Intel® MPI Library
Get help with building, analyzing, optimizing, and scaling high-performance computing (HPC) applications.
Announcements
FPGA community forums and blogs on community.intel.com are migrating to the new Altera Community and are read-only. For urgent support needs during this transition, please visit the FPGA Design Resources page or contact an Altera Authorized Distributor.

adding further compute nodes

Gert_Lindner
Beginner
1,848 Views

Hi,

Is there a need to re-install Intel Studio even in the case then I added further compute nodes to my cluster? There exists two infiniband -islands,  ibstat is:

CA 'mlx4_0', CA type: MT4099 and CA 'mlx4_1', CA type: MT26428. The latest  compute nodes are associated to MT4099.

These provider -errors are only present in the 'newer node context'

[2] MPI startup(): dapl fabric is not available and fallback fabric is not enabled
[10] MPI startup(): dapl fabric is not available and fallback fabric is not enabled
[12] MPI startup(): dapl fabric is not available and fallback fabric is not enabled
node009:UCM:2d97:570fa700: 1249 us(1249 us):  open_hca: device mlx4_0 not found
node009:UCM:2d9f:1626f700: 1262 us(1262 us):  open_hca: device mlx4_0 not found
node009:UCM:2da1:7f214700: 1102 us(1102 us):  open_hca: device mlx4_0 not found

 

Regards

Gert

0 Kudos
1 Reply
TimP
Honored Contributor III
1,848 Views
One way or another you must assure that all required shared libraries are available on all nodes. As far as impi is concerned, repeating the Mpi part of the installation with an updated node list would take care of it, (after taking care of dapl and hca). Evidently, installing psxe on a shared drive simplifies this aspect of adding nodes.
0 Kudos
Reply