mpi_abort does not terminate all processes under torque

mpi_abort does not terminate all processes under torque

We are running the Nasa Overflow code on a large linux cluster and have found that if the code calls MPI_ABORT it does not terminate as

expected.  We are running version 4.1.027 of Intel MPI.  We running under the Torque resource manager.

Bernie

3 posts / 0 new
Last post
For more complete information about compiler optimizations, see our Optimization Notice.

We have the same problem using IntelMPI 4.1.0.024. MPI_Abort hang until a newline is send.

Bernd

We have the same issue with PBS as job scheduler and mpi version 5.0.3.048.
So the code sends an MPI_ABORT and the processes are not killed correctly that the job hangs in the queue.

Is there a solution to this problem?

Sebastian

 

Leave a Comment

Please sign in to add a comment. Not a member? Join today