Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

"Assertion failed in file src/mpid/ch4/src/init_comm.c at line 72: in_use == 0" fails #7200

Open
colleeneb opened this issue Nov 7, 2024 · 0 comments · May be fixed by #7201
Open

"Assertion failed in file src/mpid/ch4/src/init_comm.c at line 72: in_use == 0" fails #7200

colleeneb opened this issue Nov 7, 2024 · 0 comments · May be fixed by #7201
Labels

Comments

@colleeneb
Copy link
Collaborator

colleeneb commented Nov 7, 2024

Multiple applications running on Aurora have seen the following fail:

Assertion failed in file src/mpid/ch4/src/init_comm.c at line 72: in_use == 0
/opt/aurora/24.180.1/CNDA/mpich/20240717/mpich-ofi-all-icc-default-pmix-gpu-drop20240717/lib/libmpi.so.12(+0x1d86514) [0x14763ac07514]
/opt/aurora/24.180.1/CNDA/mpich/20240717/mpich-ofi-all-icc-default-pmix-gpu-drop20240717/lib/libmpi.so.12(+0x1c128b1) [0x14763aa938b1]
/opt/aurora/24.180.1/CNDA/mpich/20240717/mpich-ofi-all-icc-default-pmix-gpu-drop20240717/lib/libmpi.so.12(+0x1c3c49a) [0x14763aabd49a]
/opt/aurora/24.180.1/CNDA/mpich/20240717/mpich-ofi-all-icc-default-pmix-gpu-drop20240717/lib/libmpi.so.12(+0x1d38332) [0x14763abb9332]
/opt/aurora/24.180.1/CNDA/mpich/20240717/mpich-ofi-all-icc-default-pmix-gpu-drop20240717/lib/libmpi.so.12(+0x1d0486b) [0x14763ab8586b]
/opt/aurora/24.180.1/CNDA/mpich/20240717/mpich-ofi-all-icc-default-pmix-gpu-drop20240717/lib/libmpi.so.12(+0x1c3c75c) [0x14763aabd75c]
/opt/aurora/24.180.1/CNDA/mpich/20240717/mpich-ofi-all-icc-default-pmix-gpu-drop20240717/lib/libmpi.so.12(+0x18bc6c6) [0x14763a73d6c6]
/opt/aurora/24.180.1/CNDA/mpich/20240717/mpich-ofi-all-icc-default-pmix-gpu-drop20240717/lib/libmpi.so.12(+0x18bbfbb) [0x14763a73cfbb]
/opt/aurora/24.180.1/CNDA/mpich/20240717/mpich-ofi-all-icc-default-pmix-gpu-drop20240717/lib/libmpi.so.12(+0x18ad199) [0x14763a72e199]
/opt/aurora/24.180.1/CNDA/mpich/20240717/mpich-ofi-all-icc-default-pmix-gpu-drop20240717/lib/libmpi.so.12(+0x18dad7c) [0x14763a75bd7c]
/opt/aurora/24.180.1/CNDA/mpich/20240717/mpich-ofi-all-icc-default-pmix-gpu-drop20240717/lib/libmpi.so.12(+0x18dab99) [0x14763a75bb99]
/opt/aurora/24.180.1/CNDA/mpich/20240717/mpich-ofi-all-icc-default-pmix-gpu-drop20240717/lib/libmpi.so.12(MPI_Init+0x6) [0x147638f38216]
./a.out() [0x40161c]
/lib64/libc.so.6(__libc_start_main+0xef) [0x14763891524d]
./a.out() [0x40154a]
Abort(1) on node 40128: Internal error
x4609c2s6b0n0.hostmgmt2609.cm.aurora.alcf.anl.gov: rank 40128 exited with code 1
x4609c2s6b0n0.hostmgmt2609.cm.aurora.alcf.anl.gov: rank 40129 died from signal 15

The smallest reproducer is

#include <mpi.h>
#include <stdio.h>

int main(int argc, char** argv) {
    // Initialize the MPI environment                                                                                                                                                                                                                  
    MPI_Init(NULL, NULL);

    // Finalize the MPI environment.                                                                                                                                                                                                                   
    MPI_Finalize();
}

It does not seem to be reproducible at 1 or 4 nodes but we've seen it at 4096 nodes however. (Adding @KennethEJansen since he reported it)

raffenet added a commit to raffenet/mpich that referenced this issue Nov 7, 2024
If the special init comm used for roots only address exchange still has
pending operations at destroy time, it could cause an assertion failure
during MPI_INIT. Instead, we should wait for pending ops to complete to
avoid a crash. Fixes pmodels#7200.
raffenet added a commit to raffenet/mpich that referenced this issue Nov 7, 2024
If the special init comm used for roots only address exchange still has
pending operations at destroy time, it could cause an assertion failure
during MPI_INIT. Instead, we should wait for pending ops to complete to
avoid a crash. Fixes pmodels#7200.
raffenet added a commit to raffenet/mpich that referenced this issue Nov 7, 2024
If the special init comm used for roots only address exchange still has
pending operations at destroy time, it could cause an assertion failure
during MPI_INIT. Instead, we should wait for pending ops to complete to
avoid a crash. Fixes pmodels#7200.
raffenet added a commit to raffenet/mpich that referenced this issue Nov 7, 2024
If the special init comm used for roots only address exchange still has
pending operations at destroy time, it could cause an assertion failure
during MPI_INIT. Instead, we should wait for pending ops to complete to
avoid a crash. Fixes pmodels#7200.
raffenet added a commit to raffenet/mpich that referenced this issue Nov 7, 2024
If the special init comm used for roots only address exchange still has
pending operations at destroy time, it could cause an assertion failure
during MPI_INIT. Instead, we should wait for pending ops to complete to
avoid a crash. Fixes pmodels#7200.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

Successfully merging a pull request may close this issue.

1 participant