Intel® MPI Library
Get help with building, analyzing, optimizing, and scaling high-performance computing (HPC) applications.
2161 Discussions

Installation Problem - MOST of the Cluster is Good but Phi Nodes Not Working

Tomás_P_
Beginner
275 Views

I'm hoping you can help with the best way to fix this installation problem:

We have a centrally-installed Intel Cluster Studio XE running on an HP DL380 server and 128 compute nodes - HP SL230s servers.  At the time of initial product installation, four additional compute nodes, SL230s with Xeon Phi processors were not included in this installation.

At this time, the compilers are working on all 128 nodes and /opt/intel file system is NFS mounted to all nodes from the central installation server.  We also have the flexlm license server running on our master system and the compilers are working.  Here is the only problem and I need the least destructive way to fix this:

On the four "fat" nodes with Xeon Phi processors, we are missing binaries in /opt/intel/mic/bin - right now, only the one micnativeloadex link is in there.  The other binaries are not present.

So what is the best way to fix this?  Do I re-run the central installation and include the four systems in the machines.LINUX file?  Will this cause problems with the other systems that are already installed and compilers functional?

Thank you in advance for your help!

 

0 Kudos
0 Replies
Reply