erpnext-syntax-scheduler（erpnext 语法调度程序）

ERPNext Syntax: Scheduler & Background Jobs

Deterministic syntax reference for Frappe scheduler events and background job processing.

Quick Reference

Scheduler Events (hooks.py)

# hooks.py
scheduler_events = {
    "all": ["myapp.tasks.every_tick"],
    "hourly": ["myapp.tasks.hourly_task"],
    "daily": ["myapp.tasks.daily_task"],
    "weekly": ["myapp.tasks.weekly_task"],
    "monthly": ["myapp.tasks.monthly_task"],
    "daily_long": ["myapp.tasks.heavy_daily"],  # Long queue
    "cron": {
        "0 9 * * 1-5": ["myapp.tasks.weekday_9am"],
        "*/15 * * * *": ["myapp.tasks.every_15_min"]
    }
}

CRITICAL: After EVERY change to scheduler_events: bench migrate

frappe.enqueue Basics

# Simple
frappe.enqueue("myapp.tasks.process", customer="CUST-001")

# With queue and timeout
frappe.enqueue(
    "myapp.tasks.heavy_task",
    queue="long",
    timeout=3600,
    param="value"
)

# With deduplication (v15)
from frappe.utils.background_jobs import is_job_enqueued

job_id = f"import::{doc.name}"
if not is_job_enqueued(job_id):
    frappe.enqueue("myapp.tasks.import_data", job_id=job_id, doc=doc.name)

Scheduler Event Types

Version difference scheduler tick:

• v14: ~240 seconds (4 min)
• v15: ~60 seconds

Queue Types

frappe.enqueue Parameters

frappe.enqueue(
    method,                      # REQUIRED: function or module path
    queue="default",             # Queue name
    timeout=None,                # Override timeout (seconds)
    is_async=True,               # False = execute directly
    now=False,                   # True = via frappe.call()
    job_id=None,                 # v15: unique ID for deduplication
    enqueue_after_commit=False,  # Wait for DB commit
    at_front=False,              # Place at front of queue
    on_success=None,             # Success callback
    on_failure=None,             # Failure callback
    **kwargs                     # Arguments for method
)

Job Deduplication

v15+ (Recommended)

from frappe.utils.background_jobs import is_job_enqueued

job_id = f"process::{doc.name}"
if not is_job_enqueued(job_id):
    frappe.enqueue(
        "myapp.tasks.process",
        job_id=job_id,
        doc_name=doc.name
    )

v14 (Deprecated)

# DO NOT USE - only for legacy code
from frappe.core.page.background_jobs.background_jobs import get_info
enqueued = [d.get("job_name") for d in get_info()]
if name not in enqueued:
    frappe.enqueue(..., job_name=name)

Error Handling Pattern

def process_records(records):
    for record in records:
        try:
            process_single(record)
            frappe.db.commit()  # Commit per success
        except Exception:
            frappe.db.rollback()  # Rollback on error
            frappe.log_error(
                frappe.get_traceback(),
                f"Process Error: {record}"
            )

Callbacks

def on_success_handler(job, connection, result, *args, **kwargs):
    frappe.publish_realtime("show_alert", {"message": "Done!"})

def on_failure_handler(job, connection, type, value, traceback):
    frappe.log_error(f"Job {job.id} failed: {value}")

frappe.enqueue(
    "myapp.tasks.risky_task",
    on_success=on_success_handler,
    on_failure=on_failure_handler
)

User Context

IMPORTANT: Scheduler jobs run as Administrator!

def scheduled_task():
    # frappe.session.user = "Administrator"

    # Set explicit owner:
    doc = frappe.new_doc("ToDo")
    doc.owner = "user@example.com"
    doc.insert(ignore_permissions=True)

Monitoring

Version Differences v14 vs v15

Reference Files

• scheduler-events.md: All event types, cron syntax, configuration
• enqueue-api.md: Complete frappe.enqueue/enqueue_doc API
• queues.md: Queue types, timeouts, custom queues, workers
• examples.md: Complete working examples
• anti-patterns.md: Common mistakes and corrections

Critical Rules

1. ALWAYS bench migrate after hooks.py scheduler_events changes
2. USE job_id + is_job_enqueued() for deduplication (v15)
3. CHOOSE correct queue: short/default/long based on duration
4. COMMIT per successful record, rollback on error
5. REMEMBER that jobs run as Administrator
6. ENQUEUE heavy tasks from scheduler events, don't execute directly