Molcas Forum

adutton

Member

Registered: 2017-03-24

Posts: 3

[SOLVED] MPI error trying to run a QMMM job with Tinker

Hello,

Has anyone seen this error and know anyway to fix it? 

 *** An error occurred in MPI_Allreduce
 *** reported by process [139974332776449,0]
*** on communicator MPI_COMM_WORLD
*** MPI_ERR_IN_STATUS: error code in status
*** MPI_ERRORS_ARE_FATAL (processes in this communicator will now abort,
***    and potentially your MPI job)
--- Stop Module:  scf at Fri Mar 24 13:56:25 2017 /rc=                   -1 (Unknown) ---

I am trying to run a QMMM job linked with Tinker to model TIP3P waters.  The job crashes after about an hour and during the slapaf optimization step.

Thanks

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,011

Re: [SOLVED] MPI error trying to run a QMMM job with Tinker

Does it consistently crash at the same spot? MPI errors are often difficult to catch, they could be due to bug in the MPI libraries, or a hardware/software glitch in one of the nodes causing the whole calculation to stop. If possible, try different compiler/MPI versions too.

adutton

Member

Registered: 2017-03-24

Posts: 3

Re: [SOLVED] MPI error trying to run a QMMM job with Tinker

This did turnout to be a problem with the openmpi and we solved it by creating some symlinks and using openmpi/1.10.0.  When we installed Molcas we used the pre-compiled version.  We are hoping that if we get the source code and compile it locally we can use the appropriate openmpi version.  But is working for now.