Rucio Meeting

Thursday 15 Jan 2026, 15:00 → 16:00 Europe/Zurich

Martin Barisits (CERN)

15:00 → 15:05 News

• 39.1.0 moved to next week

15:05 → 15:25 Community News & DevOps roundtable

• ATLAS
  • Issues with list_dataset_replicas - twice a wrong issue was returned (not using --deep)
    • Did the re-sync script not run?
    • Re-evaluate if --deep can be used again? (Possibly introduce some caching)
• CMS
  • list_datasets_rse does not return
    • 4PB RSE
    • Relies on collection_replicas 
    • We need a permanent solution for collection_replicas (normal) vs --deep also in respect to list_dataset_rse
• Fermilab DUNE / RUBIN / ...
  • DUNE migrating to loadbalancers to see if this is the cause of problems in the job submissions
  • Concern about client-certificate removal
    • Not sure what we can do on Rucio side (other than push for tokens)
    • RUBIN might have a Plan B, DUNE still investigating
• ESCAPE / EOSC
  • Rucio extension now in production in SWAN @ CERN
  • Questions about istape attribute
• INFN Datalake
  • Monitoring templates fix
  • Updating development instance
• RI-SCALE
• OSCARS MADDEN / ETAP
  • Metadata storage
• DaFab
• IHEP

15:25 → 15:55 Developers roundtable

• Conventional commits to be enabled this Monday
  • Old PRs grandfathered in, new PRs must follow CC
  • See documentation
• PR Crisis: We have way too many open PRs, and the situation is not getting better
  • This is a combination due to low quality PRs, PRs that are difficult to review, bottlenecks in reviewing, and bottlenecks in merging
  • We will address this by one immediate action (PR Sprint, see later) and several changes in our development processes
  • PR Sprint
    • Starting this Monday we will start a PR Sprint to rigorously reduce the number of PRs
      • Focus on review and merging
      • Close stale things (they can be reopened)
      • EVERYONE: Please re-act quickly to comments
  • Changes in processes
    • Create a new PR template which outlines the most important points for a developer
      • Testing, coverage, contribution guide, author should pick a reviewer, etc.
      • This should make the job of a reviewer easier
    • Create a new reviewers guide
      • To guide the reviewer in reviewing a PR 
      • This will have a review template where the reviewer needs to answer some questions
        • This should help the merger to understand the level of confidence the reviewer has in the PR
    • We will (slowly) expand the merger team
    • We will introduce automation (RucioBot) of PRs
      • Stale PRs / Failing tests -> Bot will auto-close if not addressed quickly
      • PR without issue -> Close
      • Draft PRs after no activity -> Close
      • The Bot should, as much as possible, enforce the contribution guide automatically
    • For LARGE and XLARGE issues, a reviewer needs to be picked on creation of the issue
      • In principle, already part of our guideline, I will enforce this strictly on roadmap issues now
      • Discuss with the reviewer (which is more like a co-author) early on about the planned changes and the processes
        • This should avoid situations where a PR goes into wrong directions
    • Review needs to be part of the development process. Every developer should spend some time reviewing PRs!
• Reflection about ongoing Sprint 1 (Sprint Board) [Karan]
  • Less activity, but overall good
• Rucio 40 roadmap
  • Please size estimate all the priority issues!
  • Pick a reviewer for LARGE and XLARGE issues!
• Rucio 40 priority followup
  • Todo
    • Create a new API endpoint for the clients to request tokens #6638[Dimitrios]
    • Implement new token authentication for download #7029[Dimitrios]
    • More useful client docstrings and CLI help messages (Target: 15/25) #363 [Dimitris]
    • Consider migrating from Jobber to Cloudprober #152 [Eric]
    • DIDs not being shown correctly in the extension after a successful download #95 [Giovanni]
    • Make available + Add to Notebook features failing #83 [Giovanni]
    • Token refresh after expiration #73 [Giovanni]
    • Prevent multiple Reaper threads from working on the same replicas #6512 [Hugo]
    • Test stability on LTS branches #7964 child of #7667 [Karan]
    • Generate "Configuration parameters" documentation page automatically, to avoid mismatches between documentation and code #325 [Maggie]
    • Change mixed prometheus_client and probe_metric approachs to use PrometheusPusher #129 [Maggie]
    • Move implimentation of CLI to new CLI structure #8295[Maggie]
    • Possible Belle II specific code in DIRAC functionality #7824[Max]
    • Remove hard-coded one-day lifetime in DIRAC API #8172 [Max]
    • Simplify what we run on CI (test suites, OS, Python versions, DB, …) #7965 child of #7667 [Mayank, Karan]
    • [ EPIC ] : Mutation Operations and Role Based Access to pages/features (Target 5) #622 [Mayank]
  • In Progress
    • Consider using Python venvs in containers to avoid conflicts with system-installed packages #458 [Ben]
    • Do not use regex to split did, use scope extraction method #7519[James]
    • [EPIC] UX improvements (Target 22) #621 [Mayank]
  • In Review
    • Add startup self-check mechanism to block Rucio services when critical diagnostics fail #8197 child of #8011 [Dimitris]
    • Ensure PostgreSQL ENUM types created during Alembic migrations honor the configured schema #8145 child of #7737[Dimitris]
    • Wrong documentation for dids/scope/name/files call #8053[Dimitris]
    • Testing: Make tools/run_tests.sh idempotent #7737 child of #7667 [Dimitris]
    • No way to check or change ownership of a scope using Rucio client #7830 [Maggie]
  • Done
    • Listing containers with more than 10k entries breaks the kernel #102 [Giovanni]
    • Rucio silently skips importing configured policy package in case of errors #7962 [Max, James]
  • Delayed
    • Stop auto-forwarding of old-style CLI commands #8294 [Maggie]
• Other discussion
  • Rucio clients realistic lower dependency bounds (for PyPi)

15:55 → 16:00 AOB