FAS Research Computing - Legg merke til historikk

Opplever delvis svekket ytelse

Status page for the Harvard FAS Research Computing cluster and other resources.

Cluster Utilization (VPN and FASRC login required): Cannon | FASSE


Please scroll down to see details on any Incidents or maintenance notices.
Monthly maintenance occurs on the first Monday of the month (except holidays).

GETTING HELP
Documentation: https://docs.rc.fas.harvard.edu | Account Portal https://portal.rc.fas.harvard.edu
Email: rchelp@rc.fas.harvard.edu | Support Hours

The colors shown in the bars below were chosen to increase visibility for color-blind visitors.
For higher contrast, switch to light mode at the bottom of this page if the background is dark and colors are muted.
Nedsatt ytelse

SLURM Scheduler - Cannon - Nedsatt ytelse

Cannon Compute Cluster (Holyoke) - Nedsatt ytelse

Boston Compute Nodes - Nedsatt ytelse

GPU nodes (Holyoke) - Nedsatt ytelse

seas_compute - Nedsatt ytelse

Operasjonell

SLURM Scheduler - FASSE - Operasjonell

FASSE Compute Cluster (Holyoke) - Operasjonell

Operasjonell

Kempner Cluster CPU - Operasjonell

Kempner Cluster GPU - Operasjonell

Operasjonell

FASSE login nodes - Operasjonell

Operasjonell

Cannon Open OnDemand - Operasjonell

FASSE Open OnDemand - Operasjonell

Operasjonell

Netscratch (Global Scratch) - Operasjonell

Home Directory Storage - Boston - Operasjonell

Tape - (Tier 3) - Operasjonell

Holylabs - Operasjonell

Isilon Storage Holyoke (Tier 1) - Operasjonell

Holystore01 (Tier 0) - Operasjonell

HolyLFS04 (Tier 0) - Operasjonell

HolyLFS05 (Tier 0) - Operasjonell

HolyLFS06 (Tier 0) - Operasjonell

Holyoke Tier 2 NFS (new) - Operasjonell

Holyoke Specialty Storage - Operasjonell

holECS - Operasjonell

Isilon Storage Boston (Tier 1) - Operasjonell

BosLFS02 (Tier 0) - Operasjonell

Boston Tier 2 NFS (new) - Operasjonell

CEPH Storage Boston (Tier 2) - Operasjonell

Boston Specialty Storage - Operasjonell

bosECS - Operasjonell

Samba Cluster - Operasjonell

Globus Data Transfer - Operasjonell

Legg merke til historikk

des. 2024

Cluster Partially Degraded
  • Løst
    Løst
    Jobs have cleared overnight and a fix for the high load appears to be working. We will monitor for any recurrence, but all appears well at this time.
  • Undersøker
    Undersøker

    Low priority jobs are not getting scheduled despite being at the top of the queue. We are currently investigating this incident and have reached out to SchedMD regarding this.

    See https://support.schedmd.com/show_bug.cgi?id=21627 

NESE tape maintenance Dec. 11, 2024
  • Fullført
    desember 11, 2024 kl. 22:00
    Fullført
    desember 11, 2024 kl. 22:00
    Maintenance has completed successfully
  • Pågår
    desember 11, 2024 kl. 13:00
    Pågår
    desember 11, 2024 kl. 13:00
    Maintenance is now in progress
  • Ikke startet ennå
    desember 11, 2024 kl. 13:00
    Ikke startet ennå
    desember 11, 2024 kl. 13:00

    From Harvard University RC:

    NESE is scheduling a hardware repair for 12/11 to one of disk cache management servers for NESE Tape.

    The scheduled repair will not interrupt the Tape service, but you will likely experience performance degradation in interacting with the archive on 12/11 during the maintenance.

    This maintenance will not affect the NESE Disk service.

Slurm Scheduler high load
  • Løst
    Løst
    The scheduler load has returned to normal. A root cause was identified and remediated.
  • Identifisert
    Identifisert
    The scheduler is experiencing a high load at the moment which may manifest as slowness, timeout or general instability for job scheduling. We are actively working to mitigate this.
FASRC monthly maintenance - Monday December 2nd, 2024 7am-11am
  • Fullført
    desember 02, 2024 kl. 16:00
    Fullført
    desember 02, 2024 kl. 16:00
    Maintenance has completed successfully
  • Oppdater
    desember 02, 2024 kl. 12:52
    Oppdater
    desember 02, 2024 kl. 12:52

    Due to an urgent network issue which requires a restart of some network hardware, all jobs will need to be paused.

    Interactive jobs and the ability to write to some storage may be interrupted.

  • Pågår
    desember 02, 2024 kl. 12:00
    Pågår
    desember 02, 2024 kl. 12:00
    Maintenance is now in progress
  • Ikke startet ennå
    desember 02, 2024 kl. 12:00
    Ikke startet ennå
    desember 02, 2024 kl. 12:00

    FASRC monthly maintenance will occur Monday December 2nd, 2024 from 7am-11am

    IMPORTANT NOTICES

    • holyscratch01 will be set to read-only during this maintenance and will be decommissioned February 1, 2025. Please move any needed scratch data to netscratch and begin using it instead if you have not done so already. The global $SCRATCH variable will be changed to /n/netscratch

    • FASRC will be switching to the Harvard ServiceNow ticket system on Dec. 2nd. Our email addresses remain the same and no action is required on your part.
      Please do not re-open old/closed tickets after Dec. 2nd and instead create a new ticket.

    • Cannon cluster: serial_requeue and gpu_requeue will be set to allow MPI/multinode jobs. Such jobs need to be able to handle preemption/being requeued. 

    Training: Upcoming training from FASRC and other sources can be found on our Training Calendar. at https://www.rc.fas.harvard.edu/upcoming-training/

    Status Page: You can subscribe to our status to receive notifications of maintenance, incidents, and their resolution at https://status.rc.fas.harvard.edu/ (click Get Updates for options).

    Upcoming holidays: Thanksgiving Nov. 28th and 29th. Winter break Dec. 23rd through January 1st 

    MAINTENANCE TASKS
    Cannon cluster will be paused during this maintenance?: NO
    FASSE cluster will be paused during this maintenance?: NO

    • Set /n/holyscratch01 scratch filesystem to read-only

      • Audience: All cluster users

      • Impact: Please adoptthe new scratch filesystem /n/netscratch prior to Dec. 2nd. The $SCRATCH variable will move to /n/netscratch during this maintenance
        Data on holyscratch01 will still be readable, but not writable, and will be fully decommissioned on Feb. 1, 2025.

    • Switch ticketing system to ServiceNow. Our email addresses remain the same.

      • Audience: All FASRC users

      • Impact: All new tickets will go to Harvard'sServiceNow,our email remains the same. Existing tickets will get moved any time someone replies.

      • NOTE: From Dec. 2nd on, please do not re-open any old tickets. Create a new one instead by emailing rchelp@rc.fas.harvard.edu

    • Login node reboots

      • Audience: Anyone logged into a FASRC Cannon or FASSE login node

      • Impact: Login nodes will rebooted during this maintenance window

    • Scratch cleanup ( https://docs.rc.fas.harvard.edu/kb/policy-scratch/ )

      • Audience: Cluster users

      • Impact: Files older than 90 days will be removed. Please note that retention cleanup can and does run at any time, not just during the maintenance window.

    Thank you,
    FAS Research Computing
    https://docs.rc.fas.harvard.edu/
    https://www.rc.fas.harvard.edu/upcoming-training/

nov. 2024

holyscratch01 degraded
  • Løst
    Løst
    holyscratch01 is operational. These failing object stores will continue to cause issues. Please move your scratch data to /n/netscratch if you have not already. Holyscratch01 will be set READ-ONLY on Dec. 2nd. https://www.rc.fas.harvard.edu/blog/announcing-netscratch/
  • Undersøker
    Undersøker

    OST2b has wedged again. We are in the process of troubleshooting this incident.

holylfs06 stability issues
  • Løst
    Løst
    holylfs06 issues resolved.
  • Identifisert
    Identifisert
    We are working on an issue with holylfs06 Expect some instability or slowness.
Citrix (TMS/RCapps) down
  • Løst
    Løst
    Citrix/RCapps is up and working again.
  • Identifisert
    Identifisert
    rcapps.rc.fas.harvard.edu Citrix is down. We are investigating.
Globus - Boston collection unavailable
  • Løst
    Løst

    The transfer nodes for Boston globus are working again. We will follow up with the vendor for any additional information on root cause.

    Thanks you for your patience.

  • Identifisert
    Identifisert
    This issue has not yet been resolved. We have a ticket in with Globus and will update once we know more.
  • Undersøker
    Undersøker
    The Boston collection appears unavailable in Globus. We are investigating.
Netscratch - some directories stalled/unavailable
  • Løst
    Løst
    Netscratch is working normally. The vendor has collected logs and is determining next steps for prevention.
  • Identifisert
    Identifisert

    We are seeing a recurrence of this issue. We are working to determine the cause and resolution.

  • Løst
    Løst
    The instability has been identified and resolved. Netscratch has returned to normal, full operaiton.
  • Undersøker
    Undersøker
    We have reports that some directories in netscratch are unavailable or intermittent. We are working with Vast to determine the cause. This is a new filesystem which is under heavy load for cluster use as well as migration from holyscratch01. We appreciate your understanding of these teething pains and will update you as soon as we have more details or resolution.

okt. 2024

holyscratch01 OST2b issues
  • Løst
    Løst
    This incident has been resolved.
  • Undersøker
    Undersøker

    OST2b (one of the bricks that make up holyscratch01) has wedged again. We are in the process of troubleshooting.

holyscratch01 OST2b issues
  • Løst
    Løst
    This incident has been resolved.
  • Undersøker
    Undersøker

    OST2b on holyscratch01 (one of the bricks that makes up holyscratch01) is hung up. We are failing it over the rectify scratch performance issues.

holyscratch01 Instability
  • Løst
    Løst
    The instability has been resolved.
  • Identifisert
    Identifisert
    We are continuing to work on a fix for this incident.
  • Løst
    Løst
    This incident has been resolved.
  • Undersøker
    Undersøker
    We are currently investigating this incident.
holyscratch01 reboot
  • Løst
    Løst
    This incident has been resolved.
  • Undersøker
    Undersøker

    We are rebooting holyscratch01 to clear a stuck state.

holyscratch01 degraded
  • Løst
    Løst
    OST2b failed over
  • Overvåker
    Overvåker
    holyscratch01 - OST2b failing over

okt. 2024 til des. 2024

Neste

For issues not shown here, please contact FASRC via
https://portal.rc.fas.harvard.edu or email rchelp@rc.fas.harvard.edu