OSG Technology Area Meeting, 3 December 2024¶

Coordinates: Conference: +1-415-655-0002, PIN: 146 266 9392, https://morgridge-org.zoom.us/j/91987518094 (password sent separately)
Attending: Brian, Tim, Mat, Matt

Announcements¶

Triage duty shifts Tue-Mon

(delta since 11/05 meeting)

EL9 repo status
- AI (Mat): Finish announcement email
- AI (Matt): Review PR for migration script
Glidein dir cleanup status: temporarily disabled due to holding up Glidein startup at a site with a large number of dirs to clean up on a shared filesystem. Todd T, Matt, and Mat had a brainstorming discussion, for which Todd will write up a design.
AI (Mat): Create new VMU disk images so we can run VMU tests again.
AI (Mat): Build XRootD 5.7.2
AI (Mat): Review OSG 24 upgrade instructions PRs
AI (Matt): Remaining OSG 24 image builds:
- docker-xcache: Mat to review PR
- docker-compute-entrypoint: requires a new osg-ce build, though hosted CEs should already be using the version of osg-ce from main-testing or upcoming-testing. Matt should ask hosted CE operations if they are running the new version
AI (Matt): VMU tests on ARM; CHTC has an ARM machine that is restricted to running jobs from BatLab and CHTC staff; CHTC operations will review the policy.

Thoughts on how we can reduce Hosted CE IPv4 usage? Need to discuss with the HTCondor development team if unique IPv4 addresses are necessary; Tim thinks yes, for reverse DNS, but Mat thinks reverse DNS isn't used anymore

Tim suggests running each Hosted CE on a different port but Brian says that would be verfy painful to manage and would essentially lock CEs to specific nodes.

Brian will investigate IngressRouteTCP.

SURF (Mat): LIGO site using an xcache-based cache; they got a new cert and need a LIGO test file to test with because their cache does not support the OSG VO.
IceCube (Mat): condor_gpu_discovery not discovering Intel GPUs. Mat will discuss with the HTCondor dev team.

None this week