3.10.5.57

[PalNilsson]

• When the worker node has a GPU, pilot will collect detailed info using nvidia-smi and report to PanDA
  • Discussed in ATLASPANDA-1368 and ATLASPANDA-1441
  • Reported fields are defined here: update_worker_node_gpu/post
  • Related: RC dev jobs now have GPU payload (https://hammercloud.cern.ch/hc/app/atlas/test/20319685/)
• EIC and generic pilot users
  • Now reporting glidein_site and remote_schedd_name with destinationSite and remoteSite with updateJob
  • Requested by X. Zhao
• Corrected logic for verifying max running time on push queues
  • Previously, on PUSH queues, the code would fallback to max running time verification that should only be used on PULL queues
  • Requested by R. Walker
• Cgroups updates (for HTCondor versions >= 24.0.7)
  • Subprocesses (i.e. all processes started by the pilot) are moved into their own cgroup - the pilot process is in a separate group
  • Memory and PID controllers are enabled
  • Monitoring of cgroups added (report once a minute)
  • Discussed in JIRA ticket ATLASPANDA-1251