Basically, pooled scratch space works by following this order of precedence:
- the first scratch disk in the list will be used to volume capacity,
- then the next scratch disk in the node confirmation file pool will be used to volume capacity and, so forth, through all defined scratch spaces in the pool,
- When all scratch disks defined on each node are consumed, $TMPDIR is used as the scratch disk to volume capacity, and
- When $TMPDIR used to volume capacity, then the /tmp location is used.