The cluster has the following queues:
| Queue | Description |
|---|---|
| devel | Accessible by all users. Used for interactive and debugging use, short duration only. Initially 16 separate nodes. |
| urgent | NR users may submit to this queue if they feel their job is more urgent than some other NR jobs. They should be prepared to justify (if asked) on the mailing list, or to the Director, why the jobs are urgent. Non-NR users should ask on the mailing list first if there are objections to running in this queue, and say how much they expect to use and for how long. If consensus cannot be reached, the Director decides. |
| nr | Only for NR users |
| normal | Accessible by all users |
| fillin | Accessible by all users. Can run longer jobs but they may be preempted (killed) by jobs in higher priority queues. See Preemption. |
Minerva provides three separate filesystems. Initial disk quotas are:
| FS | Total size | Disk quota | Inode (file*1) limit |
|---|---|---|---|
| /home | 45 TB | 10 GB | 1 Million |
| /work | 90 TB | 100 GB | 4 Million |
| /scratch | 320 TB | None*2 | None*2 |
*1 directories count 3-fold; *2 provided disk space is available.
On /scratch, users are encouraged to delete or archive old data, either to /work or to Archive Storage. /scratch is not automatically purged, however users having data older than 3 months may get reminders to clean up.
Note: Only the home file system is backed up
/home is backed up nightly/scratch and /work should be considered volatile systems; data is not backed up, and is at risk of filesystem crashes. The AEI has access to a tape archiving system. This should be used for preserving and long-term archiving of data according to the AEI policy on storage of scientific data.