frappe-impl-scheduler（frappe impl 调度程序）

Frappe Scheduler & Background Jobs - Implementation

Workflow for implementing scheduled tasks and background jobs. For exact syntax, see frappe-syntax-scheduler.

Version: v14/v15/v16 compatible

Main Decision: scheduler_events vs frappe.enqueue

WHAT ARE YOU BUILDING?
|
+-- Runs at fixed intervals/times?
|   +-- YES --> scheduler_events (hooks.py)
|   |           Task receives NO arguments
|   |           See: Workflow 1-2
|   |
|   +-- NO --> Triggered by user action or code?
|              +-- YES --> frappe.enqueue()
|              |           Pass any serializable data
|              |           See: Workflow 3-4
|              |
|              +-- NO --> Reconsider requirements

Which Scheduler Event Type?

Rule: ALWAYS use *_long suffix for tasks exceeding 5 minutes.

Which Queue for frappe.enqueue?

Rule: ALWAYS specify queue= explicitly. NEVER rely on the default.

Implementation Step 1: Scheduler Event

# myapp/tasks.py
import frappe

def daily_cleanup():
    """Daily cleanup - NO parameters allowed."""
    cutoff = frappe.utils.add_days(frappe.utils.nowdate(), -30)
    frappe.db.delete("Error Log", {"creation": ("<", cutoff)})
    frappe.db.commit()

# hooks.py
scheduler_events = {
    "daily": ["myapp.tasks.daily_cleanup"]
}

After editing hooks.py: ALWAYS run bench migrate.

Implementation Step 2: Background Job (frappe.enqueue)

# myapp/api.py
import frappe
from frappe.utils.background_jobs import is_job_enqueued

@frappe.whitelist()
def process_documents(doctype, filters):
    job_id = f"process_{doctype}_{frappe.session.user}"

    if is_job_enqueued(job_id):
        return {"message": "Already in progress"}

    frappe.enqueue(
        "myapp.tasks.process_batch",
        queue="long",
        timeout=1800,
        job_id=job_id,
        enqueue_after_commit=True,
        doctype=doctype,
        filters=filters
    )
    return {"status": "queued"}

Testing Scheduled Tasks

Method 1: bench execute (direct)

# Run the function directly (no queue involved)
bench --site mysite execute myapp.tasks.daily_cleanup

Method 2: bench scheduler (full scheduler test)

# Check scheduler status
bench --site mysite scheduler status

# Enable scheduler
bench --site mysite scheduler enable

# Trigger all pending scheduler events NOW
bench --site mysite scheduler trigger

# Run specific event type
bench --site mysite execute frappe.utils.scheduler.trigger --args "['daily']"

Method 3: bench console (interactive)

bench --site mysite console
>>> frappe.enqueue("myapp.tasks.my_task", queue="short", now=True)
# now=True executes synchronously for testing

Method 4: Check Scheduled Job Type

1. Go to: Setup > Scheduled Job Type
2. Find: myapp.tasks.daily_cleanup
3. Verify: Frequency correct, Stopped = No
4. Click "Run Now" to trigger manually

Monitoring

Scheduled Job Log (UI)

Setup > Scheduled Job Log
- Shows every scheduler run with status
- Filter by: status (Success/Failed), creation date
- Check execution time to detect slow tasks

RQ Dashboard

# Start RQ monitor (development)
bench --site mysite rq-dashboard
# Opens at http://localhost:9181

# Show background job status
bench --site mysite show-pending-jobs
bench --site mysite show-failed-jobs

Programmatic Health Check

def scheduler_health_check():
    failed = frappe.db.count("Scheduled Job Log", {
        "status": "Failed",
        "creation": [">=", frappe.utils.add_to_date(None, hours=-1)]
    })
    if failed > 5:
        frappe.sendmail(
            recipients=["admin@example.com"],
            subject="Scheduler Alert: Many failures",
            message=f"{failed} scheduler jobs failed in last hour"
        )

Error Handling in Scheduled Tasks

Per-Record Error Isolation

def sync_all_orders():
    orders = get_pending_orders()
    success, errors = 0, 0

    for order in orders:
        try:
            sync_to_external(order)
            success += 1
        except Exception as e:
            errors += 1
            frappe.db.rollback()
            frappe.log_error(
                f"Sync failed for {order}: {e}",
                "Order Sync Error"
            )
    frappe.db.commit()
    frappe.logger("sync").info(f"{success} ok, {errors} errors")

Rule: ALWAYS wrap per-record processing in try-except. NEVER let one failure stop the entire batch.

Long-Running Job Patterns

Self-Chaining Pattern (>25 min tasks)

def process_batch(offset=0, batch_size=500, total=None):
    if total is None:
        total = frappe.db.count("Sales Invoice", {"custom_processed": 0})

    records = frappe.get_all("Sales Invoice",
        filters={"custom_processed": 0},
        pluck="name", limit=batch_size)

    if not records:
        return  # Done

    for name in records:
        process_single(name)
    frappe.db.commit()

    remaining = frappe.db.count("Sales Invoice", {"custom_processed": 0})
    if remaining > 0:
        frappe.enqueue(
            "myapp.tasks.process_batch",
            queue="long",
            offset=offset + batch_size,
            batch_size=batch_size,
            total=total
        )

Rule: ALWAYS split tasks >25 min into self-chaining batches.

Common Implementation Patterns

Email Digest (weekly summary)

# hooks.py
scheduler_events = {
    "cron": {
        "0 8 * * 1": ["myapp.newsletter.send_weekly_digest"]
    }
}

See references/examples.md Example 4 for complete implementation.

Data Cleanup (daily maintenance)

scheduler_events = {
    "daily_long": ["myapp.maintenance.daily_database_maintenance"]
}

See references/examples.md Example 1 for batch deletion pattern.

Report Generation (user-triggered)

frappe.enqueue(
    "myapp.tasks.generate_report",
    queue="long",
    timeout=3600,
    job_id=f"report::{frappe.session.user}",
    user=frappe.session.user
)

See references/workflows.md Workflow 6 for progress reporting.

Critical Rules

1. Scheduler tasks receive NO arguments - Use settings or hardcoded values
2. ALWAYS bench migrate after hooks.py changes - Required to register events
3. Jobs run as Administrator - ALWAYS commit explicitly
4. Commit in batches - NEVER per-record (every 100-500 records)
5. ALWAYS use job_id for user-triggered jobs - Prevents duplicates
6. Use enqueue_after_commit=True from document events - Ensures data exists
7. Scheduler events should be thin - Enqueue heavy work to background

Version Differences

Reference Files

See Also

• frappe-syntax-scheduler - Exact syntax reference for hooks and enqueue
• frappe-errors-serverscripts - Error handling patterns
• frappe-impl-hooks - Hook configuration patterns
• frappe-ops-bench - Bench commands for scheduler management
• frappe-ops-performance - Performance tuning for background jobs
• frappe-testing-unit - Testing scheduled task logic