CFX on Windows HPC: Confused about cfx_ccsoptions.txt
 

Hi All,
I built up a cluster for running CFX on with Windows HPC. I have my network set up with Server 2012 on the headnode and Windows 7 64 Professional two workstation nodes. I am following the directions from the link below to enable the sending of jobs from CFX to Windows HPC:

http://www.ansys.com/About+ANSYS/Par...tallation+FAQs

I am confused about the setup of the text file. It says to create a \%USERPROFILE%\.cfx\cfxccs_options.txt .

What does the "\%USERPROFILE%\.cfx\" part mean? Where do I put the text file so that CFX will find it? I guess the file path given above doesn't make sense to me. Any help would be greatly appreciated.
Frank

So I have been doing some reading. Apparently %USERPROFILE% is an "environment variable." It looks like it points to the "Administrator" folder level. I have found the path then to the .cfx folder. I made the text file as outlined in the link in my initial post and inserted it in to this folder.

When I create the environment variable as described in the link in my previous post, how does CFX know how to call the variable to read the text file? It shows in the example to set the variable but it doesn't say what to name it.

I just called the environment user variable "HPC". I can get CFX to run through Windows HPC as long as I don't exceed the number of cores on the host machine (12 cores, 6 real, six virtual, Ivy-Bridge 4390k). I get an error message from CFX:

[I]"[8] fatal error
Fatal error in PMPI_Bcast: Other MPI error, error stack:
MPI_Bcast(buf=0x0000000142090244, count=1, MPI_INTEGER, root=0, MPI_COMM_WORLD) failed
[ch3:sock] failed to connnect to remote process 4C02BC26-2C82-4d84-A114-060C95E2D5CC:12
unable to connect to 192.168.0.19 on port 55440, exhausted all endpoints
unable to connect to 192.168.0.19 on port 55440, A connection attempt failed because the connected party did not properly respond after a period of time, or established connection failed because connected host has failed to respond. (errno 10060) [9-12] terminated"


All three computers are definitely connected. 192.168.0.19 is the server IP. Perhaps I didn't change the .NET Framework settings properly and that's why they are not talking? I know I am answering my own post but this might help some people trying to do the same thing. Of course, any input would be welcome.

The error above is a red herring. I have 28 cores available in my cluster. It seems that although Windows HPC is trying to allocate to all 28 cores, CFX is partitioning the mesh on to one CPU in the cluster only. For example; one computer in the cluster has 4 cores. If I specify the simulation to run on 28 cores, CFX will try to create 28 partitions on that one computer with only 4 cores while Windows HPC trys to allocate among all the cores in the cluster. The job fails. Does anyone know how to tell CFX to partition on to all the cores?