Hi Mehdi, The issue manifests itself as follows. If I have libpsm2-2/11.2.68-1 installed, any MPI program (whether a big program or a small MPI hello world example) has a 15-second delay when I run it. And for each rank the following error message is displayed:
cyberdyne.25917hfi_wait_for_device: The /dev/hfi1_0 device failed to appear after 15.0 seconds: Connection timed out (cyberdyne is the host name of my machine.) If I downgrade libpsm2-2 to 10.3.58-2 the MPI programs run instantly (without a 15 sec delay) and no error message about /dev/hfi1_0 is displayed. I'm on Debian testing with libopenmpi3/3.1.2-6. Thanks for looking into this and please let me know if you need more information. Thanks, Jonas -- Jonas Lippuner, PhD Scientist Computational Physics and Methods, CCS-2 Center for Theoretical Astrophysics Los Alamos National Laboratory jlippu...@lanl.gov 505-667-1646 http://jonaslippuner.com -----Original Message----- From: Mehdi Dogguy <me...@dogguy.org> To: "Lippuner, Jonas" <jlippu...@lanl.gov>, 910...@bugs.debian.org Subject: Re: Bug#910485: Confirm issue with libpsm2-2/11.2.68-1 Date: Tue, 16 Oct 2018 10:10:04 +0200 Hi Jonas, On 2018-10-15 19:54, Lippuner, Jonas wrote: > I'm having the same issue with libpsm2-2 version 11.2.68-1. > Downgrading > to 10.3.58-2 fixes it for me. Can you please explain how you experienced the bug? I've understood Drew's case, but maybe yours is slightly different.
