Learning Forum

[marufhossain]

Hello,

I have been struggling with setting up a cluster in HFSS 2021 R2. I have gone through a number of forum posts and it seems everything is setup correctly, yet the simulation is stuck at the first step 'Determining memory availability on Distributed machines".

The task we are trying to simulate is an encrypted automotive antenna which goes out of memory on a single windows server, therefore we are trying to distribute it across two windows servers, connected via a domain network. I made sure I installed the same version of HFSS (2021 R2) on both machines. Same version of intel MPI is also installed, and tested that the machines are talking (attached MPI_test screenshot).

[marufhossain]

Hi, can anyone please answer?

[randyk]

The following should resolve the issue:
INITIAL Setup of Intel MPI:
1.. Are both computers on a domain?
If not, then make sure customer creates an account with the exact same username and password -- on each computer.
2.. log into each machine as the same user
3.. Did customer install Intel MPI from AnsysEM installation media, there will be a\Prerequisites\Intel MPI 2018.3\w_mpi_p_2018.3.210.exe
4.. After installing Intel MPI, start the Hydra services and register user/password(from an elevated CMD -- rt-click on CMD and "run as admin")
>"%I_MPI_ROOT%"\em64t\bin\mpivars.bat
>hydra_service -install
>hydra_service -start

5.. Register your credentials, enter:
From a normal CMD:
>mpiexec -register
Confirm your credentials were correct, enter:
>mpiexec -validate

6.. set default tempdirectory to c:\scratch on both computers
-- Createc:\scratchfolder
-- edit "C:\Program Files\AnsysEM\AnsysEM21.2\Win64\config\default.cfg" so contents are:
$begin 'Config'
tempdirectory='C:/scratch'
$end 'Config'

7.. The most recent version of AnsoftRSMService should be installed and all versions AEDT registered with AnsoftRSMService
-- note: ANSYS RSM is not AnsoftRSMService and is not compatible with AEDT.
-- AnsoftRSMService is included in the AEDT installation media whereas ANSYS RSM is a download on the ANSYS Customer Portal.

[randyk]

TESTING Intel MPI

Open a CMD and run the following commands on each machine - share the output:

ping
--make sure name resolves to IP

ipconfig
-- make sure all machines in the same network
-- confirm all machines have a single active network(if multiple networks would need to set env vars to push traffic thru a specific connection)

set u
-- make sure the same user/password on each machine

set i
-- looking for I_MPI_ROOT


turn firewalls completely off on all machines for initial testing
From Powershell CMD, run:Get-NetConnectionProfile
-- confirm all computers NetworkCategory match (private/public/domain)


"%I_MPI_ROOT%"\em64t\bin\mpiexec -version
-- confirm version Intel MPI same on all machines

"%I_MPI_ROOT%"\em64t\bin\mpiexec -n 2 -host localhost "C:\Program Files\AnsysEM\AnsysEM21.2\Win64\schedulers\diagnostics\Utils\intelmpi_test.exe"

Output should be something like:
Intel MPI
Hello world! I'm rank 0 of 2 running on computer1
Hello world! I'm rank 1 of 2 running on computer1

(correct "Computername2" to reference the real computer name of remote machine)
"%I_MPI_ROOT%"\em64t\bin\mpiexec -n 2 -host localhost "C:\Program Files\AnsysEM\AnsysEM21.2\Win64\schedulers\diagnostics\Utils\intelmpi_test.exe" : -n 2 -host Computer2 "C:\Program Files\AnsysEM\AnsysEM21.2\Win64\schedulers\diagnostics\Utils\intelmpi_test.exe"

Output should be something like:
Intel MPI
Hello world! I'm rank 0 of 4 running on computer1
Hello world! I'm rank 1 of 4 running on computer1
Hello world! I'm rank 2 of 4 running on computer2
Hello world! I'm rank 3 of 4 running on computer2

[randyk]

AEDT setup

Tools -> Options -> General Options -> Remote Analysis ->
[x] specified user
enter username, password, domain

[randyk]

AEDT setup (continued)

Your Analysis configuration should have the localhost listed first and then the remote computer:

Please let me know if this resolves the issue