Skip to content

IUSCA/rhythm

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

21 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Rhythm

Rhythm allows you to design and control workflows made of Celery tasks. A workflow is a sequence of steps to run one after the other. Rhythm simplifies the process of executing workflows consisting of long-running tasks with reliability.

The following are the features of Rhythm workflows:

  • If a workflow consisting of three steps (S1, S2, and S3) encounters a failure while executing S2 (even after retries by Celery), it is possible to resume the workflow later. Resuming the workflow with restart S2 with previous arguments and after its completion S3 will be run.
  • A workflow can be paused and resumed later.
  • You can keep track of which step is currently running, as well as its progress.

Installation

pip install sca-rhythm

see on pypi

Prerequisites

Celery app should be configured with a mongo database backend.

Create Tasks with WorkflowTask class

import os
import time

from celery import Celery

from sca_rhythm import WorkflowTask

app = Celery("tasks")


@app.task(base=WorkflowTask, bind=True)
def task1(self, batch_id, **kwargs):
    print(f'task - {os.getpid()} 1 starts with {batch_id}')
    # do work
    time.sleep(1)

    # update progress to result backend
    # sets the task's state as "PROGRESS"
    self.update_progress({
        'done': 2873,
        'total': 100000
    })

    # do some more work
    return batch_id, {'return_obj': 'foo'}

⚠️ Task Constraints ⚠️

  1. The task signature must contain **kwargs for the workflow orchestration to function.
  2. The return type must be of list / tuple type and the first element of the return value is sent to the next task as its argument.

Create Workflows with Workflow class

from celery import Celery

from sca_rhythm import Workflow

steps = [
    {
        'name': 'inspect',
        'task': 'tasks.inspect'
        'queue': 'q1',
    },
    {
        'name': 'archive',
        'task': 'tasks.archive',
        'queue': 'q1',
    },
    {
        'name': 'stage',
        'task': 'tasks.stage',
        'queue': 'q2',
        'priority': 5
    }
]

wf = Workflow(app, steps=steps, name='archive_batch', app_id='app')
wf.start('batch-id-test')

The provided code defines a workflow consisting of multiple steps, each representing a task to be executed in a specific order. The workflow is initiated with a unique identifier, and its steps are configured with task names, associated queues, and optional priorities.

Each step is represented as a dictionary with the following properties:

  • name: A descriptive name for the step.
  • task: The task to be executed, specified as a string containing the task's import path.
  • queue: The Celery queue to which the task should be sent.
  • priority (optional): An integer (between 0 and 9) indicating the priority of the task in the queue. If not provided, the priority is set to the step's position in the workflow. If there are more than 9 tasks, tasks in positions 10 and above will all recieve priority 9.

Priority Scheme: The priority scheme is designed to optimize the execution of tasks within the same workflow. Tasks with higher priorities are executed before those with lower priorities. If no priority is specified, the default priority is set to the step's position in the workflow. This scheme ensures that tasks within a workflow are executed sequentially with increasing priority, minimizing the likelihood of interweaving tasks from different workflows.

Pause and Resume Workflows

Pausing a workflow stop the current running task and resuming a workflow will restart the stopped task with the same arguments.

wf = Workflow(app, workflow_id='2f87decb-a431-472b-b26e-32c894993881')

wf.pause()

wf.resume()

Build & Publish

poetry install
poetry publish --build

Task Status

  • PENDING: Task state is unknown (assumed pending since you know the id).
  • STARTED: Task was started by a worker (task_track_started = True)
  • SUCCESS: Task succeeded
  • FAILURE: Task failed
  • REVOKED: Task was revoked
  • RETRY: Task is waiting for retry.

Workflow Status

The workflow status is a summative status that is determined by the status of the initial step that is not marked as " SUCCESS," which is referred to as a "pending step".

  • PENDING - the pending step is the first step in the workflow and its status is pending.
  • STARTED - the status of the pending step is one of STARTED, RETRY, PENDING.
  • REVOKED - the pending step was revoked, the workflow can be resumed.
  • FAILURE - the pending step was failed, the workflow can be resumed.
  • SUCCESS - all steps have succeeded.

Status Groups:

  • DONE: { SUCCESS, FAILURE, REVOKED }
  • ACTIVE: !DONE

Releases

No releases published

Packages

No packages published

Languages