xqute

A job management system for Python, designed to simplify job scheduling and execution with support for multiple schedulers and plugins.

Features

• Written in async for high performance
• Plugin system for extensibility
• Scheduler adaptor for various backends
• Job retrying and pipeline halting on failure
• Support for cloud-based working directories
• Built-in support for Google Batch Jobs, Slurm, SGE, SSH, and container schedulers

Installation

pip install xqute

A Toy Example

import asyncio
from xqute import Xqute

async def main():
    # Initialize Xqute with 3 jobs allowed to run concurrently
    xqute = Xqute(forks=3)
    for _ in range(10):
        await xqute.put(['sleep', '1'])
    await xqute.run_until_complete()

if __name__ == '__main__':
    asyncio.run(main())

API Documentation

Full API documentation is available at: https://pwwang.github.io/xqute/

Usage

Xqute Object

An Xqute object is initialized as follows:

xqute = Xqute(...)

Available arguments are:

• scheduler: The scheduler class or name (default: local)
• plugins: Plugins to enable/disable for this session
• workdir: Directory for job metadata (default: ./.xqute/)
• forks: Number of jobs allowed to run concurrently
• error_strategy: Strategy for handling errors (e.g., halt, retry)
• num_retries: Maximum number of retries when error_strategy is set to retry
• submission_batch: Number of jobs to submit in a batch
• scheduler_opts: Additional keyword arguments for the scheduler
• jobname_prefix: Prefix for job names
• recheck_interval: Interval (in seconds) to recheck job status

Note: The producer must be initialized within an event loop.

To add a job to the queue:

await xqute.put(['echo', 'Hello, World!'])

Using SGE Scheduler

xqute = Xqute(
    scheduler='sge',
    forks=100,
    scheduler_opts={
        'qsub': '/path/to/qsub',
        'qdel': '/path/to/qdel',
        'qstat': '/path/to/qstat',
        'q': '1-day',  # or qsub_q='1-day'
    }
)

Keyword arguments starting with sge_ are interpreted as qsub options. For example:

'l': ['h_vmem=2G', 'gpu=1']

will be expanded in the job script as:

#$ -l h_vmem=2G
#$ -l gpu=1

Using Slurm Scheduler

xqute = Xqute(
    scheduler='slurm',
    forks=100,
    scheduler_opts={
        'sbatch': '/path/to/sbatch',
        'scancel': '/path/to/scancel',
        'squeue': '/path/to/squeue',
        'partition': '1-day',
        'time': '01:00:00',
    }
)

Using SSH Scheduler

xqute = Xqute(
    scheduler='ssh',
    forks=100,
    scheduler_opts={
        'ssh': '/path/to/ssh',
        'servers': {
            'server1': {
                'user': 'username',
                'port': 22,
                'keyfile': '/path/to/keyfile',
                'ctrl_persist': 600,
                'ctrl_dir': '/tmp',
            }
        }
    }
)

Note: SSH servers must share the same filesystem and use keyfile authentication.

Using Google Batch Jobs Scheduler

xqute = Xqute(
    scheduler='gbatch',
    forks=100,
    scheduler_opts={
        'project': 'your-gcp-project-id',
        'location': 'us-central1',
        'gcloud': '/path/to/gcloud',
        'taskGroups': [ ... ],
    }
)

Using Container Scheduler

xqute = Xqute(
    scheduler='container',
    forks=100,
    scheduler_opts={
        'image': 'docker://bash:latest',
        'entrypoint': '/usr/local/bin/bash',
        'bin': 'docker',
        'volumes': '/host/path:/container/path',
        'envs': {'MY_ENV_VAR': 'value'},
        'remove': True,
        'bin_args': ['--hostname', 'xqute-container'],
    }
)

Plugins

To create a plugin for xqute, implement the following hooks:

• def on_init(scheduler): Called after the scheduler is initialized
• def on_shutdown(scheduler, sig): Called when the scheduler shuts down
• async def on_job_init(scheduler, job): Called when a job is initialized
• async def on_job_queued(scheduler, job): Called when a job is queued
• async def on_job_submitted(scheduler, job): Called when a job is submitted
• async def on_job_started(scheduler, job): Called when a job starts running
• async def on_job_polling(scheduler, job, counter): Called during job status polling
• async def on_job_killing(scheduler, job): Called when a job is being killed
• async def on_job_killed(scheduler, job): Called when a job is killed
• async def on_job_failed(scheduler, job): Called when a job fails
• async def on_job_succeeded(scheduler, job): Called when a job succeeds
• def on_jobcmd_init(scheduler, job) -> str: Called during job command initialization
• def on_jobcmd_prep(scheduler, job) -> str: Called before the job command runs
• def on_jobcmd_end(scheduler, job) -> str: Called after the job command completes

To implement a hook, use the simplug plugin manager:

from xqute import simplug as pm

@pm.impl
def on_init(scheduler):
    ...

Implementing a Scheduler

To create a custom scheduler, subclass the Scheduler abstract class and implement the following methods:

from xqute import Scheduler

class MyScheduler(Scheduler):
    name = 'mysched'

    async def submit_job(self, job):
        """Submit a job and return its unique ID."""

    async def kill_job(self, job):
        """Kill a job."""

    async def job_is_running(self, job):
        """Check if a job is running."""