AI_INFN Technical meeting – Minutes and actions

Date: 2024-07-15 (offline)

Tracked developments:

Automation of RKE2 deployments in INFN Cloud

• NTR
• [Open since last week] We need to schedule the upgrade of kubernetes versions in the platform

Develop monitoring and accounting infrastructure (R. Petrini)

• [Open since last week] We need an automated backup of the postgres database used for the accounting

Environment setup (M. Barbetti, S. Giagu, S. Bordoni, L. Cappelli)

• A dedicated meeting was held on Monday July 8th.
• There is a problem with the order of loading the libraries of PennyLane and Tensorflow. Using the right order, things work.
• Matteo will study the problem and will try to find a more beautiful solution.
• Tested cirq following the tutorial, but it seems not to work with GPUs. Is not particularly urgent for users.
• Laura and Matteo will try to setup a GNN with quantum backend.

Offloading tests with virtual kubelets (G. Bianchini, D. Ciangottini)

• InterLink has been updated to version 0.3.0, with breaking changes in the APIs
• We have updated succesfully:
  • the ReCaS backend (kueue plugin on INFN Cloud K8s)
• Still broken:
  • the RTX-5000 backend at Cloud@CNAF (kueue plugin on INFN Cloud K8s)
  • the T4 backend at Cloud@CNAF (docker plugin on a VM)
  • the CloudVeneto backend (still some failures to be followed up)
• We have not started the procedure yet for:
  • The kueue plugin in interlink CE
  • The virtual node in the platform

In this summer days the effort to maintain this complicated infrastructure is limited.

Acquisto FPGA

• Macchinone 4 is being installed (An RTX is being installed together with the two V70).

Advanced Hackathon

• From Andrea Paccagnella:

I spoke to Marco Zanetti from Padua and he told me that the secretariat had found two classrooms with around 45 seats. One for three days and one for two afternoons.
I replied to let him know that we would need a classroom for four days and one for two afternoons.
Tomorrow he will return to Padua and go to the secretariat.

Status legend

Active
Priority
Problems
Postponed or Blocked by others
Completed