Grid Queueing

Interface for configuring and submitting jobs on a computing cluster. Do not use this class directly. Instead use one of the batch plugins that inherit from this class.

iceprod.server.grid.get_host()

Cache the host fqdn for 1 hour

class iceprod.server.grid.GridTask(*args, **kwargs)

Protocol for grid task dataclass

dataset_id: str | None

task_id: str | None

instance_id: str | None

class iceprod.server.grid.BaseGrid(cfg, rest_client, cred_client)

Interface for a generic job distribution system. Do not use this class directly. Use one of the plugins.

async run()

Override the ActiveJobs and JobActions batch submission / monitoring classes

async dataset_lookup(dataset_id: str) → Dataset

Lookup a dataset fromn the REST API.

Uses caching with TTL 60 seconds.

Parameters:

dataset_id – dataset id

Returns:

iceprod.core.config.Dataset object

async get_tasks_to_queue(num: int) → list[Task]

Get new tasks to queue from the REST API.

Parameters:

num – number of tasks to retrieve

Returns:

list of iceprod.core.config.Task objects

async get_tasks_on_queue() → list

Get all tasks that are “assigned” to this queue.

Returns:

list of tasks

async task_idle(task: GridTask)

Tell IceProd API a task is now idle on the queue (put back to “queue” status).

Parameters:

task – IceProd task info

async task_processing(task: GridTask, site: str | None = None)

Tell IceProd API a task is now processing (put in the “processing” status).

Parameters:

• task – IceProd task info

• site – computing site the task is running at

async task_reset(task: GridTask, reason: str | None = None, stats: dict | None = None)

Tell IceProd API a task should be reset back to the “waiting” status.

Parameters:

• task – IceProd task info

• stats – task resource statistics

• reason – A reason for failure

async task_failure(task: GridTask, reason: str | None = None, stats: dict | None = None, stdout: Path | None = None, stderr: Path | None = None)

Tell IceProd API a task should be put in the “failed” status.

Stats format:

{

resources: {cpu: 1, gpu: 1, …}, …

}

Parameters:

• task – IceProd task info

• reason – A reason for failure

• stats – task resource statistics

• stdout – path to stdout file

• stderr – path to stderr file

async task_success(task: GridTask, stats: dict | None = None, stdout: Path | None = None, stderr: Path | None = None)

Tell IceProd API a task was successfully completed.

Parameters:

• task – IceProd task info

• stats – task resource statistics

• stdout – path to stdout file

• stderr – path to stderr file