Re: [Denovoassembler-devel] Ray app in BaseSpace

[Boisvert, S. <boi...@an...>]

[I CC'ed devel mailing lists]

Can you provide the full command (cat RayCommand) ?

I have 2 ideas to troubleshoot (waiting 17 hours is not a good approach
to troubleshooting).

# Idea 1

Disabling the SM btl in Open-MPI's MCA (assuming you are using Open-MPI and not
MPICH).

Before:
mpiexec -n <ranks> Ray -k <k> -p <file1> <file2>

After:
mpiexec -n <ranks> --mca btl ^sm Ray -k <k> -p <file1> <file2>


# Idea 2

RayPlatform has built-in debugger (it is mentioned in the doc).

Before:
mpiexec -n <ranks> Ray -k <k> -p <file1> <file2>

After:
mpiexec -n <ranks>  Ray -debug -k <k> -p <file1> <file2>

This will give you plenty of information in the standard output.



My bet: SM in the cloud is slow / buggy and there is a bug in the runtime there
(below RayPlatform).