Skip to content

ALPS PLM component should not be used when using Cray XC nativized SLURM #1182

@hppritcha

Description

@hppritcha

SLURM has recently been enhanced with a set of plugins from Cray that allows for srun to be used for launching jobs on Cray XC systems. However, for software wishing to obtain RDMA credentials to use the HSN, the ALPS query/library methods were retained. This is fine, but it does cause issues if the ALPS PLM component is built. mpirun attempts to use aprun to launch the orted daemons and fails.

Turns out Cray has wlm detection software that can be used at runtime to determine whether the SLURM PLM or ALPS PLM should be used. This can be used in the ALPS PLM component to decide whether or not to be selected at job launch. Using this approach will allow for a site having both Cray XCs with aprun and srun to use the same Open MPI installation.

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

Milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions