Hi users,
I'm having some problems setting up the PVM daemon for the use of a slave machine in a distributed parallel problem in CFX-5.5.1. The problem is that I can't even add manually the slave host under the pvm.exe utility. The output says that the system could't find the specified file(wish I could know which file it is). The rsh service is running ok but I can't start the daemon in my desired hosts. The error is shown below.


***need more ? *** hoster() 1 to start, wait id 262158 0. tc0000 sargento so="" Bound to port 1021 Bound to port 1020 socket closed (?) after reading from :

The system cannot find the file specified. socket closed (?) after reading from sargento: Error doing $PVM_ROOT/lib/pvmd -s -d0 -nsargento 1 8f6a2690:05c0 4080 3 8f6a269d:0000 on sargento phase1() rsh failed for host sargento
Here, sargento is my slave hostname. So, is there a troubleshooting guide available for it? Any hints?
Thanks in advance, cfd guy

All the required configuration files are ok? It means files PVMhosts, hosts.ccl, CFXhosts in C:\CFX\CFX-5.5.1\Config\5.5.1 directory.

Ribeiro

All done Ribeiro. It was a problem with the PVMhosts file. Now it's ok... Thanks !
cfd guy.

What was the exact problem?

Regards, Cujo

Hi cfd guy.

Right now I am having the exact same problem and similar error message as you just had. What was your problem with the PVMhosts file? I am trying to run a distributed parallel run with one master and one slave. I have only included a PVMhosts file on the master node. My PVMhosts file looks like:

&pc1 dx=d:/cfx/cfx-5.5.1/tools/CYGWIN32_NT/pvm3.4/lib/win32/pvm3d.exe &pc2 dx=d:/cfx/cfx-5.5.1/tools/CYGWIN32_NT/pvm3.4/lib/win32/pvm3d.exe

Where pc1 is the master node, and pc2 is the slave nodes. I've tjecked that the stated home directories for the pvmdaemon in the PVMhosts file is in agreement with the cfx installation.

Thanks in advance, and best regards Jesper.

Hi Guys, I have a quick question regarding parallel setup. If I were to set up a linux cluster of 4 pc's would it be beneficial to have them all as dual CPU machines with 2Gb of ram on them or would it be better to say have separate machines each with a single CPU and 1GB of ram but say only have 6 pc's (as cost may not allow 8pc's) ?? Bob

Hi Jesper and cujo,
The whole problem was that I was not configuring correctly the three files in the config subdirectory of CFX-5 installation. Jesper, for your specific case PC1 and PC2 hostnames, the files should be something like this:
------- PVMhosts File -------

&PC1 dx=d:/cfx/cfx-5.5.1/tools/CYGWIN32_NT/pvm3.4/lib/win32/pvm3d.exe &PC2 dx=d:/cfx/cfx-5.5.1/tools/CYGWIN32_NT/pvm3.4/lib/win32/pvm3d.exe
-----------------------------


------- hosts.ccl File -------
EXECUTION CONTROL: PARALLEL HOST LIBRARY: HOST DEFINITION: PC1 Number of Processors = 1 Installation Root = D:\CFX\CFX-5.5.1 END # HOST DEFINITION PC1 HOST DEFINITION: PC2 Number of Processors = 1 Installation Root = D:\CFX\CFX-5.5.1 END # HOST DEFINITION PC2 END # PARALLEL HOST LIBRARY END # EXECUTION CONTROL
-----------------------------
------- CFXhosts File -------
EXECUTION CONTROL: PARALLEL HOST LIBRARY: HOST DEFINITION: PC1 Number of Processors = 1 Installation Root = d:\CFX\CFX-55~1.1 END # HOST DEFINITION PC1 HOST DEFINITION: PC2 Number of Processors = 1 Installation Root = d:\CFX\CFX-55~1.1 END # HOST DEFINITION PC2 END # PARALLEL HOST LIBRARY PC1 1 1.00 & d:/CFX/CFX-55~1.1/bin/5.5.1/winnt/solver-pvm.exe PC2 1 1.00 & d:/CFX/CFX-55~1.1/bin/5.5.1/winnt/solver-pvm.exe -----------------------------
Pay attention with the hostnames, they are CASE SENSITIVE (take a look at CFX-Community in the FAQ's section).
Well, I believe this will fix your problems, at least it fixed mine. Good luck!
cfd guy

Hi Bob,
Once I fixed my PVM problem, now I'm running 5 processes where 4 processes are running in two different DUAL PC's and one in a single PC. I believe that you spend some precious time due file transferring over the intranet. However running in 5 processes it saved me a lot of memory. I also believe that is cheaper you buy 4 Dual PC's than a little cluster containing 8 CPU's (if I'm wrong please correct me). And in your case I would rather buy 4 Dual PC's than buying 6 single ones due file transferring and physical space.
Regards, cfd guy

Hi cfd guy! You really saved my day. It all works now. The problem was in the CFXhosts file. The slave node was not defined properly.

Best regards Jesper.