- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Installed Parallel Studio XE 2016 Update 2 Cluster edition on Windows 7 Pro x64 (installed on two systems) and attempting to run the MPI test program.
Each system can run the test program on itself, but I cannot run on other system from either system.
hydra_service.exe can be -remove and -install correctly.
I can see and copy files from each system to the other system (to shared folder), test program installed in shared folder.
C:\Test\TestMPI_CPP\TestMPI_CPP>mpiexec.exe -n 8 -hosts thor \Downloads\testmpi_cpp.exe Hello world: rank 0 of 8 running on Thor Hello world: rank 1 of 8 running on Thor Hello world: rank 2 of 8 running on Thor Hello world: rank 3 of 8 running on Thor Hello world: rank 4 of 8 running on Thor Hello world: rank 5 of 8 running on Thor Hello world: rank 6 of 8 running on Thor Hello world: rank 7 of 8 running on Thor C:\Test\TestMPI_CPP\TestMPI_CPP>mpiexec.exe -n 8 -hosts i72600k \Downloads\testmpi_cpp.exe Error connecting to the Service [mpiexec@Thor] ..\hydra\utils\sock\sock.c (270): unable to connect from "Thor" to "i72600k" (No error) C:\Test\TestMPI_CPP\TestMPI_CPP>dir \\i72600k\Downloads\TestMPI_CPP.exe /b TestMPI_CPP.exe
The results above are same (with name changes) when run on other host.
I've run mpiexec -register on both systems.
Any hints would be helpful
Jim Dempsey
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Do you have any firewalls between the two systems? Can each ping the other by name?
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Yes. Both can ping each other by name.
This would have been self-evident by DIR \\name\folder working.
Windows Firewall on both systems have checks on
Intel(R) MPI Library Process Manager, Intel
Anything else need to be enabled? (added)
Jim Dempsey
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
You will probably need to set an exception for the program you are trying to run as well. See https://software.intel.com/en-us/articles/firewalls-and-mpi for general steps for working with a firewall.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
James,
After a bit of work, I had to add a Firewall exception to allow port 8679 incoming TCP packets.
At least the sample test program now runs.
Jim Dempsey
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Well this is interesting
While on system A I can mpiexec run on system A or B
and
While on system B I can mpiexec run on system B or A
From either, I cannot run on both
The documentation gives conflicting information as to the command line for multiple hosts.
Can you provide a correct command line? (run Foo.exe on systems A and B)
Jim Dempsey
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
There are several different ways to do that. Here are a few examples, see https://software.intel.com/en-us/articles/controlling-process-placement-with-the-intel-mpi-library for more. The article is aimed at Linux*, simply substitute mpiexec.hydra for mpirun to run in Windows*.
>mpiexec.hydra -n 2 -ppn 1 -hosts sysA,sysB Foo.exe
>type hosts.txt sysA sysB >mpiexec.hydra -n 2 -ppn 1 -f hosts.txt Foo.exe
>type machines.txt sysA:1 sysB:1 >mpiexec.hydra -n 2 -machinefile machines.txt Foo.exe
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
The examples shows using mpiexec, not mpiexec.hydra.
My results:
C:\Program Files (x86)\IntelSWTools>mpiexec.hydra -n 16 -ppn 8 -hosts Thor,i72600k \Downloads\TestMPI_CPP.exe Error connecting to the Service [mpiexec@Thor] ..\hydra\utils\sock\sock.c (270): unable to connect from "Thor" to "i72600k" (No error) C:\Program Files (x86)\IntelSWTools>mpiexec.hydra -n 16 -ppn 8 -hosts Thor,i72600k \Downloads\TestMPI_CPP.exe Hello world: rank 0 of 16 running on Thor Hello world: rank 1 of 16 running on Thor Hello world: rank 2 of 16 running on Thor Hello world: rank 3 of 16 running on Thor Hello world: rank 4 of 16 running on Thor Hello world: rank 5 of 16 running on Thor Hello world: rank 6 of 16 running on Thor Hello world: rank 7 of 16 running on Thor Hello world: rank 8 of 16 running on i72600K Hello world: rank 9 of 16 running on i72600K Hello world: rank 10 of 16 running on i72600K Hello world: rank 11 of 16 running on i72600K Hello world: rank 12 of 16 running on i72600K Hello world: rank 13 of 16 running on i72600K Hello world: rank 14 of 16 running on i72600K Hello world: rank 15 of 16 running on i72600K
Note, for some reason first attempt failed, second succeeded. Third attempt succeeded.
Any thoughts?
Jim Dempsey
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Do you have anything in the Windows Event Viewer indicating problems at the time of the failed run.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
If that shows up again, I will look in the logs (both systems).
I think that may have occurred the first time after a re-boot of one of the systems. I haven't seen it since.
I may also have to re-enable the Firewall logging too.
Jim Dempsey
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page