SSH particle caching on a cluster-only environment

We have a CryoSPARC deployment that only allows for cluster runs via Slurm. Individual compute nodes are not registered as workers.

I saw the “–ssdpath” and “–ssdreserve” options in worker registration step, but we don’t connect individual nodes in this way.

Is there a way to configure SSD caching from the master only? Are there env variables that correspond to these options in the master’s config.sh perhaps?

Thanks in advance!

Hi,

You probably want to set cache_path in cluster_info.json (examples) when registering your scheduler as a cluster lane (reference). This can include an environment variable associated with the submission if cache staging is managed by prolog/epilog.

Cheers,
Yang

1 Like