FAS Research Computing - Cannon cluster down – Detalhes do incidente

Experimentando Desempenho Parcialmente Degradado

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.

Cannon cluster down

Resolvido
Totalmente indisponível
Começou em há 26 diasDurado 3 dias

Afetado

Cannon Cluster

Totalmente indisponível undefined 8:21 PM para 9:13 PM, Desempenho degradado undefined 9:13 PM para 2:45 PM

SLURM Scheduler - Cannon

Totalmente indisponível undefined 8:21 PM para 9:13 PM, Desempenho degradado undefined 9:13 PM para 2:45 PM

Cannon Compute Cluster (Holyoke)

Totalmente indisponível undefined 8:21 PM para 9:13 PM, Desempenho degradado undefined 9:13 PM para 2:45 PM

Boston Compute Nodes

Totalmente indisponível undefined 8:21 PM para 9:13 PM, Desempenho degradado undefined 9:13 PM para 2:45 PM

GPU nodes (Holyoke)

Totalmente indisponível undefined 8:21 PM para 9:13 PM, Desempenho degradado undefined 9:13 PM para 2:45 PM

seas_compute

Totalmente indisponível undefined 8:21 PM para 9:13 PM, Desempenho degradado undefined 9:13 PM para 2:45 PM

Atualizações
  • Resolvido
    Resolvido

    Slurm crashed on 4:30p on Friday due to a user running a large sacct query against the Slurm database. This caused the database host to run out of memory and crash the scheduler. To prevent this from reoccurring we are reducing the time range that users are permitted to query at one time to 7 days. Thus if you need to cover a month you would need to query in four 7 day increments.

    We do ask users to be judicious in their querying of the Slurm. Only ask for those fields that you require. Please also ensure any AI agents you have running limit their queries appropriately.

  • Identificado
    Identificado

    To temporarily stabilize the situation, we have reduced the maximum query time for sacct and other Slurm commands to be 1 day. We have filed a ticket with SchedMD to further analyze the issue.

    The cluster is back up and the scheduler is accepting new jobs.

    We will continue to monitor for emergencies over the weekend, and resume in-depth troubleshooting on Monday.

  • Investigando
    Investigando

    The Slurm scheduler is experiencing an error which is impacting jobs. The Cannon cluster will be inaccessible while we troubleshoot.

    We are currently investigating this incident.