tardis.adapters.batchsystems.slurm module

SLURM Batch system Adapter

class tardis.adapters.batchsystems.slurm.SlurmAdapter[source]

Bases: BatchSystemAdapter

SlurmAdapter implements the TARDIS interface to dynamically integrate and manage opportunistic resources with the SLURM Batch System.

async disintegrate_machine(drone_uuid: str) → None[source]

SLURM does not require any specific disintegration procedure (at least in Freiburg).

Parameters: drone_uuid (str) – Uuid of the worker node, for some sites corresponding to the host name of the drone.
Returns: None

async drain_machine(drone_uuid: str) → None[source]

Drain a machine in the SLURM batch system, which means that no new jobs will be accepted

Parameters: drone_uuid (str) – Uuid of the worker node, for some sites corresponding to the host name of the drone.
Returns: None

async get_allocation(drone_uuid: str) → float[source]

Get the allocation of a worker node in SLURM, which is defined as maximum of the ratios of requested over total resources (CPU, Memory, Disk, etc.).

Parameters: drone_uuid (str) – Uuid of the worker node, for some sites corresponding to the host name of the drone.
Returns: The allocation of a worker node as described above.
Return type: float

async get_machine_status(drone_uuid: str) → MachineStatus[source]

Get the status of a worker node in SLURM (Available, Draining, Drained, NotAvailable)

Parameters: drone_uuid (str) – Uuid of the worker node, for some sites corresponding to the host name of the drone.
Returns: The machine status in SLURM (Available, Draining, Drained, NotAvailable)
Return type: MachineStatus

async get_resource_ratios(drone_uuid: str) → Iterable[float][source]

Get the ratio of requested over total resources for a worker node in Slurm. The CPU ratio is calculated as allocated CPUs / total CPUs and the memory ratio is given by allocated memory / total memory.

Parameters: drone_uuid (str) – Uuid of the worker node, for some sites corresponding to the host name of the drone.
Returns: Iterable of float containing the ratios
Return type: Iterable[float]

async get_utilisation(drone_uuid: str) → float[source]

Get the utilization of a worker node in Slurm, which is defined as minimum of the ratios of requested over total resources (CPU, Memory, Disk, etc.).

Parameters: drone_uuid (str) – Uuid of the worker node, for some sites corresponding to the host name of the drone.
Returns: The utilization of a worker node as described above.
Return type: float

async integrate_machine(drone_uuid: str) → None[source]

SLURM does not require any specific integration procedure (if the Drones take care of it themselves)

Parameters: drone_uuid (str) – Uuid of the worker node, for some sites corresponding to the host name of the drone.
Returns: None

property machine_meta_data_translation_mapping: AttributeDict

The machine meta data translation mapping is used to translate units of the machine meta data in TARDIS to values expected by the Slurm batch system adapter.

Returns: Machine meta data translation mapping
Return type: AttributeDict

async tardis.adapters.batchsystems.slurm.slurm_status_updater(options: AttributeDict, attributes: AttributeDict, executor: Executor) → dict[source]

Slurm status update.

Parameters

options (AttributeDict) – Additional parameters for the sinfo command
attributes (AttributeDict) – Formatting options for sinfo

Returns

Dictionary containing the output of the sinfo command

Return type

dict