Overlapping communication with computation will be more important than ever on future exascale architectures, Fanfarillo says, because of high data transfer and synchronization costs and the need to scale applications on billions of cores. In this seminar, he will explain the communication/computation overlap in MPI in theory and practice with multiple examples involving non-blocking and one-sided MPI routines. His research focuses on how to exploit heterogeneous architectures (CPU+Accelerators) and Partitioned Global Address Space (PGAS) languages (coarray Fortran, in particular) for scientific purposes. He is also the lead developer of OpenCoarrays, the open-source library that implements the coarray support in the GNU Fortran compiler.

The seminar will be webcast here. See the UCAR Software Engineering Assembly site for more information.

Oct. 9, 2017

No Downtime: Yellowstone, Cheyenne, HPSS, Geyser_Caldera, GLADE

Oct. 6, 2017

CISL supercomputer users who would like assistance with visualizing their data can ask for help from staff experts. This service is available to researchers using specialized visualization applications on the Geyser or Caldera clusters.

Staff members have helped science teams produce visualizations featured in such venues as the PEARC17 Conference (Practice & Experience in Advanced Research Computing) Visualization Showcase, the Supercomputing Conference Visualization Showcase, American Geophysical Union and American Meteorological Society events, the SIGGRAPH Electronic Theatre, and many others.  

See Advanced visualization support to learn how to submit a request, what criteria must be met, and other details.

Oct. 3, 2017

Cheyenne: Downtime Tuesday, October 3, 8:00 am - 12:00 am

HPSS: Downtime Tuesday, October 3, 7:30 am - 9:30 am

No Downtime: Yellowstone, Geyser_Caldera, GLADE

Oct. 3, 2017

The Cheyenne cluster will be unavailable from 8 a.m. to midnight MDT on Tuesday, October 3, to allow CISL staff to perform maintenance on several critical system components.

A system reservation will be put in place on batch jobs Monday afternoon. Jobs submitted with specified job times that overlap the reservation period will remain on hold until the system is restored to service. Running jobs that have not finished when maintenance begins will be killed and will need to be resubmitted after the maintenance period.

Users will be informed via the CISL Notifier service when the systems are returned to production.

Sep. 27, 2017

Allinea Software will hold a free webinar on Thursday, October 5, dealing with memory-bound code in high-performance and scientific applications. Attendees will see how the right checks and ongoing tuning can improve software run-times and identify bottlenecks that can slow down applications.  Click here to register or to find out more.

Pages