On 05/08/2018 05:33 PM, Christopher Samuel wrote: > On 09/05/18 10:23, Bill Broadley wrote: > >> It's possible of course that it's entirely an openmpi problem, I'll >> be investigating and posting there if I can't find a solution. > > One of the changes in OMPI 3.1.0 was: > > - Update PMIx to version 2.1.1. > > So I'm wondering if previous versions were falling back to PMIx v1 > support in PMIx v2 whereas now it's trying to use v2? > > Could explain that OPAL_PMIX_V1 definition in previous versions. > > What is your default MPI type? > > scontrol show config | fgrep MpiDefault > > If it's not pmix_v2 then you'll probably need to specify that, you can > test with "srun --mpi=pmix_v2".
Ha, very good catch, thank you: bill@headnode:~/relay$ srun --mpi=pmix_v2 -N 2 -n 2 -t 1 ./r2 1 c2-31 c2-33 size= 1, 16384 hops, 2 nodes in 0.02 sec ( 1.28 us/hop) 3046 KB/sec I updated the system wide MpiDefault and now it works great without the --mpi= It tends to be the simple things that get me. Thanks again.
