Closed remyd1 closed 7 years ago
Hi Remy
thanks for the feedback. Not sure yet whether there is anything I can do on the side of pbmpi’s code to fix that, but I am taking note.
best regards
nicolas
Le 16 nov. 2016 à 15:26, DERNAT Rémy notifications@github.com a écrit :
Hi,
First of all, I am not sure this bug is related to pb_mpi, but anyway, you will be able to mention it in documentation.
So, JFYI, I have some issues with pb_mpi since I installed docker as daemon on our cluster. The problem is that mpi broadcast seems to send messages to other interfaces (like "lo", "docker0"). Then, pb_mpi seems to run, but in fact, it does not do anything.
I am now able to have a pb_mpi working again with:
mpirun --mca btl_tcp_if_exclude lo,docker0 -np .... pb_mpi .... instead of:
mpirun -np .... pb_mpi .... Kind regards, Remy
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/bayesiancook/pbmpi/issues/1, or mute the thread https://github.com/notifications/unsubscribe-auth/AHwIoFzXHnLczg3K2AMb_Q8C1ytuYuv6ks5q-xKtgaJpZM4Kz7QD.
Nicolas Lartillot Laboratoire Biométrie et Biologie Evolutive UMR CNRS 5558, Université Lyon 1 43 Bld du 11 Novembre 1918 69622 Villeurbanne cedex France
+33 (0)4 72 43 12 87
The Bayesian kitchen http://bayesiancook.blogspot.fr/
Hi,
First of all, I am not sure this bug is related to pb_mpi, but anyway, you will be able to mention it in documentation.
So, JFYI, I have some issues with pb_mpi since I installed docker as daemon on our cluster. The problem is that mpi broadcast seems to send messages to other interfaces (like "lo", "docker0"). Then, pb_mpi seems to run, but in fact, it does not do anything.
I am now able to have a pb_mpi working again with:
instead of:
Kind regards, Remy