(Click to open topic with navigation)
This topic provides details regarding the file system configuration.
Users will normally submit Nitro jobs to a scheduler. A Nitro job runs in the user's workspace on the nodes allocated to the job by the scheduler. There are two Nitro output file locations (task log and job log) that must be available to the hosts running the workers and coordinator.
In this topic:
3.3.1 Nitro Launch Script Location
The Nitro launch script must be installed in a shared file location or a location that is accessible to the users that will be running Nitro jobs. While users won't need to run the Nitro application itself, they should have read access to the Nitro launch script.
The task file a Nitro job will use must be accessible to the job. Most schedulers will copy the user job script, but will not copy the task file the user must provide. Users may need to stage this task file into a location accessible to Nitro job hosts. Nitro only needs read access to this file.
If the user does not specify a Nitro job directory (using "‑‑job‑dir" command line option), Nitro defaults to the $HOME/nitro/<job id> directory.
This directory contains the Nitro job and task log files, Nitro diagnostic log files, and temporary files Nitro uses to recover a canceled job. Users' home directories on these hosts should be mapped to a shared file location accessible to the users either from the system's login hosts, or outside of the cluster. Nitro will need to create directories and files in the Nitro job directory.
3.3.4 Recommended NFS Settings
If the job directory and task file are located on an NFS, then depending on the NFS cache settings, system administrators and users should be aware of the following:
Related Topics