- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
The latest version of Intel Cluster Studio-2013.1.046 is installed on a Cent OS 6.5 Linux system having two Intel Xeon Phi MIC 7120P cards. A sample helloworld (host only, No mic) mpi program throws following error:
$ mpiexec.hydra -genv I_MPI_DEBUG 5 -np 4 ./hellompi
[mpiexec@mac2.host1.com] HYDU_getfullhostname (./utils/others/others.c:136): getaddrinfo error (hostname: mac2.host1.com, error: Name or service not known)
[mpiexec@mac2.host1.com] HYDU_sock_create_and_listen_portstr (./utils/sock/sock.c:999): unable to get local hostname
[mpiexec@mac2.host1.com] HYD_pmci_launch_procs (./pm/pmiserv/pmiserv_pmci.c:313): unable to create PMI port
[mpiexec@mac2.host1.com] main (./ui/mpich/mpiexec.c:877): process manager returned error launching processes
What could be the issue here? Is this a bug with Intel MPI?
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
What do you get when you run hostname on this system? What if you ssh to mac2.host1.com, does this work?
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
First you need to fix your hosts.
vi /etc/hosts
add your hostname (in your case, mac2.host1.com) to the 127.0.0.1 line.
Example:
127.0.0.1 localhost mac2.host1.com
You should now be able to ssh to mac2.host1.com. If not, restart your networking (the way you do this varies on what OS you are using).
PS: I would recommend that you watch the introductory Intel MPI videos from this link: https://software.intel.com/en-us/intel-mpi-library
You will lean that Intel's MPI relies on networking. Therefore, this is not a bug.
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page