Configuration

Use the rjm_config command to configure RemoteJobManager, which will walk through the configuration options and write the configuration file.

By default rjm_config configures the Globus stack (Globus Transfer + Globus Compute). To configure the experimental SSH/SFTP (Paramiko) backend instead, pass -s / --ssh:

rjm_config --ssh

The configuration file lives at ~/.rjm/rjm_config.ini. Its main sections are:

• [COMPONENTS] selects the backend (runner and transferer).

• [GLOBUS_TRANSFER] and [GLOBUS_COMPUTE] hold the Globus endpoint ids and remote path for the Globus stack.

• [PARAMIKO] holds the SSH key path, remote address, remote user, remote base path, and job script for the SSH stack.

• [POLLING] controls how often RJM polls the remote for job state.

• [SLURM] sets the Slurm script filename (used by the Globus stack) and, optionally, a default account that RJM passes to sbatch --account when the Slurm script does not already specify one.

• [FILES], [RETRY] and [LOGGING] control upload/download lists, retry behaviour, and per-logger log levels.

Globus authentication tokens are cached at ~/.rjm/rjm_tokens.json and are not used by the Paramiko backend.

Default Slurm account

When the Globus stack is configured via rjm_config, the NeSI project code entered at the prompt is persisted as [SLURM] account in rjm_config.ini. At job submission time, the Globus Slurm runner passes this value to sbatch as --account=<account>, so the Slurm script no longer needs to set the account itself.

If the Slurm script already specifies an account (either via #SBATCH --account=<value> or #SBATCH -A <value>), the directive in the script takes precedence and RJM does not add --account to the sbatch command. To change the default account for all jobs, edit [SLURM] account in rjm_config.ini or rerun rjm_config. The key is optional, so removing it (or leaving it blank) restores the previous behaviour of relying entirely on the Slurm script.