Partitions

Total PROTEUS nodes are grouped by common properties or functionality into several queues or partitions. These partitions are logical divisions and have nothing to do with processor families.

Partitions are a feature of the SLURM queue manager. Each partition has a set of attributes that define and limit the behaviour of the jobs running in it. Each node belongs to one or more partitions.

PARTITIONS IN PROTEUS

_

Currently, PROTEUS has 2 partitions, and these are differentiated by their maximum execution time:

TIMELIMIT DEFAULT TIME NODES
short 3 days 3 days metis (6)
htc-general* 90 days 90 days calypso, kratos, hermes, metis

htc-general is the default partition. If no partition is specified in the SLURM script, the job will go to that queue.

TIMELIMIT is the maximum execution time of a job. After this time, if it has not finished, the job is cancelled.

DEFAULT TIME is the default time assigned to the job. You can indicate a different time (always less than TIMELIMIT), if you have an estimate of the execution time. After this time is exceeded, the job is cancelled. Specifying an execution time can help SLURM to do a more efficient job scheduling.

The short partition is intended for short programs. Nodes have been reserved for this partition so that there is availability for these types of jobs and they do not have to wait for longer jobs.

The long-preempt partition, on the other hand, is appropriate for very long jobs (several months). It must be taken into account that the job must have some mechanism that allows it to pick up where it left off in case of failure (power failure, node failure, etc.).

The list of partitions and their properties can be consulted with:

sinfo

which returns the following columns:

  • PARTITION: partition name
  • AVAIL: whether available (up) or not (down)
  • TIMELIMIT: maximum execution time
  • NODES: number of nodes
  • STATE: status of these nodes. This can be:
    • idle: available
    • alloc: in use
    • mix: part available, part in use
    • resv: reserved
    • drain/drng: not available for technical reasons

This list of partitions may be modified in the future, so it is recommended to visit this page frequently.

_

Do you need more info?

_