mpich icon indicating copy to clipboard operation
mpich copied to clipboard

hydra: sessions not correctly tracked in pbs environments

Open mpichbot opened this issue 9 years ago • 2 comments

Originally by balaji on 2013-04-19 02:08:49 -0500


We do not correctly track session IDs on the pbs_mom node with Hydra in pbs environments. We use tm_spawn() even for launching the proxy on the local node, causing it to be launched as a separate process session.

The attached file (provided by Bharath Ramesh @ Virginia Tech) gives more information on the error.

mpichbot avatar Oct 14 '16 17:10 mpichbot

Originally by balaji on 2013-04-19 02:09:00 -0500


Attachment added: mpi_ps_axf.out (4.3 KiB)

mpichbot avatar Oct 14 '16 17:10 mpichbot

We use tm_spawn to launch hydra_pmi_proxy, which seems to always launch into a new process session. I guess we need check whether the proxy is on the same node as mpiexec and use fork instead when it's on the local node.

hzhou avatar Aug 10 '22 22:08 hzhou