Skip to content

OSG Technology Area Meeting, 6 August 2024

Announcements

Triage Duty

Triage duty shifts Tue-Mon

  • This week: Matt
  • Next week: TimT
  • 3 (-1) open FreshDesk tickets
  • 0 (+0) open GGUS ticket

Jira (as of Monday morning)

# of tickets Δ State
220 +2 Open
17 +1 Selected for Dev
32 +3 In Progress
16 +0 Dev Complete
1 -3 Ready for Testing
1 +1 Ready for Release

OSG Software Team

What's the process for adding patching XRootD in Pelican containers?

  • AI (Matt): Kuantifier
    • Status of NET2 and Prometheus auth?
      • Follow up again re: prometheus API access
    • Derive ProbeName from SUBMIT_HOST
    • nodeSelector + toleration support
    • GHA from osg helm-charts repo
    • Require that pods set a CPU resource request
      • We use this to derive the "Processors" field
  • AI (Mat): Set up ARM Koji builder
    • Need to updated systemd unit file, then create production builder VM
  • AI (Mat): EL9 repo
  • Contribute VOMS patches upstream?
    • Want to drop from OSG software
  • Topology "StashCache" GHA has been failing for some time
    • Validates topology cache list against cvmfs
  • New Yubikeys are in, BrianL working on purchasing hubs

Discussion

None this week

Support Update

  • University of Tennessee at Chattanooga (BrianL): NVIDIA MPS server (provides a CUDA interface) seems to be causing glideins to hang upon condor_gpu_discovery. NVIDIA recommends against using MPS across multiple jobs (but may be useful for a single user).
  • Add check to see if it's enabled? Show up as a process under ps
    • Probably not exposed to containers
  • JLab (Mat): Support in setting up Pelican Origin, setting up weekly meeting

DevOps

None this week

OSG Release Team

  • Ready for Testing
    • htcondor-ce
    • osdf-server-7.9.3
    • xrootd-5.7.0
    • igtf-ca-certs
  • Ready for Release
    • GlideinWMS 3.10.7

Discussion

  • IGTF CA certs will be released next week due to the holiday
  • Full support periods for HTC23 and OSG23 are not fully aligned
    • Support for OSG 23 lasts several months longer
  • Marco has been using OSG noarch packages successfully in ARM