Here follow three questions linked to improvements for the MACCS processor wrapper.
The MACCS processor is run in parallel to the SLURM pool. Is there a reason to that ? Is it a lot of work to transform the wrapper to run it in a sbatch like the others ?
Apparently, there is a “working directory” option in the MACCS exe. Is it possible to link it to a user-defined parameter in the DB ? We are observing a I/O bottleneck on our servers and want to use a SSD space. We also want to use a volume shared via NFS for the output storage, larger than the space available on a single server (S2 images are quickly huge), and don’t want to saturate that volume with disk write. In brief, the ideal is a local computation on SSD, then a move to the remote volume.
We have found that MAJA can be put into a docker box. It is possible to wrap MACCS into one to make it available on other servers without CentOS7 ? (e.g. via a shared SLURM pool)