On 2/10/14, 10:47 PM, "Orion Poplawski" <[email protected]> wrote:
On 02/06/2014 09:10 PM, Balaji, Pavan wrote:
Thanks. That’s very useful analysis. Would you be willing to try the attached patch to see if it solves this issue?
— Pavan
Well, it seems to prevent a hang (although I'm also updating from 3.0.4 to 3.1rc3 so not sure what is all changing here), but it does not run:
It might be easier to use the nightly snapshots to make sure you are not missing some fixes: http://www.mpich.org/static/tarballs/nightly/master/mpich/ The patch I sent, as well as a few other patches after 3.1rc3, are all included in the nightly snapshots.
============================ Fatal error in MPI_Init: Other MPI error, error stack: MPIR_Init_thread(467)..............: MPID_Init(177).....................: channel initialization failed MPIDI_CH3_Init(70).................: MPID_nem_init(319).................: MPID_nem_tcp_init(171).............: MPID_nem_tcp_get_business_card(418): MPID_nem_tcp_init(377).............: gethostbyname failed, i-00001ff8 (errno 1) Fatal error in MPI_Init: Other MPI error, error stack: MPIR_Init_thread(467)..............: MPID_Init(177).....................: channel initialization failed MPIDI_CH3_Init(70).................: MPID_nem_init(319).................: MPID_nem_tcp_init(171).............: MPID_nem_tcp_get_business_card(418): MPID_nem_tcp_init(377).............: gethostbyname failed, i-00001ff8 (errno 1)
That’s really weird. Errno 1 is "permission denied”. I don’t know how that’s happening with gethostbyname. Can you send your mpiexec command line and a small program that reproduces this error? E.g., if a program that just does MPI_INIT/MPI_FINALIZE reproduces this error, that’ll be best. — Pavan