In our lab we have MaterialStudio with Castep running on a Linux cluster with SLURM as a queuing system.
Everything works fine, but we are not able to get Castep running on multiple nodes, neither if started from the MaterialStudio interface nor as standalone version.
It's possible to allocate more than one node to the process, but Castep then only uses one of the allocated nodes.
To start the standalone version, we use this command:
\$(tsk) = number of tasks per node
and \$(np) = number of processes for Castep
I also included a sinfo.txt file, with information about the cluster and a gw_info.txt file, with the Gateway setup for MaterialStudio.
Help is very much appreciated.
Albert
Everything works fine, but we are not able to get Castep running on multiple nodes, neither if started from the MaterialStudio interface nor as standalone version.
It's possible to allocate more than one node to the process, but Castep then only uses one of the allocated nodes.
To start the standalone version, we use this command:
sbatch --nodes \$(nodes) --tasks-per-node \$(tsk) --time \$(time) RunCASTEP.sh -np \$(np)
where, \$(nodes) = number of allocated nodes\$(tsk) = number of tasks per node
and \$(np) = number of processes for Castep
I also included a sinfo.txt file, with information about the cluster and a gw_info.txt file, with the Gateway setup for MaterialStudio.
Help is very much appreciated.
Albert
