Hi Howard/all, Here is the simple code I was talking about in the meeting today: // general high-level optimistic application void main() { MPI_Session session; MPI_Session_Init(MPI_INFO_NULL, MPI_ERRORS_ARE_FATAL, &session); MPI_Group group; MPI_Group_from_session_pset(session, "mpi://world", &group); MPI_Comm comm; MPI_Comm_create_from_group(group, &comm); ret = do_stuff_with_comm(comm); if (MPI_SUCCESS == ret) { MPI_Comm_disconnect(&comm); MPI_Session_Finalize(&session); break; } else { panic(); } } // general high-level pragmatic application void main() { // additional code while (1) { MPI_Session session; MPI_Session_Init(MPI_INFO_NULL, MPI_ERRORS_RETURN, &session); MPI_Group world, failed, group; MPI_Group_from_session_pset(session, "mpi://world", &world); // additional code MPI_Session_get_proc_failed(session, &failed); // new API, seems easy to do MPI_Group_difference(world, failed, &group); MPI_Group_free(&world); MPI_Group_free(&failed); MPI_Comm comm; MPI_Comm_create_from_group(group, &comm); // <-- the detail-devils live here MPI_Group_free(&group); ret = do_stuff_with_comm(comm); MPI_Comm_disconnect(&comm); MPI_Session_Finalize(&session); if (MPI_SUCCESS == ret) { break; // all done! } else if (MPI_ERR_PROC_FAILED == ret) { continue; // no more panic } // additional code } // end while } Best wishes, Dan. From: mpiwg-sessions <[email protected]> On Behalf Of Pritchard Jr., Howard via mpiwg-sessions Sent: Thursday, January 25, 2024 6:07 PM To: MPI Sessions working group <[email protected]> Cc: Pritchard Jr., Howard <[email protected]> Subject: [mpiwg-sessions] Sessions WG - meet 1/29/24 Hi Folks, Let's meet on 1/29 to continue discussions related to sessions and FT. I think what will help is to consider several use cases and implications. Here are some I have * App using sessions to init/finalize and create at least one initial communicator with MPI_Comm_create_from_group, but also wants to use methods available in slice1 of ULFM proposal to shrink/repair communicators. Are there any problems? * App using sessions to init/finalize and create at least one initial communicator with MPI_Comm_create_from_group, and wants to use methods available in slice 1 of ULFM proposal to create new group from a pset and create a new communicator * App using sessions to init/finalize, etc. and when a fail-stop error is detected destroy the session, create a new session query for process sets, etc. and start all over. We should also consider the behavior of MPI_Comm_create_from_group if a process failure occurs while creating a new communicator. The ULFM slice 1 discusses behavior of MPI_COMM_DUP and process failure. We'd probably want similar behavior for MPI_Comm_create_from_group. For those with access, the ULFM slice 1 PR is at https://github.com/mpi-forum/mpi-standard/pull/947 Thanks, Howard ------- [signature_61897647] Howard Pritchard Research Scientist HPC-ENV Los Alamos National Laboratory [email protected]<mailto:[email protected]> [signature_1672648044]<https://www.instagram.com/losalamosnatlab/>[signature_2067890307]<https://twitter.com/LosAlamosNatLab>[signature_1942525183]<https://www.linkedin.com/company/los-alamos-national-laboratory/>[signature_882949974]<https://www.facebook.com/LosAlamosNationalLab/>