Author Topic: Cannot use more than 10 nodes: Licensed number of users already reached.  (Read 293 times)

0 Members and 1 Guest are viewing this topic.

Offline SDas

  • New ATK user
  • *
  • Posts: 3
  • Country: in
  • Reputation: 0
    • View Profile
Hi,

I am running a few calculations simultaneously in a supercomputer cluster. The system is quite big and I need atleast 10 nodes (48 cores each) for each calculation. But I can only run one such calculation at a time. That is, if I already have an atkpython job running on 10 nodes, and I submit another concurrently, it stops running with the following error: Licensed number of users already reached. (Full log file is attached as 'final.log').

Also, I get the same error if I attempt to run a solitary single job on 20 nodes.

I assume it is because there is a maximum limit of the no. of licenses that can be in use concurrently. But I want to understand why a single job cannot be run on 20 nodes (It would make each calculation incredibly faster). Is there any parallelization option that can help me avoid this issue?

Thanks,
SD

Offline Ulrik G. Vej-Hansen

  • QuantumATK Staff
  • Supreme ATK Wizard
  • *****
  • Posts: 413
  • Country: dk
  • Reputation: 7
    • View Profile
Yes, it does indeed look like you run out of dp licenses. Please file a ticket on SolvnetPlus, there we can better help with license-related questions.

Offline Anders Blom

  • QuantumATK Staff
  • Supreme ATK Wizard
  • *****
  • Posts: 5056
  • Country: dk
  • Reputation: 85
    • View Profile
    • QuantumATK at Synopsys
Running in parallel on many (>100) cores doesn't always make calculations run faster. It really depends if there are enough independent tasks to parallelize over, such as energy points on the NEGF contour or k-points. For most calculations, the benefit of going from 200 to 400 cores is far less than 2x so you would be better off running 2 calculations on 200 cores than 1 calculation on 400 anyway.