Partition

Note

This supersedes the pyslurm.partition class, which will be removed in a future release

pyslurm.Partition

A Slurm partition.

Setting Memory related attributes

Unless otherwise noted, all attributes in this class representing a memory value, like default_memory_per_cpu, may also be set with a string that contains suffixes like "K", "M", "G" or "T".

For example:

default_memory_per_cpu = "10G"

This will internally be converted to 10240 (how the Slurm API expects it)

Parameters:

Name	Type	Description	Default
name	str	Name of a Partition	None
**kwargs	Any	Every attribute of a Partition can be set, except for: total_cpus total_nodes select_type_parameters	required

Attributes:

Name	Type	Description
name	str	Name of the Partition.
allowed_submit_nodes	list[str]	List of Nodes from which Jobs can be submitted to the partition.
allowed_accounts	list[str]	List of Accounts which are allowed to execute Jobs
allowed_groups	list[str]	List of Groups which are allowed to execute Jobs
allowed_qos	list[str]	List of QoS which are allowed to execute Jobs
alternate	str	Name of the alternate Partition in case a Partition is down.
select_type_parameters	list[str]	List of Select type parameters for the select plugin.
cpu_binding	str	Default CPU-binding for Jobs that execute in a Partition.
default_memory_per_cpu	int	Default Memory per CPU for Jobs in this Partition, in Mebibytes. Mutually exclusive with default_memory_per_node. This can also return UNLIMITED
default_memory_per_node	int	Default Memory per Node for Jobs in this Partition, in Mebibytes. Mutually exclusive with default_memory_per_cpu. This can also return UNLIMITED
max_memory_per_cpu	int	Max Memory per CPU allowed for Jobs in this Partition, in Mebibytes. Mutually exclusive with max_memory_per_node. This can also return UNLIMITED
max_memory_per_node	int	Max Memory per Node allowed for Jobs in this Partition, in Mebibytes. Mutually exclusive with max_memory_per_cpu This can also return UNLIMITED
default_time	int	Default run time-limit in minutes for Jobs that don't specify one. This can also return UNLIMITED
denied_qos	list[str]	List of QoS that cannot be used in a Partition
denied_accounts	list[str]	List of Accounts that cannot use a Partition
preemption_grace_time	int	Grace Time in seconds when a Job is selected for Preemption.
default_cpus_per_gpu	int	Default CPUs per GPU for Jobs in this Partition
default_memory_per_gpu	int	Default Memory per GPU, in Mebibytes, for Jobs in this Partition
max_cpus_per_node	int	Max CPUs per Node allowed for Jobs in this Partition This can also return UNLIMITED
max_cpus_per_socket	int	Max CPUs per Socket allowed for Jobs in this Partition This can also return UNLIMITED
max_nodes	int	Max number of Nodes allowed for Jobs This can also return UNLIMITED
min_nodes	int	Minimum number of Nodes that must be requested by Jobs
max_time	int	Max Time-Limit in minutes that Jobs can request This can also return UNLIMITED
oversubscribe	str	The oversubscribe mode for this Partition
nodes	str	Nodes that are in a Partition
nodesets	list[str]	List of Nodesets that a Partition has configured
over_time_limit	int	Limit in minutes that Jobs can exceed their time-limit This can also return UNLIMITED
preempt_mode	str	Preemption Mode in a Partition
priority_job_factor	int	The Priority Job Factor for a partition
priority_tier	int	The priority tier for a Partition
qos	str	A QoS associated with a Partition, used to extend possible limits
total_cpus	int	Total number of CPUs available in a Partition
total_nodes	int	Total number of nodes available in a Partition
state	str	State the Partition is in
is_default	bool	Whether this Partition is the default partition or not
allow_root_jobs	bool	Whether Jobs by the root user are allowed
is_user_exclusive	bool	Whether nodes will be exclusively allocated to users
is_hidden	bool	Whether the partition is hidden or not
least_loaded_nodes_scheduling	bool	Whether Least-Loaded-Nodes scheduling algorithm is used on a Partition
is_root_only	bool	Whether only root is able to use a Partition
requires_reservation	bool	Whether a reservation is required to use a Partition
power_down_on_idle	bool	Whether nodes power down on idle after running jobs

create() method descriptor

Create a Partition.

Implements the slurm_create_partition RPC.

Returns:

Type	Description
Partition	This function returns the current Partition instance object itself.

Raises:

Type	Description
RPCError	If creating the Partition was not successful.

Examples:

>>> import pyslurm
>>> part = pyslurm.Partition("debug").create()

delete() method descriptor

Delete a Partition.

Implements the slurm_delete_partition RPC.

Raises:

Type	Description
RPCError	When deleting the Partition was not successful.

Examples:

>>> import pyslurm
>>> pyslurm.Partition("normal").delete()

load(name) staticmethod

Load information for a specific Partition.

Parameters:

Name	Type	Description	Default
name	str	The name of the Partition to load.	required

Returns:

Type	Description
Partition	Returns a new Partition instance.

Raises:

Type	Description
RPCError	If requesting the Partition information from the slurmctld was not successful.

Examples:

>>> import pyslurm
>>> part = pyslurm.Partition.load("normal")

modify(changes) method descriptor

Modify a Partition.

Implements the slurm_update_partition RPC.

Parameters:

Name	Type	Description	Default
changes	Partition	Another Partition object that contains all the changes to apply. Check the Other Parameters of the Partition class to see which properties can be modified.	required

Raises:

Type	Description
RPCError	When updating the Partition was not successful.

Examples:

>>> import pyslurm
>>>
>>> part = pyslurm.Partition.load("normal")
>>> # Prepare the changes
>>> changes = pyslurm.Partition(state="DRAIN")
>>> # Apply the changes to the "normal" Partition
>>> part.modify(changes)

to_dict() method descriptor

Partition information formatted as a dictionary.

Returns:

Type	Description
dict	Partition information as dict

Examples:

>>> import pyslurm
>>> mypart = pyslurm.Partition.load("mypart")
>>> mypart_dict = mypart.to_dict()

pyslurm.Partitions

Bases: pyslurm.xcollections.MultiClusterMap

A Multi Cluster collection of pyslurm.Partition objects.

Parameters:

Name	Type	Description	Default
partitions	Union[list[str], dict[str, Partition], str]	Partitions to initialize this collection with.	None

Attributes:

Name	Type	Description
total_cpus	int	Total amount of CPUs the Partitions in a Collection have
total_nodes	int	Total amount of Nodes the Partitions in a Collection have

load() staticmethod

Load all Partitions in the system.

Returns:

Type	Description
Partitions	Collection of Partition objects.

Raises:

Type	Description
RPCError	When getting all the Partitions from the slurmctld failed.

modify(changes) method descriptor

Modify all Partitions in a Collection.

Parameters:

Name	Type	Description	Default
changes	Partition	Another Partition object that contains all the changes to apply. Check the Other Parameters of the Partition class to see which properties can be modified.	required

Raises:

Type	Description
RPCError	When updating at least one Partition failed.

Examples:

>>> import pyslurm
>>>
>>> parts = pyslurm.Partitions.load()
>>> # Prepare the changes
>>> changes = pyslurm.Partition(state="DRAIN")
>>> # Apply the changes to all the partitions
>>> parts.modify(changes)

reload() method descriptor

Reload the information for Partitions in a collection.

Note

Only information for Partitions which are already in the collection at the time of calling this method will be reloaded.

Returns:

Type	Description
Partitions	Returns self

Raises:

Type	Description
RPCError	When getting the Partitions from the slurmctld failed.