Skip to content

June 2026 Community Conversation

This meeting was held on Tuesday, June 9, 2026 at 11:30am ET / 10:30am CT / 8:30am PT / 3:30pm GMT. Monthly community meetings are convened over Zoom to discuss topics of interest. Community members are encouraged to bring questions and ideas to share.

Meeting recording

Quick recap

This June's Coldfront Community Meeting, led by Anton with Dori absent for training. The core working group provided updates on recent demos, including Chris's storage plugin and Tufts' billing system, and discussed plans for V2 development with a focus on PERC tutorial preparation. Andrew Bruno reported that the governance working group had stalled, while Jay mentioned ongoing documentation efforts. The group discussed monitoring solutions, with Matthew sharing his exploration of OpenTelemetry for application monitoring and Jay recommending Rollbar for Python error tracking. New participants introduced themselves, including Andrew Damon from RPI evaluating Coldfront as a replacement for their homebrew orchestration system, and Natascha from KIT working on storage integration with Bcache tape storage. The discussion covered Slurm integration challenges, particularly around user deprovisioning and the differences between file-based and REST API approaches, with Sajid sharing his work on scripts to handle association removals. The conversation ended with Andrew Bruno inviting others to consider demoing their Coldfront implementations at future community meetings.

Next steps

Jacob Boschee

  • Upload and sanitize Ansible playbooks for Coldfront integration and share with the community after removing sensitive information.

Sajid

  • Investigate adding job cancellation and association removal functionality to the REST API plugin for Slurm integration, and share progress/results with the group.

Collaboration

  • Andrew Bruno / UB team: Post finalized PEARC demo and Birds of a Feather dates on the website once details are confirmed.
  • All interested parties: Contact Andrew Bruno or Anton if interested in presenting a demo (plugin, integration, or use case) at a future community meeting.

Summary

June Coldfront Community Meeting

Anton led the June Coldfront Community Meeting and noted that Dori was out for training. He decided to hold an open forum discussion since there was no formal agenda, with the possibility of hearing updates from working groups if available. The meeting was expected to be short due to the small group present.

PEARC Event Planning Updates

The core working group provided updates on recent demos, including Chris's storage plugin and Matt's work at IU, and discussed upcoming focus on the PEARC tutorial and V2 development. The group noted that the governance working group has stalled slightly, and Anton mentioned plans for revamping documentation. The team discussed preparations for the upcoming PAERC event in late July, including plans to use Coldfront V2 for the demo and organize a Birds of a Feather session. Matthew raised a question about monitoring their instance using OpenTelemetry, but no specific answers were provided in the transcript.

Application Monitoring Integration Discussion

Andrew Bruno expressed interest in integrating more application-level monitoring into Coolfront, specifically mentioning Prometheus and OpenTelemetry as potential solutions. Jacob Boschee discussed Rollbar as a commercial product for capturing Python and JavaScript errors, though not integrated into Coolfront itself, and suggested it could be added easily through configuration. The discussion focused on implementing application-level monitoring and auditing capabilities in Coolfront, with Jacob Boschee noting that Rollbar uses Python's standard error logging framework which could potentially be leveraged for auditing purposes.

Coldfront Orchestration System Evaluation

Andrew Damin from RPI introduced their team's evaluation of Coldfront as a potential replacement for their orchestration system, currently using a homebrew solution since 2007. They are exploring Coldfront alongside Vantage as alternatives, with particular interest due to their integration with the Empire AI project and the need to avoid high costs. Natascha from KIT shared their plans to use Coldfront for managing project-based long-term storage, currently working on storage integration with Bcache tape storage. Jacob Boschee reported positive early results from their 3-month Coldfront deployment, noting only minor adjustments needed for renewal processes.

HPC Automation Tooling Discussion

The group discussed their experiences with automation tooling for managing HPC systems, with Jacob Boschee sharing that his team submitted a paper for PERC and has open-source Python tooling available on GitHub, though private Ansible playbooks will be shared later after sanitizing IP ranges. The discussion covered current usage of Slurm features through Coldfront, with participants sharing that they primarily use basic storage and group management, while advanced features like QoS are not yet implemented at most centers. Andrew Bruno explained that Coldfront currently integrates with Slurm using both file format imports and a REST API plugin, noting that the REST API approach requires careful consideration of Slurm versioning and token management.

Slurm API Security Concerns

John and Andrew Damin discussed concerns about the Slurm API security, particularly regarding potential denial of service attacks, though they acknowledged it may not be as risky as other recent issues. Sajid highlighted differences between loading and removing associations in Slurm, noting that removal requires specific steps including job cancellation and setting max submit to zero, and mentioned plans to implement these features in their fork of the REST API. Andrew Bruno confirmed similar challenges with automated association removal and expressed interest in Sajid's solution. John described their current approach to managing user access through AD groups and cron jobs that monitor and cancel jobs when UIDs no longer resolve.

Slurm User Management Automation

The group discussed automated processes for managing user accounts and job cancellations in Slurm systems. John explained their approach of using an SMU account with default job restrictions to handle deprovisioned users, while Sajid shared his Python script that removes user associations before account removal. They discussed performance improvements from changing their LDAP search base to focus on active users only, reducing Slurm restart times from 45 minutes to just a few seconds. The conversation concluded with a discussion about potentially federating authentication services across multiple servers and the possibility of demonstrating various Coldfront use cases at future community meetings.