Learning Forum

[pweakliem]

We have a HPC cluster where all nodes see the same copy of Ansys software through a NFS mount point. We can run Ansys on single nodes, and it works fine. However, on another (almost identical) node it fails with a license error.

e.g.

------------------------

/software/ansys_inc/v211/AnsysEM/AnsysEM21.1/Linux64/ansysedt -ng -batchsolve -UseElectronicsPPE -machinelist numcores=4 -auto Si_test.aedt

runs fine

=============logfile (end of it) ====================

localhost RAM: 90%, 4 cores, 1 GPUs

[info] Using command line: "/sw/ansys_inc/v211/AnsysEM/AnsysEM21.1/Linux64/ansysedt.exe -ng -batchsolve -UseElectronicsPPE -machinelist numcores=4 -auto Si.aedt".

[info] Project:Si, Design:Bowtie_WithAirbox (DrivenModal), Normal completion of simulation on server: Local Machine. (4:43:33 PM Feb 02, 2022)

Stopping Batch Run: 4:43:33 PM Feb 02, 2022

======================================

However, when we try to run it on the nodes that are higher memory - which are very similarly configured, we get a license error.

=====(logfile - end of it) ======

Machines:

localhost RAM: 90%, 4 cores, 1 GPUs

[info] Using command line: "/sw/ansys_inc/v211/AnsysEM/AnsysEM21.1/Linux64/ansysedt.exe -ng -batchsolve -UseElectronicsPPE -machinelist numcores=4 -auto Si.aedt".

[error] Project:Si, Design:Bowtie_WithAirbox (DrivenModal), Failover feature 'ANSYS Electronics Solve Level 2' is not available.

FLEXlm server is not responding. Resource temporarily not available.

   ANSYSLI_SERVERS: [email protected]

   FLEXlm Servers: [email protected] (4:27:39 PM Feb 02, 2022)

[error] Project:Si, Design:Bowtie_WithAirbox (DrivenModal), Simulation was terminated by license error. (4:27:39 PM Feb 02, 2022)

Stopping Batch Run: 4:27:39 PM Feb 02, 2022

===============================

The license info is all in the same NFS mounted directory for both systems, so it's hard to understand what might be going on. I didn't see any obvious files in the user's home directory that might override license info. Both nodes can reach the license server. We see checkouts normally for the working node, but on the non-working node we see a checkout of elec_solve_level2 and an immediate check back in of it, and then the job fails.

Does this ring any bells for anybody, or is there a simple way to debug or trace the license checkouts via the command line?

[ANSYS_MMadore]

The checkout/in immediately is typically indicating a problem with communication with the vendor daemon or a hosts/DNS issue. Can you investigate any differences in the host resolution or network connectivity with the nodes and the license server?